"Real men have Fabs" - Jerry Saunders, former AMD CEO.
After all the dodging and lying about not having any plans going fabless, AMD finally admits that it is in fact going to sell its fabs. They intend to remain a minority share holder and team up with a company who haven’t got a single clue about semiconductors. But none of that matters because they have lots of money to spend. And in a time when the usual lenders are looking for a some kind of a bailout themselves, it’s either this arrangement or Chapter 11 for AMD.
"On Oct. 7, 2008, AMD and the Advanced Technology Investment Company announced the intention to create a new global enterprise, The Foundry Company, to address the growing global demand for independent, leading-edge semiconductor manufacturing. There is a strong shift to foundries occurring – particularly to foundries with the capacity to produce devices using leading-edge process technologies. With The Foundry Company, AMD will be able to unlock the value of its world-class manufacturing capability – by making it available to a growing community of fabless semiconductor companies. " - AMD's New Global Foundry page.
There's a lot already said about AMD's move. The best one that highlights the concerns more clearly is from Fabtech:
Another aspect that concerned me was the notion pumped out by AMD that demand for leading-edge foundry capacity was something that was in strong demand. Ask SMIC, Chartered or UMC how much of its capacity is allocated to 65nm-and-below production and you will find it is very small... Also, if demand for foundry capacity was that strong, then why are wafer ASPs in decline and the major foundries cutting CapEx each year?... Even worse is the fact that SMIC has struggled since birth to actually turn a profit, so why should we think that a new foundry start-up in Europe (and the U.S.) will fare any better?
The "growing trend" to go fabless is a decision forced upon companies due to the rising cost of running a Fab. This alarming trend have been identified by Intel in the late 90's. AMD knew this day was coming and set a goal of 30% market share just to avoid ending up where it is now. It's entertaining to see how AMD is making it all appear as if it was an advantageous choice.
How can it be advantageous when AMD will now have to ask another company to spend $Billions every time it wants new equipment for a new process technology. How can it not be disadvantageous when your main competitor do not have such bureaucratic problems. Short term, this move by AMD will buy them time for the next year or two. Long term, I can only see AMD becoming the next Transmeta. When The Foundry Company starts losing large sums of money they will become cost conscious and that's when problems begin. Against Intel's "tick-tock" execution, AMD will only struggle to keep up.
10.08.2008
AMD comes out the closet, says no longer a real man
Subscribe to:
Post Comments (Atom)
236 comments:
1 – 200 of 236 Newer› Newest»http://scientiasblog.blogspot.com/2008/04/amds-asset-smart-explained.html
So how close was Scientia's explanation about asset-smart?
Good to hear from you again Robo, although things had been pretty boring.
I find it telling that even with the large demands that Barcelona put on fab capacity that AMD apparently STILL couldn't keep its fabs running fulltime and needs to get outside contracts to keep things humming. That's the buried story here and it is not good for AMD.
AMD goes from being a parasitic 2nd tier chipmaker to join the rest of the fabless ranks, long overdue. Their shareholders deserve all the losses, they somehow find a reason to stick with that POS.
On a different note, the retarded Scientia never made sense when it came to chip manufacturing (read his article on low-k vs high-k for another example). Scientia is only relevant for other retards on AMD zone.
Tick Tock Tick Tock
The clock really has run out. Yes anonymous poster will start talking about Lex or something..
But this annoucement forever relegates AMD down the path of never having a leadership position in high performance CPUs. They perhaps have a chance to compete with Nvidia and more probably then not beat Nvidia for a while. But once the Arabs realize that to make money trying to compete directly with INTEL and the bleeding edge is simply flushing money in the toilet they will change their model more and more to look like TSMC and more likely then not will look like SMIC. Sink billions into a fab and find no customers.
WTF are they thinking. As Robo noted look at TSMC and see how much volume they have at 90 and 65nm and INTEL is already ramping 45nm. The cost of mask alone at 32nm will price all but the very very highest volume or the highest ASP devices out of the node. AMD has neither volume nor ASP so soon the Arabs will be tweeking the process for the Qualcomms of the world and AMD will find themselves further and further behind on process.
The clock has run out. Reminds me of the Last Emperor where the servants sat there with their manhood in a jar. Dick and Hector just handed theirs to the Arabs of all people. They could have tried to keep at least one ball and gave it to some bankers but in this day and age there are no bankers with balls either.
AMD is finished
rob: "They intend to remain a minority share holder and team up with a company who haven’t got a single clue about semiconductors. But none of that matters because they have lots of money to spend."
I couldn't help but think that there is a bit of a parallel with the current economic crisis, where financial companies were allowed to stick their necks way out (in some cases, almost forced to) because the government would be there to bail them out. The government may not know much about finances, but they can print money!
The problem is, when you use the safety net, it means you failed at what you were attempting to do.
But this may have saved AMD the company. Is there room in the market for another VIA? A company that produces low-end CPUs (and high end GPUs) and is profitable doing so? Does anyone think that AMD cannot do this? I am not sure how the Foundry Co would handle this, though. They would have to drift apart and become Just Another Foundry, wouldn't they?
This just gives AMD the chance to survive in hopes of catching Intel asleep 3 years from now while trying to design another comeback chip and hoping the lawsuit brings some money.
On 45nm, AMD seems to be making a big deal about DFM (design for manufacturing) - something every IC manufacturer has been doing, except they give it a name (like APM) and make it sound innovative. These design rules come about from close interaction between the process and integration folks and the design folks with each of them having a reasonable understanding of the impact of what they own has on the other guy.
If I may be allowed to provide an alternative view. The PCB shop I worked for did not produce any original designs. Everything we made was designed by the customers.
Here is how it worked. We published a list of design specifications. They consisted of line widths, line spacings, hole diameters/aspect ratios, board thickness, etc. Basically, if you could measure a product parameter and tie it to a process capability, we did.
The customers took this list and compared it to their design. If our capability met their design criterion, they submitted a request for a bid from us. If our bid was satisfactory, then we got a contract and we built a board for them. (This is over-simplistic, but it captures the essence.)
Often the design had issues that complicated the manufacture of the board and we would request a variance (read design change). This was often denied, and we had to struggle with making something the process was only marginally capable of producing.
The need to show cutting edge capability (and get the most profitable contracts) often led to overstating our process capability. As far as I know, the company never outright lied, but being able to do something and being able to do it consistently and reproducibly are not the same thing. So we often got products that were on the hairy edge of what we could produce.
But produce stuff we did, and all without integrated DFM between the designers and the manufacturers.
Now I'm not trying to tell you that making a PCB and making a microchip require the same degree of integration, they don't. The chip requires much more complex integration, but I don't imagine the system is a whole lot different.
The foundry will give a set of process capabilities and the designers will have to design within those capabilities. There will be very little optimization between design and process. You will run through a couple of revs until you get something that is "good enough" and off to market it goes.
I suspect this is why you see one of the oddest parts of the arrangement. AMD, which is now a design house, remains part of a manufacturing consortium. I have to believe that this is an attempt to ensure that their designers are more closely tied into the process capabilities and take those capabilities into account.
I don't believe that this will work as well as having manufacturing and design under the same roof, but I do think it will be better than nothing.
roborat64: How can it be advantageous when AMD will now have to ask another company to spend $Billions every time it wants new equipment for a new process technology. How can it not be disadvantageous when your main competitor do not have such bureaucratic problems. Short term, this move by AMD will buy them time for the next year or two. Long term, I can only see AMD becoming the next Transmeta. When The Foundry Company starts losing large sums of money they will become cost conscious and that's when problems begin. Against Intel's "tick-tock" execution, AMD will only struggle to keep up.
AMD might get the benefit of capital investment without all the disadvantages mentioned above. Allow me to give another (conspiracy) theory :)
[conspiracy start]
Under the cross licensing terms, AMD is not be allowed to be owned by other entity (50%):
quote from http://contracts.corporate.findlaw.com/agreements/amd/intel.license.2001.01.01.html
(7) the other party undergoes a Change of Control. For
purposes of this Section 6.2(b)(7), "Change of Control"
shall mean a transaction or a series of related transactions in which (i) one or more related parties who did not previously own at least a fifty percent (50%) interest in a party to this Agreement obtain at least a fifty percent (50%) interest in such party, and, in the reasonable business judgment of the other party to this Agreement, such change in ownership will have a material effect on the other party's business, or (ii) a party acquires, by merger, acquisition of assets or otherwise, all or any portion of another legal entity such that either the assets or market value of such party after the close of such transaction are greater than one and one third (1 1/3) of the assets or market value of such party prior to such transaction.
to workaround this(not sure entirely legal or not), AMD split into 2, with ATIC owning 55.6% of the Foundry Company. ATIC will invest $2.1 billion (ignore the later investment 3.6B from this calculation, as it might be in the form of loan, etc) in Foundry and Foundry will assume about $1.2 billion, making ATIC paying 2.1B + 1.2(x55.6%) = 2.77B for this. This is already more that what AMD market cap is (today its Market cap is 2.53B). Even with 20% premium, it will still making ATIC as >50% ownership, since it has already owned 8% of AMD today.
put that amount of money into AMD would violate the above clause (7). So they 'smart'ly splitting the company into 2 to avoid direct cofrontation with that clause ...hey, this might explain what asset-smart really means!! :) hahahaha
So, to the outsider, they are 2 companies, but internally they really act as one.
[conspiracy end]
ok, allow me to laugh again ... hahahaha because i have tried defining the asset smart ... but i am smart enough to define it after knowing more details ... not like some over the amdzone ... his definition is sooo off :)
Anonymous said...
http://scientiasblog.blogspot.com/2008/04/amds-asset-smart-explained.html
So how close was Scientia's explanation about asset-smart?
ROFL! Not very...
One would think that at least once in a while, ol' Sci would randomly hit upon a correct guess. I posted on here a list of about half a dozen zingers by him made in the last couple of years, so make this # 7 :).
BTW, I just have to laugh when Sci and Abi go after each other, each stating the other can't read or comprehend, etc etc. However I would have to say, on average, Abi is doing better with his predictions (only 6 zingers to date :). Although some of his were pretty hilarious as well...
DOC-
“How can it be advantageous when AMD will now have to ask another company to spend $Billions every time it wants new equipment for a new process technology.”
Far be it my place to suggest a re-wording of your excellent and flawless piece. But if I may be so bold as to humbly suggest, a slight revision, or rather, modification to your statement above.
May suggest the following, “”it wants new equipment for a new ‘unproven’ process technology”.
After all, they did bet the farm on ‘unproven’ SOI. (And lost half of it)
I know this is splitting hairs, but the final analysis, it could make a world of difference for the new ‘partnership’ to work. Indeed, the AMD/IBM partnership hasn’t developed anything revolutionary, or commercially successful as of late, and it hasn’t cost IBM a dime for AMD’s overall(tools inclusive), SOI disaster.
Conversely, AMD’s new, or shall I say additional, partner is indeed footing the bill on every ‘misstep’ AMD (and indirectly IBM) makes now, and in the foreseeable future.
BTW: “Clueless”, excellent!
Truly Yours,
SPARKS
So what happens when AMD's equity stake starts dropping as more and more capital is invested in the foundry over the next few years? Or is AMD really going to keep up their 44% stake as the majority investor pumps in 3-6Bil over the nesxt few years?
And if it drops will they continue to control 50% of the board?
I don't see the end game - you have a majority investor who knows nothing about the business and who's only contribution is money. What happens when AMD blows the trust fund they inherited? Hope they have another rich uncle? Hope that the majority investor who's stake keeps growing and is carrying the majority of the losses thinks it is wise to shoulder the financial burden but allow the minority owner to control the direction of the company?
There was an interview with AMD about the foundry aspect (I'll look for the link). Basically it seems like AMD is targeting 32nm for the main foundry (non-AMD work) business and they made it sound like for the next 1-2 years the work will be mainly AMD only.
The problem I see is the demand is actually soft on the leading edge nodes - there is minimal 45nm foundry demand compared to the older nodes. If there is soft demand for 45nm now what makes AMD think 32nm will be in demand in 2 years - my guess is that is when you will see designs migrating from 65 to 45nm. AMD doesn't seem to understand outside of CPU's and memory - the leading edge node isn't really the main foundry driver. It doesn't make sense for design chops to update designs to what will be a more expensive process and incur the mask/development costs unless they are doing significant volume or have a specific technology need.
AMD is going BK, who was that monkey talking about INTEL BK in 2006, no 2007, no 2008, no 2009.
In the end is is AMD going BK
Ahhhh.... another insightful comment.
http://www.thestreet.com/story/10441929/1/amds-ex-ceo-set-for-nice-payout.html?puc=googlefi&cm_ven=GOOGLEFI&cm_cat=FREE&cm_ite=NA
Hector gets 1.15Mil/year from the foundry, with a bonus opportunity of another 400%.
And the kicker? A "special" bonus of $3Mil for the spinoff deal. Putting aside the salaries, this special bonus is ridiculous. I would ask who's paying it but it doesn't matter - you have 2 theoretically separate companies that are just operating as one, so what's the difference.
It is amazing how much Ruiz continues to soak out of AMD... Get paid a lot, do a bad job for the owners (stockholders), force parts of the company to be sold off, get a bonus for doing so and then become chairman of the new company when you've finally worn out your welcome in the old one, and start the cycle again. I must say the guy in this sense is brilliant.
“AMD is finished”
Not so fast.
Also…….
“I must say the guy in this sense is brilliant.”
Oh brother, how true.
I am reluctant to day this but AMD has, considering the number of monumental missteps and execution failures, done pretty well. True, their vitals may be in a jar of formaldehyde, but they are far from facing bankruptcy. In fact, they’re still very much alive and kicking, make no mistake.
Don’t get me wrong. I don’t like the sons-of-bitches, however, the fact still remains they are here to fight another day. You have to give them credit. They parlayed this thing really well, (3 card Monte) and they are terrific at operating in survival mode. Given the current day economic stress when the Lehman’s, et al, are belly up floaters, you simply can’t kill these cockroaches even when their debt is twice their value!
You guys in the industry are familiar with close integration of process and architecture. Naturally, I don’t have a clue. I am as far from an industry expert as, say, Andromeda, but if you guys say it ain’t gonna fly, I’m board with ya, and we’ll see how it pans out in the long term. It will give us some food for fodder in the meantime. And the best thing is, you boys know just how the shit flies BEFORE it hits the fan
Personally, AMD’s continuing saga has been absolutely fascinating. Especially when INTC points a finger and declares, “Let the castrations begin!”, “Release a new lineup and drop the prices here and here”. Frankly, I’m enjoying the daily daytime drama.
All said, BK from AMD’s perspective was NOT an option, and they did get to keep their testicles, albeit in a jar. Further, they are taking the A-RABS for billions, Christ, it’s about time someone found a way, no? Thousands of people kept their jobs, and finally, Captain Ahab (Wrector) just may live to see his Big White Whale in court come 2010, just in time for x86 license renegotiations.
And……………………
“Hector gets 1.15Mil/year from the foundry, with a bonus opportunity of another 400%.”
“G”, WHERE’S THE KING? WHERE’S THE KING?
SPARKS
Whoops! Ah, "thats ON board with ya"
I could never be board with you guys, even LEX!
xoxo
SPARKS
Anonymous said...
AMD is going BK, who was that monkey talking about INTEL BK in 2006, no 2007, no 2008, no 2009.
Sharikook is busy fixing his old computer, which he busted 3X by upgrading to an X3 :).
Or maybe he said "boosted" instead. Anyway, his blog is about dead & buried, only took 2 years after his credibility died & was interred :).
Don't like quoting Fuddie, but this typo is priceless:
We got a confirmation from sources close to company that The foundry co. should shit to 32nm High-K process in 2010.
Maybe 'shit' wasn't the typo and the typo was "on" instead of "to"
http://www.fudzilla.com/index.php?option=com_content&task=view&id=9921&Itemid=1
Of course 45nm coming out end of 08 (kinda/sorta) add two years and you get end of '10 for 32nm (unless you think when AMD says 2010 they mean beginning 2010?!?!?)
Needless to say what this means for the "option" of high K on 45nm...
anon: "(unless you think when AMD says 2010 they mean beginning 2010?!?!?)"
Is that when they finally move ahead of Intel on process tech? :)
Well at least Fuddie reads his own articles the typo has been corrected (or maybe he reads this blog for information?)
Q3 #'s are in
EPS - $0.35 vs $0.34 street forecast (up 12% year on year)
Revenue $10.2Bil, up 1% year on year
Q4 revenue forecast was a bit light though:
10.1-10.9Bil vs street forecast of 10.4-12.2Bil
AMD announces Thursday. Street consensus is $-0.40 EPS on 1.478Bil revenue
AMD announces Thursday. Street consensus is $-0.40 EPS on 1.478Bil revenue
If that holds then 40 cents per share times 607 million shares = a loss of 240 million. If we could only sell another 699 AMDs to the petro-rich states, we would about break even on what we sent them each year in petro-dollars..
BTW, Sci hasn't been seen on UAEZone since the beans were spilled last week on what asset-light really meant, nor has he updated his blog. I suppose he is hiding in a hole somewhere, too embarrassed to poke his head out and get laughed at :).
What will be interesting on Thursday is if AMD stands behind their operating income break even by the end of the year promise. Given the current economic environment this would seem to be difficult (unless there is creative accounting which turns one time proceeds into operating revenue like AMD did with the 200mm equipment sales).
Regardless, the spinoff probably buys them a complete pass on the earnings #'s, unless they are WAY off. My guess is that Wall St may also discount the Q4 #'s as well (in terms of caring/not caring what the EPS and revenue #'s are).
At this point, >1 year after K10 "launch", it would seem whatever ASP bump AMD should have gotten would have been realized (or not realized) by now. With 45nm having no impact on the Q4 (and probably Q1) #'s, I don't see the breakeven happening unless AMD sees a serious uptick in unit sales.
Impact you say? Let’s see what Arab Micro Devices are up against in 2009.
Someone once said the only way they’d believe in the existence Alien races or civilizations would be an interview on The Jay Leno Show. I agree. Nothing illustrates the perception of reality more than mainstream TV, just before the Viagra commercial, immediately after the Massengill commercial.
So, where is that lunatic Sparks going with all this? Well. I’ll tell you, its computer hardware and all the bias nonsense surrounding it.
My biggest, most reliable source of unvarnished hardware reviews and testing is MaximunPC. Web reviews, by the usual suspects, are flawed by agendas, be it ‘advertising agreements’ or reader biases. Not so with MaximumPC. Want proof? Just look at the number of advertisers in every issue, at the very least, they don’t have friends. However, if a company gets a Kickass Award, you can bet your ass the award will be emblazoned on their products retail packaging.
No Alien conspiracy theories here, just the signature of great hardware. If it ain’t in MaximunPC, and it ain’t got a Kickass Award, I ain’t a believer.
Decembers issue’s point of interest was Nehalem. Talk about kickass? My QX9770 WILL have its ass handed to it come November, bank on it. (I know some of you industry boys already know this and are sitting back grinning like Cheshire Cats. I am not as privileged.) Incidentally, I’ve got a sneaking suspicion INTC has a soft spot for the boys at MaximumPC. They usually get the first ES samples, lucky bastards, especially for the year end ‘Ultimate KiskAss Rigs’.
May I suggest a full read in December’s issue, but, here are some highlights.
Nehalem is bringing a lot to the table.
INTC SANTIONED OVERCLOCKING!!!! In the past even mentioning the “O.C.” word within an eye shot of INTC’s Head Quarters was grounds for recrimination, is now AUTOMATICLLY SUPPORTED! Huh? Yes “Turbo Mode” ala 386, 486 is back and on steroids. Basically over clocking one core, turning down the rest if they aren’t utilized! Hell, if I could do this QX9770, I could clock just like ‘Giant’s’ E8500 (little sweetie) at 4.5 Gig!
MaximumPC made it very clear IMC was NOT an AMD invention. They even mentioned the brief history of Timna. Ah, back to reality.
Mega-parallelism unseen before in previous designs.
Bandwidth, bandwidth, and more bandwidth.
Hyper-Threading, revisited, this is not your fathers Hyper-Threading. Unlike P4, this iteration will be fully optimized. INTC claims 10 to 30%! (I left my 3.06 P4’s HT always enabled, why should I stop now?)
Improved Loop Detector Routines to detect larger loops, storing them in cache.
Polished branch prediction, to avoid pipeline flushes, increasing cache hits.
Power management up the kazoo, cache voltage is maintained separately, while core voltage can be reduced, or INCREASED (within INTC’s predefined limits, no doubt) if needed! HOO YA! This is all on the fly!
Did I mention scaling?
2008 was the great server assault (2P), 2009 will be the great HPC assault.
In short, Arab Micro Devices should be afraid, very afraid. The Perfect Storm continues. Break even my ass.
SPARKS
Sparks,
From the vicious rumors department, I've heard rumblings that the Nehalem issues will be in the chipsets. Supposedly that is why only the enthusiast part is debuting this year.
I have no idea if this is true or not, but it is what I'm hearing.
Assuming the rumors are true, I have to wonder if this isn't due in part to issues with repartitioning the chipset. The Nehalem chipset is only comprised of one chip as I understand it due to the movement of the memory controller onto the CPU and replacement of FSB with Quickpath.
And unless I'm not mistaken, Intel wasn't the first with IMC either. IBM's Power2 chip had this back in 1996. But the point remains. It wasn't invented by AMD.
I'm a little skeptical of Turbo mode. While it may be useful for Joe sixpack, for folks buying the enthusiast chip (and assuming they are true enthusiasts) - they are still probably better off OC'in on their own.
I do think it will take the architecture a little while to shine as I don't think SW is caught up enough to take advantage of it. Early on you may hear the naysayers talking about less than expected improvements, but I think SW will start to catch up and better utilize the capabilities of the chip.
This is also why I think it will shine more in server land early on (in addition to the whole IMC/QPI thing). The real question will be can Intel get its graphics in order to make an integrated chip as well as having a competitive non-discrete graphics card solution to compete with AMD
Earnings call transcript:
http://seekingalpha.com/article/99913-intel-q3-2008-earnings-call-transcript?source=yahoo
Some highlights:
- The atom "failure" accounted for 200Million in revenue in Q3 (to put this in perspective this would be ~1/7 to 1/8 of all of AMD's revenue). If AMD had a few 'failures' like this, they would probably be profitable.
- Otellini also said they couldn't keep up with Atom demand (and won't until end Q4/Q1'09) and expect Q4 atom revenue to be up. Also any cannibilization of Celeron sales due to the atom netbook actually results in a higher margin! (Meaning even though ASP's of atom are lower, Intel earns a much better margin compared to Celeron)
- shipments of Nehalem began in Q3
- Mobile is now 45% of total revenue and was up 23% from Q2.
- ASP's were slightly down, but if you exclude atom they were flat.
- The atom "failure" accounted for 200Million in revenue in Q3 (to put this in perspective this would be ~1/7 to 1/8 of all of AMD's revenue).
That probably explains why Ah-Ben-Stoopid deserves his moniker..
Sheesh, what an epic failure that guy is - almost as bad as Asset Smart know-it-all Scientia :).
“I'm a little skeptical of Turbo mode.”
See, that’s what I’m talking about, doubts. No Aliens from Zeta Reticula for you.
Turbo mode is for the “six packers”, no doubt. None of these guys will have the Fluke meters, LN2, and scopes tweaking vCore Reg. voltage peaks, of course. But if INTC wants to bring the masses into the fray, hey, more power to ‘em. ‘Everyone will be famous for five minutes.’ Call it, “Over Clocking for Dummies”.
ITK-
Yes, I’ve heard some FUD from FUD that speculated that if you raise the memory voltage above its default voltage, you take the chance of blowing the processor. I don’t know about you guys, but my SuperTalent DDR3’s recommended voltage is 2.1V@1866MHz . I wouldn’t throw that kind of vCore voltage at QX9770 with gun to my head. Ok, maybe a BIG gun, and a hundred pounds of LN2, I would.
Ludicrous, I say. That’s an awful lot of performance to take off the table after hundreds of millions in R+D. Further, INTC’s memory partners wouldn’t be too happy if INTC’s latest and greatest is stuck in the DDR3 bread and butter gulag at 1066. ‘I Believe.’
DDR3 doesn’t come into its own until you’re in say, at the very least, the 1333MHz bracket. This I know. Strapping to the FSB and humping the clock has taken me to the manufacturers recommended speed of 1866. And, it will go higher, close to 2 GIG, in fact. (7-7-7-21 2T). Corsair already has modules rated at 2133! Ah, that’s 2X 1066 for all those mathematically challenged. Ci7 Extreme not using high speed DDR3? ‘Danger, Wil’ Robinson.’
http://www.corsair.com/_datasheets/TW3X2G2133C9DF.pdf
True, the game may change with Tri-Channel, but at 1066? Moon bases on the dark side of the Moon? No way. I think it’s all a bunch of crap. Some of these guys are saying INTC is being very hush, hush. I’m sure they are. Why the hell would they cannibalize C2Q sales when a mainstream Nehalem will kick my $1500 overclocking beast to the curb? It will happen with the Extreme cut, no doubt, but not before Ci7 hits the shelves in earnest. Besides, you guys will be the first to know how well Nehalem will do against QX9770 @ 4 GIG, trust me.
Bottom line, when MaximumPC says (hints) that Ci7 will be a rocket, you can bet your MHz it will.
‘The Truth Is Out There.’
SPARKS
Still have doubts about Ci7?
Read this:
"Documents from Intel Corp. allegedly seen by a web-site earlier this month reportedly mention that Intel Core i7 processor 965 (3.0GHz) delivers 52% speed improvement in 3D games, 41% higher performance in media encoding and editing applications as well as 38% greater 3D rendering speed when compared to Intel Core 2 Extreme QX9770 (3.0GHz). If the information regarding performance improvements is correct, then it means that the new micro-architecture in its first implementation provides roughly 52% more performance at the same clock-speed compared to currently used Core 2 micro-architecture."
This thing is going to make me look like I'm going backwards.
Arab Micro Devices be afraid, very afraid.
http://www.xbitlabs.com/news/cpu/display/20081014223031_Intel_Initiates_Shipments_of_Core_i7_Microprocessors.html
SPARKS
AMD #'s are in - generally pretty decent, though there are a bunch of asterisks...
Revenue - 1.78Mil, Loss - 67Mil, -$0.10 EPS
Operating Income** - 131Mil, $0.07EPS
However the Revenues were aided by a 191Mil Technology license fee related to the sale of the 200mm equip to a Russia company last quarter.
That said without the license fee it was -$0.24EPS, which is still ahead of expectations of -0.40 (but would not meet "operating income" promises)
And there's the spin for the 4th quarter. Rivet was "pleased to have reached our goal of operating profitability this quarter". This of course included the 191Mil lincense fee..... however when he guided Q4 he said he expected revenue to be flat.
The problem?
- AMD used 1.78Bil for quarterly revenue (this included the 191Mil licensing fee). They used this # to compare to previous quarter and YoY, as well as for operating income and gross margin. This is a gray area but is a pretty sketchy way to account for the additional revenue.
- AMD then forecasted 1.58Bil revenue for Q4 - which they called "flat". So 1.78Bil Q3 to 1.58Bil Q4 is "flat"?
So, when comparing to previous quarters they use the 191Mil to make the #'s look better, and when they forecast Q4 they take it out to set the bar lower! That is VERY slimy and completely inconsistent.
Microprocessor units were down YoY (Q3'07), and ASP's were flat... yet "revenue" was magically up 8%!
So if you sell fewer units, and at the same price how does revenue go up? Good question - sneak in a $191Mil license fee and call it CPU revenue.
So what happens if you take it out? CPU revenue was ACTUALLY DOWN 6.5% YoY. But that of course does not paint as pretty a picture!
Graphics revenue was up 40% (additional 110Mil) and operating in the black (+47Mil). That and the 191Mil license fee was the reason for the good #'s - but AMD did a fair job disguising this. It will be interesting to watch the spin in Q4 when there is no 191Mil to conveniently hide into the CPU #'s to pump it up.
Ooops - here's the link to the details if folks care:
http://biz.yahoo.com/bw/081016/20081016006192.html?.v=1
Well, so much for Alien abductees/conspiracy theories and the like. Clearly, that idiot Fudzilla with his ridiculous reports concerning Ci7 memory voltage/X58 issues were completely unfounded.
The usual suspects have released a gaggle of DDR3 Tri-Channel memory choices for X58, up to and including DDR3-1600. (Ah, do I have to say these are not for Nvidiot?)
http://www.techpowerup.com/index.php?72926
A-Data's recommended retail price for the DDR3 1600 kits is $230 for the 3GB models (3x1GB), and $350 for the 6GB (3x2GB) set. CHEAP!
That said they run at a substantially lower voltage, 1.65v. Still, that’s way above 965’s Vcore.
Sure, INTC really screwed up X58 this time. They’ll be bankrupt next week.
Wait! Fuddie, I think I see a cigar shaped object over your website!
SPARKS
Speaking of the mindless drivel coming from the Fudzilla website, I offer this jewel.
" Intel obviously wants to stay away from SOI as it doesn’t like the burden that this process brings, but AMD traditionally has much better yields than Intel as AMD has only one and sometimes two operational fabs and needs to get the maximum out of it."
Will someone explain to me how needing something to be so makes it that way?
The people on the Titanic needed the ship to have been made from steel with a different temper. They needed more lifeboats. They needed the other ship in the area to come rescue them. But no matter how badly they needed these things, it just didn't happen.
Wow, the spin with the "Process Technology Revenue" is rich. Not only is this worse than the Saxony Germany subsidies which are only applied to cost of sales, they are applying the sale of old equipment to the Top line as CPU shipments.
they are applying the sale of old equipment to the Top line as CPU shipments.
Yes, that's what I was trying to say in a long-winded way in the comment above. If you strip it out CPU revenue is down 6.5% year on year and there is no "operating" profit.
The sad thing is how careful AMD is to strip the # out when they project Q4 sales to be flat! Shameless!
Earnings call transcript for AMD conference call:
http://seekingalpha.com/article/100358-advanced-micro-devices-inc-q3-2008-earnings-call-transcript?source=yahoo&page=8
A couple of interesting points and "creative accounting" and some actually decent questions:
I got a question on the balance sheet. You might have gone through this, Bob, I apologize if you did. The cash fell quarter-over-quarter given that you got the 191 million that came in I was a little surprised to see that. Could you describe what happened there?
Bob Rivet
Sure. One of the phenomena’s on the 191 is actually we collected the majority of the 191 in prior quarters. They were all, I will call it prepaid. It was until we completed the revenue process in the quarter that we could actually record it. So, the 191 did not show up just in the third quarter. It actually showed up throughout this year, which is how we manage our cash balance throughout the year. So, that is one of the phenomenas of why cash went down even though I had this nice pop in revenue particularly from a process technology license.
So basically they were collecting the 191mil license fee throughout the year but claimed it in Q3 (conveniently when they targeted/promise breakeven). It also explains the question I had which is why did their cash go down when they were supposedly profitable!?!? Funny how even though they clearly knew this they never mentioned it in previous conference calls!
Bob Rivet
Yes, I mean the zip code I have given people last week is still the same again - more details to come. We will lose around 100 million a quarter in process R&D in the AMD Design Co business that will be replaced in gross margin by higher wafer costs. Operating income is effectively neutral.
So those thinking that suddenly AMD expenses go down without manufacturing and R&D need to realize this will be offset by increased wafer cost as they are now buying it from a foundry.
David Wu - Global Crown
So basically, in Dresden, we got one SOI fab and one bulk sea Moss fab?
Bob Rivet
That is maybe a little too simplistic but you are not too far off.
So f38 will not be CPU?
And the questioner award goes to Citi who saw through the 191Mil smokescreen (though Bob Rivet totally ignored the question and blew smoke):
Glen Yeung - Citi
Thank you. If I look at your operating margins in computing and I ex-out the gains that you had from the licensing revenues, it looks like the profitability actually went down there. I wonder if you would talk a little bit about what drove that because it sounds like servers did pretty well for you, I would imagine that is good margins. So, if I had to guess I am going to assume that mix had an issue or played a role in this impact. I wonder if you can just clarify if that is true.
Bob Rivet
This is Bob. If you takeout the technology and try to do it from that perspective the mix is always a challenge. We are introducing new products. It is a combination of all those things above that just move things around. In total, we are feeling pretty good about the progress we have made in server, notebook, desktop, etc.
Notice he did not disagree with the assessment that CPU actually WENT DOWN! He just babbled and said we are happy.
The easiest way to distill all this down: What would AMD's operating income of $41Mil be if you took away the one time revenue of 191Mil? And that Sparks (if you've read this far) is the trick behind this 3 card monte game! Where's the one time 191Mil card... find it, find it.... ahhh it's hidden in CPU sales.... obviously!
ITK - I also how Fudzilla concludes AMD's yields are traditionally better based on relative fab count!
Ab-in-stupid did a similar "calculation" (I use this term loosely) and concluded AMD's yield was nearly 100% better than Intel, but realizing this might be a bit absurd put an arbitrary correction factor to get AMD in the range of 25-50% better.
The problem with simply counting fabs, and looking at relative market share is that it ignores the SIZE of the fab (they are not all sized the same), it ignores the aggregate die size (mix between products types, quad vs dual ratio may be different ratios, etc) and it ignore the fact that Intel happens to make more than just CPU's in house (chipset, wireless, etc).
But come on ITK - what do you expect from Fudzilla?
BTW it is an odd day of the week - does that mean Fuddie is going to say AMD is using SOI for 32nm or abandoning it?
“And that Sparks (if you've read this far) is the trick behind this 3 card monte game! Where's the one time 191Mil card... find it, find it....”
Kudos, you called it, months ago.
It’s still lipstick on a pig from corporate slug who talks out the side his mouth in light of years of bad judgment. Now, he’ll expect a pat on the back for what a great job they’ve done (And some fat bonuses, while the poor bastards on the floor are worried about feeding their kids.)
“Good job Bob………’pat…pat’…..”
SPARKS
anon: "So what happens if you take it out? CPU revenue was ACTUALLY DOWN 6.5% YoY."
Do investors and analysts fall for this sort of thing? Sure, AMD will try to spin the numbers as positively as possible, as any company will do. You expect analysts and experienced investors to read between the lines and understand what it all means.
You can dress up the numbers as much as you want, but reality has a way of being implacably persistent. AMD expects Q4 to be "flat," but aren't they introducing new CPUs any day now?
Tick Tock Tick Tock watch AMD slip away.
Well things are a bit better for the men in green. The bleeding has slowed but they are still in the red.
They have found a new sugar daddy willing to sink billions with hope of learning the silicon business.
I gotta ask, now that AMD has found a new banker how are they going to turn it around? The business has gotten more expensive each generation. AMD's margins are lower then many a foundry. Their new sugar daddys wants in oun that business. How can AMD give their sugar daddy a cut and still make money when without that daddy they couldn't make money.
Soon the sugar daddy will realize this old skanky company is a money pit and kick them out of the fab and fill it with higher margin customers like Nokia, nvidia, etc.
As the the old silver haired founder said "real men have fabs" These days amd is a pussy
You expect analysts and experienced investors to read between the lines and understand what it all means.
Typically, I think you'll see some euphoria and the price will pop - then a few days later the "real" analysts will come in and the price will settle back to reality (you saw this phenomena after Asset light was announced).
The stock ran 10-15% up after hours; it's now dropped back down to up 2%, so it looks like the market read things much quicker this time around.
Analysts and and AMD fanbois are idiots. They see clock speeds, benchmarks, or overlcoking and their eyes glaze over and they spring a hardon.
If an analysts had a clue and any competence they would have a real job in the business versus watching from the sidelines.
AMD is finished, their fate was cast 4 years ago.
Well, here it is. Fudzilla’s official back tracking. Coolaler and the boys have just BROKEN the memory bandwidth record with X58/965. Further the moron writes this:
“One of the interesting things is that memory was able to achieve that OC with voltage set at 2.05V while the CPU VCore was set at 1.576V which means only one thing, that MSI's Eclipse, unlike some other X58 based motherboards, can separate CPU and memory voltage and the 1.65V limit doesn't apply.”
YA THINK????
http://www.fudzilla.com/index.php?option=com_content&task=view&id=10009&Itemid=1
ITK- I love ya, brother, but stay away from those conspiracy theorists and web sensationalists. They’d do anything to tarnish INTC at this juncture; it is simply the greatest company in the world. That’s a fact and not fanboism. The numbers, the technology, and execution prove it.
I am proud to be a shareholder, especially now.
God bless you guys and your families. Well done!
SPARKS
Well, here it is. Fudzilla’s official back tracking. Coolaler and the boys have just BROKEN the memory bandwidth record with X58/965. Further the moron writes this:
“One of the interesting things is that memory was able to achieve that OC with voltage set at 2.05V while the CPU VCore was set at 1.576V which means only one thing, that MSI's Eclipse, unlike some other X58 based motherboards, can separate CPU and memory voltage and the 1.65V limit doesn't apply.”
Poor Fuad, never have I witnessed a 'hardware' news site so consistently get every technical detail of the inner workings of a computational device wrong.... and consistently wrong.
JEDEC standards specify frequency, voltage, and latency.... memory settings are programmed in the SPD of the memory, voltage that drives that memory is different than what drives the voltage of the CPU .... every MB made that supports DDR, DDR2, or DDR3 posses a separate VRM for memory and sets a separate voltage (unrelated to the CPU) for that memory ... it is commonly referenced Vdimm where as the CPU voltage is Vcore (or Vcc).... even early leaked photos of ALL and I mean ALL x58 MBs show the power MOSFET and capacitor for the Vdimm VRM aside the DIMM slots.
It is rather sad ... such ignorance gets perpetuated on that site ... there is no other explanation, just pure, unadulterated, concentrated, inexcusable ignorance.... the one thing that can be said of it ... it makes one laugh.
The idiot even thinks the chipset still determines the memory compatibility even on AMD systems...
Jack
Kindly avert your eyes. Watching the troll feed is disturbing to some.
If an analysts had a clue and any competence they would have a real job in the business versus watching from the sidelines.
Man, you got that all wrong. Why would they want to bail on a job that is paying them 6 figures for one that pays 5. Last time I checked, engineering wasn't the highest paid profession out there.
Engineers may not be starving, but they ain't exactly making the kind of cash these supposed experts make.
LOL, if you are any good as a engineer you should easily make 6 figures as an engineer.
Like many professions if you are in the middle of the pack you probably are a high double figure, but if you are really in the top 10% then 6 figures is a given.
I know a few short time analysts and they were clearly middle of the pack engineers looking for green pasteurs. Analyzing is far easier then having to solve hard problems. Reading tea leaves who the fuck cares if you are right or wrong in the end. If you are wrong as an engineer you end up being someone like the managers and designers of Barcelona or something, LOL
*yawn*
I may correct your trollish comments once, but I'm not interested in playing your game.
“Reading tea leaves who the fuck cares if you are right or wrong in the end.”
Ahhh…., not so fast, pal.
This may be true a year ago before the bottom dropped out, when the pigs were feeding at the trough, not now.
You should have seen the sons-of-bitches wrapped around the block in lower Manhattan looking for work at one companies yearly ‘job fair’. It looked like a scene from those old depression photos from the early 30’s, like soup kitchen lines.
They had their best gear on, too. Dressed in black, it looked like they were in mourning, like a frig’n wake or some kind of twisted death march.
It wasn’t pretty.
I’ve mentioned my LIRR commuter buddy who I ride the train with. He tells me many of these guys are looking for jobs selling house wares and brassieres at Wall mart, and the like. They’ll be SOL as Christmas shopping is not going to be very merry this year.
One beaut’, who worked for Lehman, had his 3 million plus retirement fund melt down to 60K. Another guy was on suicide watch at Belleview. Companies like Barclays are bottom fishing for prime fillet, on the cheap.
The skies over Manhattan are dark, swarming with buzzards, looking for some good pick’ns.
The list is endless, and there are a lot of carcasses. AIG was an Insurance vulture’s wet dream, they wanted the business, not the people.
I don’t know how things are in sunny Arizona, and I wouldn’t give you 2 bits for a half dozen anal ists, but let me give you a news flash, they ain’t making that kind of money anymore. Even the hot shots at Golden Sacks are looking over their shoulders, and that saying something.
You’re too arrogant to be a mediocre engineer, I’ll give you that. But, if you do know any who are, tell them to sharpen their pencils and forget finances, presently, stick to humping pods, and noxious chemicals. Plenty of sucking up couldn’t hurt either.
SPARKS
Jack- I’m glad you pointed out the JEDEC DDR3 standard and the voltage requirements. To be honest, I knew about it and should have realized it and said so, as did you. Then again, you’re in the business, I can afford to drop the ball.
In any case, that idiot has been spewing this nonsense for weeks.
After your post, I looked into the V requirements of DDR3. Officially, they must work at 1.95V without issue. There is some headroom built in, as mine are maxed out at the DIMM manufactures recommended 2.1V, clearly exceeding the standard. By the way, the heat spreaders are cool to the touch at these speeds and voltages.
Further, as you no doubt know, motherboard makers like Asustek are very resourceful in stretching the envelope on high performance boards. This is what I live (and pay) for.
I’ve seen the some X58 boards; they are beauties and will no doubt be tweaker’s dream come true.
Fuddie, it seems, doesn’t get this either.
The first one comes with SAS! Raid 0 SAS will even give SSD’s a run for their money.
http://forums.bit-tech.net/showthread.php?t=155947
http://www.tomshardware.com/news/intel-x58-nehalem,5829.html
http://www.techpowerup.com/index.php?73499
This one comes with a control panel!
http://xtreview.com/addcomment-id-6604-view-Asus-X58-motherboards-List.html
SPARKS
The guys that lost their shirts at Lehman and the many other firms downsizing are more then analysts.
Call them traders, brokers, M&A specialist, bean counters or whatever. Sure there are a few analyts among them. The most famous of the infamous that comes to mind is Henry Blodget ( was that his name ). Look where all his headline predictions are today.
It is sad in this greedy world that the financial firms pluck many of the best and brightest physics and math majors into wallstreet seducing them with 6 and 7 figure salaries instead of letting them go discover something great, or build that better mouse trap.
Its ugly these days with more then one house for sale on every block and some developments in the middle of the desert with nothing but one or two unsold houses. Analyize that you so smart 6 figure analyst, weren't you suppose to see it all coming?
Average engineers are a dime a dozen for 5 figure salary, they do CE just barely. Its the smart ones that invent
"financial firms pluck many of the best and brightest physics and math majors into wallstreet."
Well, there ya go. You learn something everyday. Honestly, I didn't have a clue, nor did I realize the academic requirements so closely related.
It must be the number the crunching discipline. Frankly, I’ll have to concur, it is sad and waste of valuable talent. However, there are alternatives, and this is one amazing woman, with a Harvard Degree in Applied Mathematics. She’s found something she loves to do, and she does it well. She’s found a brilliant formula, on both counts.
http://www.youtube.com/watch?v=VlPSxemBhYU
SPARKS
"Its the smart ones that invent"
Actually these days scientists do most of the major invention work - there is a difference between scientists and engineers.
And as for it being "sad" that greedy Wall St "plucks" the best and brightest. I must have missed the part where these folks are FORCED to take these jobs out of school. Is it not greed on the part of these people to go for the high salary? If these people are the "best and the brightness" they must understand the job they are taking, no?
As usual there is no longer any sense of personal accountability... everything is blamed on the evil oil company or evil Wall St or the government or some other evil entity. People that make decisions every day to enable these entities are mere unwitting pawns who are seduced, tricked or strong armed into everything.
End RANT!
That's not a rant. That's a sobering, well grounded dose of reality. Sometimes, we all need a shot in the arm.
That why I think your tops!
SPARKS
Scientist discover and maybe ivents.
It is the engineer that takes these discovers and invetions and make them a reality. Look only at the lowely transistor. It is the sweat of many a smart engineer to continue shrinking them. It was the scientist that long ago said oxide couldn't go thinner, or lithography couldn't print smaller, or that the thermodynamics would prevent the manufacturing of stable High-K metal gates.
But the engineers always find a way.
Average engineers are a dime a dozen for 5 figure salary, they do CE just barely. Its the smart ones that invent
Go ahead and change the argument if you'd like - but the above were YOUR words. Silly me for not seeing that "invent" meant "make it work" or "finds a way". I don't disagree with your last comment (that is pretty much the definition of "engineering"); but it is nothing like what your previous comment stated.
I love when people change the argument mid-stream as opposed to just saying I was wrong... you're not related to hyc by any chance?
As usual there is no longer any sense of personal accountability... everything is blamed on the evil oil company or evil Wall St or the government or some other evil entity.
You have no idea how much I wish I could disagree with this statement. But it is all to true.
Go ahead and change the argument if you'd like - but the above were YOUR words.
Which is why I dropped the argument in the first place. I made a claim that analysts were generally better paid than engineers.
This was twisted into a comment on my ability as an engineer with no basis in fact or knowledge of what I make.
At that point the conversation was over as far as I was concerned.
How good an engineer are you?
You seem a bit sensitive there for someone so intheknow, or are you so sensitive because you are usually the last to know, LOL
There is an interview in this issue of The Overclocker with an Intel rep (from France, I presume). Some cool stuff about Nehalem and a mild dig at AMD.
"It is possible to increase the base clock of Core i7, even on the lower-frequency Bloomfield."
"As you probably notice in the industry, no High-K means no big OC."
Heh...
http://www.reuters.com/article/marketsNews/idINLK44073820081020?rpc=44
"France's Soitec issues warning; H1 sales drop"
"Uncertainty regarding macro economic conditions together with indications of short-term demand reductions from some large customers results in an overall lowering of the second-half sales revenue guidance," it said in a statement.
Soitec supplies silicon-on-insulator wafers to semiconductor companies and has been plagued by delays in orders to customers such as Advanced Micro Devices
Keep in mind short term on substrates means mid-term on production. Either AMD has not started as many wafers today and has been banking the SOI for a rainy day, or they expect sales in Q1/Q2 to dropoff as SOI ordered this quarter and next would not be wafers out of the fab until Q1 and Q2.
ITK - don't feed the troll; he has nothing to add from a constructive viewpoint and lives to attempt to make himself look better by trying to bring those around him down. (emphasis on attempt and trying)
Why would a company have been banking SOI when it had a crap process running a crap design?
Of course AMD and IBM are dropping their orders for SOI substrates. Perhaps they have realized that all the noise about SOI being superior then bulk was nothing more then marketing pomp from the IBM fellows. In the end a few other insignificant players have far larger business in bulk and for many generations produced far superior transistors.
It is likely that decision of when one changes to High-K / Metal Gate and the resulting huge change in the models for the designers gives you pause to reconsider their silly adventure. Figure in moving to High-K it makes little sense to continue to try and tune to SOI.
Or maybe its because you are still trying to do that silly combination and sales truely are in the shitter, and why would that be a surprise either.
I'm still anxiously awaiting AMD SOI SION SIOC wonderfuly new 45nm what has it been more the a years since they started rolling of the lines at a few other companies and still nothing from AMD nor IBM. Gotta make you wonder if High-K / Metal gate will ever come from those silly guys in NY
Wow, you truly lack any strategic planning capabilities.
Just curious what you think will happen when tri-gate / FINFET devices come down the line (22nm). How about when 1C/1T SRAM cells or other alternate memory technologies come down the line?
Intel was smart not to switch to SOI as they had the foresight to understand that the benefit would sart falling off after 90nm. But why would AMD switch off of SOI on 32nm and then potentially switch back to it on 22nm?
From a foundry purpose you will see a 32nm bulk process, but I don't see why AMD would go off of 32nm SOI for CPU's at this point - given the technologies coming down the pipe on nodes after it. The main reason for moving off SOI would be for cost purposes, but if it is a one generation thing then the increased variable cost of the subtrate will be offset by the fixed costs associated with the switch. (Which is why Intel did not go with it for 1 or 2 nodes around the 90nm node)
But I guess when one is purely interested in bashing companies and people, instead of analyzing them, what should I expect.
I also would be interesting in hearing the EXACT specifics on how SOI is responsible for AMD's process issues. While it may not be helping, I would enjoy hearing you educate the masses and lowly engineers on this blog on how it actually HURTS the process. AMD's process is not on a cliff because of SOI.
http://www.tomshardware.com/news/Moorestown-Intel-MID,6513.html
Intel demonstrated the world’s first working Moorestown platform at the 2008 Intel Developer Forum (IDF) in Taipei, Taiwan. The demonstration was comprised of a validation board using three-day old Moorestown silicon.
This is the first "real" atom and is the chip Intel has been targeting all along. If this is the real deal then the 32nm shrink should be a player against some of the high end ARM based devices. This will also probably put some serious separation in the netbook space as the power #'s will no longer have the 945 chipset power anchor attached to it.
“Look only at the lowely transistor. It is the sweat of many a smart engineer to continue shrinking them.”
Transistors, “lowely” (lowly)? On the contrary, the big ones can be just as refined and beautiful as the little guys. Take the 200W variety; say 40 of them, in full Class A (per channel) in one of these…..
http://www.passlabs.com/images/xa.5%20images/xa5_images.htm
I’m sure as engineer you’ve heard of Nelson Pass. Frankly, I’m no engineer, but from a broader perspective, there is no such thing as a lowly transistor. In hi end audio; switching speeds are irrelevant, in fact, detrimental to sonic purity (Class AB), and there’s no need to shrink them. Bigger is better. They get the job done, exactly for what they been designed to do.
How about the “lowely” vacuum tube (my forte), say one of these beauties in Single Ended full Class A.
http://www.kraudio.cz/pic/300bxls.jpg
Then again there’s this. Sure it’s way out there, but the possibilities have been explored.
http://www.freepatentsonline.com/6437360.html
Who knows, maybe its back to the future.
It all depends on how much you’ve broadened/expanded your horizons, and how you appreciate them all, of course.
Like people,….if you know what I mean.
Check that, I don’t think you do.
SPARKS
For audio nothing beats the sweet sound of them glowing tube class A amplifiers
Now for your next generation phone or high powered quad-core computer, or your low power eepc nothing beats a little dose of nanometer transistors served with a healthy dose of High K and Metal gate.
SOI and SION need not come to the party
Tick Tock Tick Tock is that an arab with a bundle of SUV addicated american oil dollars that is buying those castrated men in green?
Alright - anonymous time to put or shut up....
What exactly is the NEGATIVE impact of using SOI in terms of power or performance?
You're doing exactly what AMD does - lumping 2 changes together (SOI and SION) and attributing both variables to the change.
It's funny how you dodge the questions that can't be easily googled. I would like for the "knowlegdeable" engineer/scientist/porn star wannabe to explain this to us?
This is the first "real" atom and is the chip Intel has been targeting all along.
As a trial balloon, I have to say I think Silverthorne has been a stunning success. But it needs what Moorestown brings to the table to be a serious contender. Namely better power characteristics.
From the few reviews of the Poulsbo chipset I've seen, Intel still has some work to do there. Hopefully the Lincroft package will give what Intel is promising. Numbers I've seen so far are around 5 hours of operating use in a mid/phone configuration. Figure in downtime with 20mW idle power draw and you should get pretty much a full 8 hour day out of one of these. Still a bit power hungry compared to most of the ARM line, but not completely out of the ballpark.
It sounds like Intel is planning a 32nm shrink on Moorestown a year later as well (2010). If they stick to the current cadence of another design rev in 2011 and a shrink to 22nm in 2012, I think they might well hit their target to be a serious competitor to ARM.
And as for it being "sad" that greedy Wall St "plucks" the best and brightest. I must have missed the part where these folks are FORCED to take these jobs out of school.
And this brings us to a pet peeve of mine. You have the tech industry (Craig Barrett has been very vocal on this lately) screaming that government needs to "fix" our educational system to produce more math and science majors. When in reality, it is within the tech industry's power to fix this.
They let the "best and brightest" escape to the financial industry by not offering competitive wages. High-tech is responsible for the situation they find themselves in.
Rather than crying to the government, they should out bid the other industries for the talent they need. If they don't there is nothing anyone else can do to save them. Only the offer of a competitive wage (compared to other professional positions) is going to bring more students into engineering.
It is a tough disciple to get through and nobody wants to go through that just for an average wage.
If a single brilliant engineer or scientist could make a difference then it probably is worth paying for them. But these days things like designing chips, developing a 22nm process take scores of engineers working across many diciplines. One brilliant architect or one brilliant device engineer will still be limited by the weakest designer or process engineer on the team.
At startups or wallstreet its easy for brilliance to shine and thus they get paid more by those industries.
Overall US is at a huge disadvantage. In a place like India or China how much of the population study as a way up. How many go to cramm schools. In the US the weekends are about baseball, soccer, or football. Is it no wonder our universities still the best and envy of the world are filled with foreigners. We educated them, once they used to stay but these days the more often then not take their education back home as times are booming back home.
US land of the sub-prime mortage and greedy people who will take a loan that they dont know they can't afford. In most other countries people pay cash or put a huge percentage down to buy their house.
There are actually very few things that the US lead in these days.
High-tech is responsible for the situation they find themselves in.
The industry also lobbies government hard for more H1 Visas which allows companies to bring in more engineering talent to the US at a lower wage. While not as severe as outsourcing this work to development centers overseas, it still helps to suppress the wages for engineering positions.
The other fundamental problem is the complete lack of accountability of the public school system. You have to be GROSSLY incompetent or nearly commit a crime to be removed. Of course the teachers union is one of the stronger lobbying groups in politics so this will never change; with the likely outcome of the 08 elections, things will likely get far worse in this area.
I do disagree with it being entirely within the tech industry's control - while higher wages will certainly help; especially encouraging students at the college level, it won't help the quality of the K-12 education. Just look at what passes as high school classes these days as opposed to 10 years ago.
Who is this anonymous guy in love with the Porn star?
Why oh why is SOI on the cliff for AMD. Just go ask them them if you can't figure it out.
To turn it around tell us how AMD's process is a solid and sound choice with a good roadmap.
AMD is BK, but they did find a good surgar daddy. Wait till that surgar daddy wants some dough from that multi-billion dollar fab. AMD margins and preference for capacity will go poof. AMD couldn't make money without a middle man, not they got a pimp how the hell are they going to be better off now?
To turn it around tell us how AMD's process is a solid and sound choice with a good roadmap.
Just as I thought - able to lob grenades and that's about it. Find a single comment where I said the above. I said SOI is not the cause of AMD's process issues and that somehow implies AMD has a solid roadmap and made a good choice with SOI? Delusional much?
You need to "turn it around" because you can't defend your own ignorant statements - so as usual you change the argument; and then twist others' words in an absurd way and argue against that.
Again since you are incapable of reading - Tell us how SOI SPECIFICALLY is responsible for AMD's process issues
It's not the SOI that puts the process on a cliff, but as you seem to know otherwise, ONCE AGAIN PERHAPS YOU CAN EDUCATE US! Maybe you can throw out some official sounding transistor parametric terms and dazzle us with your Google abilities.
It really is just sad.... you don't even have the applied knowledge to know why AMD's 65nm process is on a cliff. You pretend you do, but when confronted you scurry away on to the next argument.
I'll save you the trouble and not shatter your reality - it is SOI that causes AMD's problems - it causes the thermal problems, causes the inability to get higher frequencies and it is Intel's avoidance of SOI that is the primary enabler for their better process. You are, as always 100% right and I was completely and utterly wrong (as my past tack record on this blog has born out). (To the rest of the readers not living in bizarro world there are a few hundred other process steps which may also be important, in fact a few of them may even be a wee bit more important than SOI vs non-SOI)
"I'm laughing at the superior intellect" (name that movie)
Let me guess - next you will "turn it around" and ask me to tell you why AMD's process is on a cliff because you will be unable to ask the question posed here.
Sparks - Could you do me a favor and hook up some 480V, 3 phase to this clown and let me throw the knife switch when you are done?
Why don't you dazzle us, I'm totally unimpressed.
I feel for you, so much bitterness. Whats the matter wished you and your job mattered, but are actually aren't?
I'll think about whetting your appetite if you can really throw some more insults
Just as I thought...unable to answer what should be a simple question given you 'extensive' knowledge and background. Do you need more time to Google? Something like this should take you 5-10 minutes to type up and explain if you really knew.
It would actually be kind of funny if it weren't so sad - get banned at other blog sites and try to come her and pose as someone who knows their stuff only to get exposed.
Come on, we're all waiting... why is it that SOI puts AMD process on a cliff? Come on throw out some terms and generalities to "whet" our appetites (for the folks at home "whet" translates to: built in excuse for not being able to provide detail as the type of detail needed here would not be found on the web)
And for folks with the popcorn out and enjoying this, next up is "yawn" with excuse B (or are we up to C now?). But make no mistake, there is no specific answer coming. It will either beneath him, too hard for us to understand, a waste of his time, or some other vague response
Sorry you need to do better sucker
But make no mistake, there is no specific answer coming. Because there is none - the poser was exposed.
"Sparks - Could you do me a favor and hook up some 480V, 3 phase to this clown and let me throw the knife switch when you are done?"
My buddy is on Phase 3 of pulling out the last remaining 400A Knife Switches out of W.A. on 49th + Park.
If I can pry one of those lovely, very lovely, old copper 'details' out of his fingers before it hits the 'mongo' bin, it's yours.
SPARKS
The industry also lobbies government hard for more H1 Visas which allows companies to bring in more engineering talent to the US at a lower wage.
And then cries when no U.S. students want to go into engineering. The tech industry has managed to artificially reduce wages with the use of visa's, but it is not a good long term strategy. It is another indication of the short term greedy approach that has become characteristic of our society.
They stuck their head in this particular noose and if they aren't careful, it will choke them.
I do disagree with it being entirely within the tech industry's control -
The other fundamental problem is the complete lack of accountability of the public school system. You have to be GROSSLY incompetent or nearly commit a crime to be removed.
....while higher wages will certainly help;...it won't help the quality of the K-12 education.
I have kids in the public school system and I am not impressed with it either. So I will concede that at least that portion of the problem is not the fault of the tech industry.
But isn't putting the onus on the school system just another way to shift responsibility from those that really own it, the parents?
My kids do their homework, they keep their grades up, and my wife and I make sure that they learn the things that are important for them to get by in the world. Because someday they are going to find that nobody is looking for excuses to call them successful.
It is disturbing how many teachers comment on the fact it is obvious we work with our children at home.
I warned you this whole thing was a pet peeve of mine. :)
intheknow: "But isn't putting the onus on the school system just another way to shift responsibility from those that really own it, the parents?"
They share the blame. On the one hand, teacher's unions allow bad teachers to remain on the job far longer than they deserve. This costs money and efficiency suffers as well.
On the other hand, there are students who manage to excel in the same schools and the same classes where their classmates fail miserably. I consider this a failure of parenting, which in many ways is much worse than the union issue. How can you fail your own kids?
New York City is a prime example of the results of both of those problems. We spend something like $18,000 per year per student, and their reading/math scores are abysmal. The solution? Throw more money at the problem! It's downright depressing.
I guess for Mr Anonymous he is hoping to get the golden nuggest of much debate and reverse engineering. Sorry that stays where it must.
You can say what you want it really matters not to me. AMD's path has and was flawed when it went SOI. If it was so superior should not the designers have delievered a capable Barcelona with similar cache sizes as intel 65nm parts. The fact AMD has struggled mightly says it all. The details are for those that need it or maybe clever enough to deduceit. Are you one such person anonymous? If you are share it here with those others. Trust me, it really won't change my outlook as I already know.
Why aren't there more joe the plumber kids going to engineering school and then on to graduate schoo? Its because Joe the plumber, and Joe sixpack in general have far more passion about their kids sports activities then promoting excellence in math and science. People who think that making their kids do their homework and keep their grades up is all that is needed complete misses the point
Look at schools like Berkeley, MIT, Stanford, Cal Tech and go check on the ethnic makeup versus the general US population. It is something that goes far deeper then doing homework, keeping the grades up. Its a value system that underpins the whole family unit.
Look at the Far East and how the family values education versus the US. It isn't about doing homework or keeping the grades up. Saying that already tells me a lot.
The details are for those that need it or maybe clever enough to deduceit.
Did I tell you guys - not quite the exact excuse from my list of choices but pretty close. Additionally, the 'SOI is making the process bad' argument has now morphed into 'SOI was the wrong choice' argument. Typical intellectual cowardice. Where did I say the AMD process in general was superior? I believe I've been saying on this blog for over a year it is inferior technically (and behind on schedule). I also said there would be a cliff at 2.8GHz on 65nm (I was wrong on this one as the cliff is a little lower)
If you are share it here with those others.
I had this one bang on - this comes from the Dementia school of reasoning (apparently he was around that blog long enough to pick up the bad habits before he got banned!). So let's see make an outlandish and wrong statement. When you get called on it and asked to support it - you either change the argument or ask someone to prove you wrong.
Again, what about SOI is responsible for AMD's process issues?
Tock Tick Tock Tick.... make no mistake, there is no specific answer coming. Intellectual BK!
Tonus, ITK you are both dead on.
As I've been preaching personal accountability the ultimate responsibility comes down to the parents and kid. However, having some accountability at the school level would increase the "error margin" and perhaps catch kids that otherwise fall through the cracks.
The problem with most gov't run systems is there is always political pull, unions, lobbyists, etc. While these groups serve some purpose, they confuse their job and think any concession is bad and their purpose is to protect every individual (which it is not). If any of the gov't agencies were run as businesses (whether it is the cost over-runs on Defense contracts, public schools, emergency response) - the company would flounder, revenues would decrease and either the underperforming people would get laid off or the company would die.
Throwing more money at the school systems will, believe it or not, make things worse. It will only entrench the corruption and waste, and will only re-enforce that the perception that the current system is working. The only way to truly fix this is with some tough love - if you cut the budget (I know this sounds insane); you will force the tough choices, clean out the underperforming teachers, clean out unneeded levels of bureaucracy, force innovation and drive new efficiencies. When this happens, then you re-invest.
If your 15 year old is running up crazy bills and getting into trouble, do you double their allowance to fix things? You do if you are the US gov't. Money is always the answer... (especially when you can just collect more taxes rather than actually trying to fix things)
It killing ya isn't it, not knowng the true reasons and issues behind the various technologies!
Tick Tock Tick Tock, or is it knock knock, time to have your milks and cookies before you go to bed, its past your internet bedtime.
Poor little anonymous poster, LOL
It killing ya isn't it, not knowng the true reasons and issues behind the various technologies!
Yeah, you got it.. perhaps you can explain how SOI is the CAUSE of AMD's process issues?
Once again, folks... the answer is never coming. (Because there is none, all you will hear is 'poser-ese' where the argument shifts, and the inevitable I could tell you but you wouldn't understand comes out)
So it is not killing me because your comment that SOI is the reason behind the process issues is flat out wrong (and demonstrates your true lack of knowledge). There are other reasons behind the issues which I and others on the board had predicted lone before 65nm cam out... and it ain't SOI...
But I'm sure you'll get the last word in as you are an insecure, wannabe who'll change the argument and mix up the words. Remember though it is better to keep your mouth shut and have people think you a fool then to open it up and prove it.
To others on the board my apologies for trying to expose this idiot. Outside of AMD BK, AMD sucks, AMD process sucks what actually is this guying adding to the board? It is the same old schtick, no actual logical arguments and no actual information.
All------“Unions”, please preface that with Teachers Unions. They are definitely not, Electricians, Iron workers, Plumbers, Elevator constructors, Fitters, Laborers. You see, we (me inclusive) are in the PRIVATE sector. We are NOT an extension of a local government (s). We work for a living.
Believe me, with every essence of your being, if we don’t PERFORM or turn a PROFIT for the owners we are SOL, done. In fact, yours truly is frequently called upon to run jobs with “low margins” because I can. Occasionally, they’ll throw me a bone, give me a really ‘good’ job, and let me turn 30 to 40 percent profit, piece of cake.
Conversely, teachers have no such scale to determine their performance, and “Tenure” insulates them from all but the worst transgressions. Not so with us. You’re only as good as your last job. Further, if you think I am going to hang my balls out for a slack, think again, I’ve got kids to feed (and chips to buy).
The word Union nowadays is wrought with implications of entitlements, of being overpaid and under worked. Yet the Teachers Union by virtue of higher educational professionalism has set a new standard for the definition. Clearly, what other union employee receives full wages FOR NOT WORKING 3 MONTHS OF THE YEAR, and not be judged by any comprehensive performance standard.
Not in my Union you don’t. We simply don’t have a polished education, clever talk and a nice smile to get by on. (Well, most of us don’t :) ) Further, we don’t share the blame with our customers because of an inability to get the job done. We find a way.
Besides, are you familiar with the term ‘educated idiot’. Hmmph!
“Schtick!” Oy Vey, such words! Watch ‘m “G”, he’s a real schmageggi!
SPARKS
Insecure? Ouch how did you know? You're so right you see thru it all. I am nothing in this world, hiding behind alias and google for fancy words on the internet. I live in my moms basement and work at bestbuy telling people to buy Intel over AMD.
Tell me one technology thing you know beside asking me for the answer you twit. Impress me, give it a go.. test me you silly geek.
I don't give office hours to silly people like you
Tell me one technology thing you know beside asking me for the answer you twit.
The next informative thing you say on this blog will be your first.
How about telling us all about how "analog I/O" was the limiter in the ability to get 2X scaling on a new node? Do you remember that "gem" you gave us? I almost spat out the coffee when I saw you write that...
Not only are you an idiot...you can't even insult me with any originality. The phrase not the sharpest spoon in the draw comes to mind (yes I meant spoon).
Still waiting to hear how SOI hurts AMD's process... maybe it's the whole I/O analog scaling issue? Oh wait I forgot that is the key issue limiting scaling (stifles laughter) or maybe a DIBL thing or a Miller capacitance thing or [insert latest googled term here to make it sound like I know what I'm talking about]?
All------“Unions”, please preface that with Teachers Unions
My bad Sparks - didn't mean to lump all unions in together. Unlike some, ah hem, folks, I'll own up to a mistake.
But tell me were you on your fourth or fifth mandated coffee break when you responded?
(Oh come on, I'm kidding! That was a little funny, no?)
Unions are bad, anytime when you get collective bargaining is a recipe longterm for the slow build of inefficiencies.
Unions at the time they were invented made sense. In western developed countries unions and any collective bargain group will always result paying more then what the labor is worth.
Teacher Unions, UAW, trade unions it doesn't matter.
Take the port workers, do they do important jobs that deserve 6 figure salaries? Yes, do they deserve getting what they get at the moment? I think it is a bit too much.
Calling Lex's fluffer let me know a good private way for for us to have a dialog. If you prove worthy I just might consider helping your education about silicon technlogy and being more then my fluffer ;-)
So transparent... so sad...
I pity you buddy
I pity you buddy
It's just downright painful... no originality, no intelligence, nothing to say. He even finally owns up to being Lex.
Tell us, oh wise one - why again is SOI hurting AMD's process?
Oh that's right you're incapable of answering that question and having an intelligent discussion - so just continue to lob insults and grenades and refuse to admit you are wrong on this one. I believe we are up to excuse # 6 on why he can't answer that question.
“Unlike some, ah hem, folks, I'll own up to a mistake.”
Even if you made me look like a monkey, you’ve given me too much to ever piss me off. (Like the reason why AMD had issues with SOI)
“But tell me were you on your fourth or fifth mandated coffee break when you responded?”
Funny you should mention it. In fact, we take one ‘coffee’ per day from 9:30A to 9:45 ~ 9:50A. I’m uncomfortable going any closer to 10.
Strictly speaking, the break is not in our contract. It’s a tradition that goes back to an earlier time when Edison and Westinghouse were pissing on each other during the AC/DC wars.
“For those about to rock, we salut…………” Sorry, got carried away.
Owners/Contractors have continued to honor the break freely, without issue. Actually, it’s the one and only thing that is never mentioned by either side during contract negotiations.
It’s the State labor laws that mandate certain rest intervals for all workers and they take precedent over any contract, be it non-union/union/office worker etc., just like minimum wage. But, “coffee” is a sacred tradition, on both sides, like English “Tea”.
On a more interesting note is this:
“And then cries when no U.S. students want to go into engineering. The tech industry has managed to artificially reduce wages with the use of visa's, but it is not a good long term strategy. It is another indication of the short term greedy approach that has become characteristic of our society.”
ITK- Brilliant! If ENGINEERS organized the way the IBEW did, they couldn’t pull this shit and I’d put up our engineers against any in the world. (Our guys even made an SS officer/scientist brainstorms work. They even gave Wernher a dozen or so Atlas ICBM’s to kick start NASA).
Then there was the BRILLIANT genius GOD of engineering, Kelly Johnson. Kelly never believed in the division between his technicians and his engineers. They worked together, hand in hand, side by side.
There’s no “I” in team.
SPARKS
“Take the port workers, do they do important jobs that deserve 6 figure salaries? Yes, do they deserve getting what they get at the moment? I think it is a bit too much.”
This is a simplistic, one dimensional perspective, indicative of pure ignorance.
I’m sure from this narrow perspective you’ve only heard what they make, and assume what they do. I ask you, when was the last time you’ve actually visited a port? I’m willing to wager never.
Factor this, do you think with the current system of containerization, transit lifting, and automation there are the same number of workers doing the same jobs back, say, 20 or thirty years ago? Don’t bet on it. This is all done with far fewer people than ever before. If you dispute this, bring some facts to the table.
Most of the work is with, mechanized gantry lifting, cargo manifest, import duties, and the biggest responsibility, of all, inspection. Yes, dirty nukes, and all the other lovelies that may go along for the ride. They ALL must be carefully and rigorously inspected by, you guessed it, people.
Of course, if you live in Long Beach, California, maybe you would be happy with untrained $8.50 hour personnel, working in the harms way, checking the containers? Do I hear a time bomb ticking? That’s not baby power, that’s Anthrax.
Perhaps you would like to tell a Long Shoreman who is permanently disabled, due to some unfortunate accident; he was making too much money. Tell, his kids while you’re at it.
That was a stupid example, short sighted, and ridiculous, like its author.
SPARKS
Jack-
ITK-
Well, Fuddie has back tracked as far as possible. He will not admit what he published previously was wrong, but at least he is publishing more accurate information regarding Ci7.
http://www.fudzilla.com/index.php?option=com_content&task=view&id=10088&Itemid=1
........AND, it is a MONSTER, just like MAXIMUMPC said it was.
I will need to keep QX9770 overclocked to stay in the same league, that's a fact.
http://www.fudzilla.com/index.php?option=com_content&task=view&id=10081&Itemid=1
The boys at INTC never, absolutely never, looked better.
Those morons at Golden Sacks can down grade INTC till half the finacial institutions in the world turn into belly up carcasses.
INTC is running like one of these!
http://www.arcent.army.mil/cflcc_today/2005/february/images/feb01_03/tank2.jpg
SPARKS
8.5 an hour? No No, I think they are worth more then that. Are they worth six figures, NO!
Entry level should be like anyother trades. Once trained and with a few years under their belt, I think 60 thousands is reasonable.
I think teachers are underpaid for what they do for society, but that too many of the teachers don't deserve to be teachers, and the unions and guranteed job are the problem.
The rest of the world works on pay for performance and so should teachers. Unions shield people from pay for performance, and that is simply wrong in a free economy.
Somehow AMD thinks of itself like a bad teacher the right to continue to put out crap. And if it doesn't perform its somebody elses problem, their own problem, NO. Its INTEL, or if its teachers class with low scores its the parents. AMD is just like a union worker who thinks he deserves pay even when he is incompetent or puts out crap.
“Unions shield people from pay for performance.”
“Just like a union worker who thinks he deserves pay even when he is incompetent or puts out crap.”
You sound some maniacal distraught lesbian who over generalizes about men and why they hate them. What/which unions are you talking about? Where are these union workers who are incompetent and putting out “crap”? What crap? Do you have specific references, besides teachers, who are protected by a union who are not fulfilling their obligations?
“Once trained and with a few years under their belt”
60K????
Do you know that different unions have different requirements? Which one?
ALPA? Next time you’re on your business trip you tell your Captain, his union sucks, he gets paid too much, and he’s protecting/covering up the real bad pilots.
AP Mechanics-Hey, just hire these guys from Joe’s Salvage Yard, right outside Amarillo.
Tell an IBEW, during a freezing blizzard 40 bucks an hour is way to much money to restore power to keep your fat ass comfortable at 2 in the morning. Better yet, you climb the pole and tie in the live 4160.
Maybe you are talking about Steel Workers, climbing around on eight inch beams 30 or 40 stories up. You’d piss your pants at 10 feet off the ground, pal.
Maybe, you are talking about the UAW?
Why can Toyota (and others) take the same members in this country, produce and excellent product, pay them the same wage, and be VERY successful auto makers? You’ll find the difference isn’t with the union workers. In fact, in a Toyota plant, every member on the line can do any job, on that line, from frame to polish. That how much THEY slack!
Take at step back, from your anti union, simple minded mentality, think about that last question objectively, if you can, you find the answer isn’t so clear. Or perhaps you would be happy with a Pinto, Vega, or Fairmont.
Sure, union guys designed those monstrosities. Wait, those were ENGINEERS! Fuck it; we’ll blame those UAW workers for putting “out crap.”
While your at it, factor this, with that Clockwork Orange ticker. NOT ONE Union worker contributed ONE minute or ONE cent to the current historic financial debacle. It was ALL corporate, and ALL management.
Union workers don't have 'Golden Parachutes', just a lot of hard concrete.
Can you rationalize how the unions contributed to this mess with your biased, twisted, convoluted, pretzel logic?
Finally, and be honest, do you own an American car? If you don't, you’ve outsourced.
SPARKS
Toyota doesn't pay the same as GM and they make a far better product with cheaper labor too.
Don't type too much or you only show your ignorance there buddy.
By the way who needs a golden parachute when you got a job for life?
Sparks said...
If ENGINEERS organized the way the IBEW did, they couldn’t pull this shit and I’d put up our engineers against any in the world.
With all due respect if engineers formed a union, I'd be looking for a new profession. Not for the reasons you might expect, though. Mostly, the restrictive environment would drive me nuts. You summed it up beautifully when you said:
If you are one of the Generals, Forman’s, or Straws you don’t touch your tools, you supervise. (General’s and Forman NEVER touch tools, the straws will, occasionally, but they could be reprimanded. I have)
The idea that as I progress in my profession I have to move away from what I do well and love doing is nothing short of repugnant. Not to mention, having to look at a job that I could do and waiting for someone to come do it would drive me round the bend.
And back to tech news (not that I mind the other discussions).
AMD stock has now gone from $4.2 to $3.28/share in the 5 days since announcing "better" (and I use that term loosely) than expected earnings (or I should say less of a loss).
The market saw through the $191mil accounting gimmick; understood the high margin portion of the business (CPU's) is actually DOWN year on year and I think it is clear to the market that the foundry spinoff does nothing to AMD short term except smooth out the cash expenditures. Bottom line the foundry is not suddenly going to make chips cheaper in the short term and now they will take a cut on each chip so AMD has to take a smaller cut/chip to keep prices flat.
BTW - for the last month AMD was more or less tracking with the DOW and NADSAQ, until the last 5 days (Dow ~flat, NASDAQ down ~4-5%, AMD down ~25%)
That said unless you truly think AMD is going BK (like one poster here)... at $3.28/share you have to wonder if the stock is worth a flier (for a short term trade) at this point? (full disclosure - I own no AMD stock put am pondering a couple thousand shares)
It is rather hard to believe the market cap is now under $2Bil.
That’s it.???? Ignorance, that’s the best you can do? You concede the rest?
Listen, do yourself a favor, and don’t argue a point in which you unwittingly concede by default. Obviously, you agree that Toyota “….. make a far better product” in this country than GM.
Oh, by the way, the pay scale is basically the same.
http://www.payscale.com/research/US/Employer=Toyota_Motors,_Torrance_California/Hourly_Rate
http://www.payscale.com/research/US/Employer=General_Motors_Corp/Hourly_Rate
And you don’t own an American car.
Go back to your cushy desk job. Run a new program in that skull, you’ve got a corporate rhetoric nested loop going on. Your over generalizations are pathetically predictable. Don’t bash unions like some prepubescent nerd/bigot on the high school debate team. You know nothing of skilled organized labor.
You’re way out of your league.
(ITK-The Engineer’s in Nuclear Power Plants are IBEW members.)
("G"-Sorry, I'm done with this discussion. Your right about AMD, the share price surprized me too.)
SPARKS
(ITK-The Engineer’s in Nuclear Power Plants are IBEW members.)
All 2 of them? ;-)
AMD's stock price down over 9%, to $2.96 now. I'm tempted to buy some as I think the Yabba Dabba Doobie investors won't let it go much lower without buying more.
INTC down by about 6% to $13.62.
Still no sign of Scientia over at UAEZone...
Still no sign of Scientia over at UAEZone...
He's in the fab working on the evolution of APM to Asset Leann to Asset Smart. This whole foundry thing and spinoff, is really just an elaborate ruse.
All 2 of them? ;-)"
ITK- During lunch, I opened this blog at work on an in house computer. My partner asked me what the hells so funny after I LOL'd.
Thanks, it was a welcome relief from a particularly miserable day where!!!MANAGEMENT!!! delivered 27 buckets (modular switch gear, HOA)with no fuses, and no motor O.L. heaters, idiot.
The day was shot.
Thanks again for the ironic twist, very appropriate.
SPARKS
Moose!- Market Share at All COST!!!!!
http://www.xbitlabs.com/news/cpu/display/20081024231525_AMD_s_Market_Share_Drops_to_3_Year_Low_amid_Intel_s_Share_Hitting_3_Year_High.html
DOOH!
SPARKS
From the XBit Labs article:
"However, the main problem for AMD, which lost 1% of the market, is relatively narrow mobile products family."
That strikes me as awfully ironic. I think it was Guru who pointed out that one of AMD's missteps in the desktop market was creating way too many SKUs. Looks like they forgot to diversify their mobile line while they were busy diversifying their desktop offerings to death.
Unless what they are saying is that AMD simply did not have products that could compete at all ends of the mobile spectrum. Good thing for them that Atom was such a failure. :)
I think it was Guru who pointed out that one of AMD's missteps was creating way too many SKUs.
So folks do actually read the comments! The desktop scene is just ridiculous (esp quads) and will likely get worse during the 65-45nm transition, but perhaps AMD will learn and cut down the # of 45nm SKU's? I suspect with some better clocks on 45nm, AMD might be able to spread out the speed bins a bit more and cut down on the $10 increments between chips... but who knows if they will.
Unfortunately for AMD the problem in mobile, in my humble view, is branding, which carries a lot more weight (but will probably lessen over time as mobile continues to grow). They have a decent platform, but the CPU is weak - and that's because the CPU strengths (scaling, IMC) don't translate well into the mobile world and it's deficiency (the efficiency of the K8/K10 core compared to Core2) is magnified.
I'm not sure how AMD overcomes this without a new mobile CPU. I suspect they will continue to try to price their way to market share, but as we've seen this is a questionable strategy unless you have superior manufacturing (which AMD doesn't). Regardless, I don't think the 'gaming with an IGP' strategy is going to catch on - those who do want decent graphics on a mobile (which is still a limited target audience) are typically willing to pay more and will get a discrete card (and want a higher end CPU).
The only potential game changer I see on the horizon is on chip graphics - but we won't see that impact until 2010 and unless Intel stumbles badly on their solution, it probably will just more or less end up being a wash.
“If you are one of the Generals, Forman’s, or Straws you don’t touch your tools, you supervise.”
ITK- Sorry, I should have been clearer. The structured chain of command is the ideal for a really big job (like a fab). Those jobs need to be highly structured in order to maintain control and order. However, jobs of this magnitude are few and far between.
It all depends on the number of men you have and the size of the job. For example, take the job above. Upon its completion, we all go to different jobs where they’ll be three or four of us, and I’m running the work.
Maybe, they’ll send me the General and ask me to put him to work. Sure, he’s lost some skills, but he IS A VERY POWERFUL ASSET for the owner(s) of our company.
“SPARKS, Can you take “Pete” for a couple of weeks till his next big job starts?” “Make sure he doesn’t kill himself”. Even though he was a General on the last Job, he’s on MY job now, He works with his tools and he works for ME! Usually, as you surmised, they’ll appreciate the “forced demotion” away from the meetings, stresses, paperwork and get lost in the pleasures of “working with his tools”.
“Hey, “Sparks’ it’s been a while, can you give me the color codes for RJ-45”
http://www.bb-elec.com/tech_articles/EthernetCables_illustration.asp
Or……….
“Sparks, are you using expansion shields or thunder bolts on this switch gear?”
Most of the time, they are so skilled and knowledgeable they’re great to have around. They “run themselves” and help the other guys. More importantly, they give me the same respect and courtesy I gave them on the multi-million dollar project.
The other mechanics love it because they can work one on one with the “big guy”, and they’ll develop a personal rapport they otherwise would never have.
Further, running work is optional and voluntary. I can do both. General’s love me because I’m good at doing both. “Where’s Sparks? Get him down here to trouble shoot that VFD.” In fact, personally, I prefer working with my tools. That’s why I got into trouble. A Steward told me to put down my tools; I got in his face, and told him to go fuck himself (on the big job). Naturally, the General was there to explain why I was doing, and what I was doing. He smoothed things over. Then, the General pulled me aside and said, “Jesus, Sparks you could be a little more diplomatic.” “Yeah, yeah, yeah, I’m sorry, Pete.” Pete loved it.
Then I went back to work figuring out why the 1400 ton chiller wasn’t communicating with the rest of the building.
http://english.carrier.com.cn/products/product.aspx?navmenu=2&iProductId=23
(The one pictured is without the reactor and drive assembly. It stands about 14 feet high at the turbine end.)
Most of the time I’m on small Jobs (less than 5 men) where I can show guy’s the subtleties of control work, and in contrast, bending compound offsets in four inch ridged pipe on an hydraulic table bender.
http://www.arizonatools.com/tools/pipe-benders/detail/GRN881CT-E980/
If you ever wonder what ‘A-Journeyman’ stands for, the ‘A’ means ANYTHING!
I love it. The only way they’ll get the tools out of my hands is to pry them away from my cold dead fingers, just like my QX9770 and my shares in Intel Corporation.
SPARKS
I can just imagine a foreman trying to calm a nervous client by telling him "don't worry, Mister Smith, the transformer was a concern but I just saw Sparks down there and... sir? Are you alright? You're so pale all of a sudden..."
I'm not sure how AMD overcomes this without a new mobile CPU. I suspect they will continue to try to price their way to market share, but as we've seen this is a questionable strategy unless you have superior manufacturing (which AMD doesn't).
And yet another perspective on mobile: according to Ed at Overclockers, AMD's mobile marketshare amounted to just 11.3% for Q3 2008. And of that small amount, AMD only got 72%of the revenue that Intel got on average per notebook chip. As Ed points out, notebook/mobile is the most rapidly growing segment.
So I guess Puma didn't exactly set the market on fire, despite Sci's blather about how it would be a game-changer or something equivalent. And AMD's decision to cancel or delay Bobcat now looks kinda stupid.
I've said before that the tone of much of the comments at AMD Zone seems like what they hope to see happen, as opposed to what they think is realistic. I think that's why you can look back and find predictions that fall short. This is fueled by AMD's tradition of playing things very close to the vest and giving out just enough information to allow for all kinds of possibilities, even totally unrealistic ones.
When you are hoping against hope for a specific outcome, and you are getting carefully-vetted bits of information, it's easy to look past the realities of the situation and cling to alternative scenarios.
So I guess Puma didn't exactly set the market on fire, despite Sci's blather about how it would be a game-changer or something equivalent.
What are you talking about - he was dead on. Puma has been as much a game changer as DTX was predicted to be! The problem is you are interpreting 'game changer' differently. Or maybe he meant "name" changer and it was just a typo?
What are you talking about - he was dead on. Puma has been as much a game changer as DTX was predicted to be! The problem is you are interpreting 'game changer' differently. Or maybe he meant "name" changer and it was just a typo?
LOL - I forgot about his DTX prediction already :). As for name-changers, I hear AMD is now planning a Bullcat and Bobdozer release in 2010 that will set our pants on fire :)
Since their Q3 earnings report, AMD's stock has shed over 1/3rd of its value - closed at $2.94 today.
Tonus said...
I've said before that the tone of much of the comments at AMD Zone seems like what they hope to see happen, as opposed to what they think is realistic.
It's the green Koolaid they are sworn to guzzle, or else get banished by der Ghost. In fact, looks like he's been a bit too free with the ban stick - some of the greenest posters there got banned and haven't come back, so posting activity is way down. If they depend on any advertising dollars, I suspect Ghostie might get the business end of his own stick soon...
LOL - I forgot about his DTX prediction already :). As for name-changers, I hear AMD is now planning a Bullcat and Bobdozer release in 2010 that will set our pants on fire :)
While reminiscing, I thought this projected schedule would be useful:
Q2 06 - Woodcrest
Q3 07 – Barcelona Trailing by 5 quarters.
Q4 07 - Penryn
Q3 08 – Shanghai Trailing by 3 quarters.
Q4 08 - Nehalem
Q2 09 – Bulldozer Trailing by 2 quarters.
Q4 09 - Westmere
Q1 10 - 32nm Bulldozer Trailing by 1 quarter.
Did I miss the Shanghai launch last quarter? Did I miss the roadmap which has Bulldozer out in ~6 months? Do you think AMD will convert to 32mn ~1-1.25 years after starting to move to 45nm?
This is the kind of "analysis" you get from someone who has no real understanding of process tech and manufacturing. It is also indicative of trying to drive to a desired conclusion and working backward (wanting AMD to "catch up" to Intel) instead of trying to do a ground up analysis.
What are you talking about - he was dead on. Puma has been as much a game changer as DTX was predicted to be!
Yeah, DTX was a much better prediction than my claim that Atom would be a big deal. :)
EETimes has given us our first peek at Intel's 32-nm process. Key details are:
According to the IEDM paper, Intel built a functional 32-nm, 291-Mbit SRAM array test chip with a 0.171-micron2 cell size. The device has nearly 2 billion transistors, and an array density of 4.2-Mbit2. The test chip operated at 3.8-GHz at 1.1 Volt....
EETimes has given us our first peek at Intel's 32-nm process.
No mention of tri-gate - I'm standing by original prediction of conventional planar transistor for 32 nm (to enable a 2nd generation of highK/replacement flow) and look for trigate at 22nm with possibly SOI (and outside chance of change to the traditional 6T/8T SRAM cell). SOI/trigate may also enable (or require?) a gate first high K/MG flow... I'm not sure how well replacement flow would work with trigate transistor design as you are no longer filling simple transistor gate geometries, not to mention the ever smaller fill features.
ITK - do you know what speed the 45nm SRAM chip ran at? I don't know (one way or the other) if the speed indicated is what would be expected for a logic device.
If Intel is presenting at IEDM, it would appear 32nm is relatively on schedule. First tapeouts/Si for logic maybe Q1-Q2'09?
And again good call on Atom! Is your back OK - you are patting it rather hard! :) I'm still thinking Intel must be considering this gravy as clearly the SOC design is probably what Intel was targeting for volume/significant revenue. Intel speaking specifically to a 200Mil revenue stream at the conference call (which is only ~2% of revenue) is in my view an indication that Intel was surprised by the amount of revenue this early on.
And I still think Atom will level off due to the 945chipset limitations on power consumption (which the SOC design will address); but then again I was completely off on the popularity and adoption rate of the first gen, so take anything I say in this area with a huge grain of salt!
ITK - do you know what speed the 45nm SRAM chip ran at?
According to the paper presented at IEDM for the 45nm process the voltage and speed on the SRAM was the same.
The 45nm yield learning vehicle was a 153Mbit SRAM featuring a 0.346μm2 SRAM cell and over a billion transistors. The first fully functional 45nm 153Mb SRAM was reported in Jan 2006; yields are now at mature levels. The SRAM has demonstrated 3.8 Ghz operating frequency at 1.1V power supply and stable low voltage operation (Fig. 17).
Regarding Atom, I was more poking fun at the way it was dismissed with the claim that a new PCB layout was a bigger deal than Atom claiming any prescience. I've missed more than my share of times too. But this idea really struck a chord with me.
And I have to agree that Intel is probably viewing the profits from Atom as gravy. The fact that they have repeatedly had to dial up there manufacturing projections says they were caught by surprise too.
I saw Otellini on a video clip claim that Atom had better margins than Celeron. It seems most analysts miss that detail, look at the low price and assume low margins. So any sale of an Atom is a bonus vs the sale of a Celeron.
And yes, Menlow would seem to be the big play, as this finally gets Intel into the ballpark with high end cell phones. Cortex A8 is looking pretty good from what I've read so far as well, so I think we are still looking at 22nm to get real power/performance parity with ARM.
Sparks - you'll love this :)
From http://www.dailygazette.com/news/2008/oct/27/1027_amd/
AMD applies for sales tax exemption
BALLSTON SPA — Advanced Micro Devices asked the Saratoga County Industrial Development Agency today for a sales tax exemption on its planned $4.6 billion computer chip factory at the Luther Forest Technology Campus in Malta and Stillwater.
IDA approval would allow AMD - through a newly formed subsidiary, AMD Fab Technologies US - to save nearly $28 million in sales tax on building materials. That would be in addition to $1.2 billion in previously negotiated state incentives for the plant, where 1,465 people will work.
If it were federal tax exemption I'd guess AMD now considers itself either a charitable organization or a religious entity :)...
If it were federal tax exemption I'd guess AMD now considers itself either a charitable organization
Well they've been a non-profit organization for quite some time now, so might as well take it up a step to full on charity. :) I assume this is a NY state sales tax exemption and you were just throwing out the fed tax part for the charitable comment?
Hey Sparks that's not a hand you feel in your pocket anymore that's someone cutting a hole in the bottom so you're money just bleeds out quicker and on a consistent basis! Just think of the whole 1400 jobs that will be created at the end of all of this!
AMD just had the sale price of the CE division cut down by 51million at the deal close due to decreased expected revenue for that division. Not sure if this was why the stock was down $0.10, but in a day when the DOW is up near 900 points, you have to really work at it to be down.
But mark words... you will see another 141Mil one time revenue stream, creatively blended into the quarterly Q4 #'s. You have to hand it to AMD; they've spread out the equipment sale (190Mil, Q2); the tech license deal (191Mil, Q3) and eventually this CE division sale (141mil, Q4) rather brilliantly - managing to even blend the Q2 and Q3 #'s into gross margin and CPU revenue totals. It is not clear to me if they can do it again with the CE business sale - it may be stretching the accounting rules a bit too far to squeak that into "CPU revenue", but I'm sure they will somehow creatively blend it into something.
The problem will be next year when they have to compare to these artificially inflated #'s.... oh wait they will be a split entity, but still reporting consolidated #'s so they will be able to introduce all sorts of double speak (and probably tell folks they should compare to #'s without the one time revenue additions).
I must say the financial game being played is absolutely brilliant (it's about the best they can spin the #'s given the current environment and results), but the strategy is hinged on the business picking up in the future If/when it doesn't (whether it be economic conditions or competitive issues with Intel) then the house of cards will come crashing down when they won't have these one time revenue streams. It's a dangerous game, but they probably have no other choice.
Interestingly, the new AMD manufacturing spinoff will be incorporated in the Cayman Islands, situated in the western Caribbean, which, surprise, surprise, have no corporate income tax.
http://www.theinquirer.net/gb/inquirer/news/2008/10/28/amd-fab-spinoff-sniffs-possible
So basically an OFFICIALLY non-US company is going to get 1.2Bil (and change) from the state of NY... and a company that will be paying no corporate taxes... this thing just keeps getting better and better!
Sparks - can you write to Schumer/Clinton and ask why they are giving 1.2Bil to a company that is not even incorporated in the US and won't be paying corporate taxes. Aren't the Dems supposed to be against these loopholes?!?!?
Sparks - you'll love this :)
It’s only 28 million on 4.5B! That’s peanuts.
Let’s do the math.
8K in property taxes
7.5K in Federal
4.5K IN STATE!!!!
8.5% on every purchase.
I’m frig’n cottage industry for government sponsored shake down artists.
Arab Micro Devices gets a pass!
Give it away
Give it away
Give it away, now.
Give it away
Give it away
Give it away, now.
Give it away
Give it away
Give it away, now.
http://www.youtube.com/watch?v=WESs2U_avdU
SPARKS
...........AND!!!!!!!!!!!
If you try to dump 20% of your gross in a 401K, the banks will reem your a**hole there, too!
HOO YA!
Ya, gotta love it. You just can't make this stuff up!
SPARKS
About ATOM and 945 chipset, it gives a new meaning to the Odd Couple.
Would it have been more difficult for INTC to design ATOM (perhaps in the future) with an IMC? Further, why didn’t they just cut to chase, and build one in initially? They pulled it off with Timna nearly ten years ago.
Wouldn’t the power consumption be reduced further?
Correct me if am wrong, but by the way you guys are talking, and in addition to the high demand, it sounds as if ATOM was the most successful (and lucrative) test vehicle that ever fell out the back door
ITK was the only one who had these things pegged, at least here, anyway. Honestly, I had no clue. If it ain’t cooking at 4 Gig, well, you know the rest.
BTW: If INTC says they’re taping out 32nm, I believe it.
SPARKS
"Aren't the Dems supposed to be against these loopholes?!?!?"
Nah, they're just "spreading the wealth".
Ha Ha Ha, come Jan 1, welcome to my world boy's!! Flush twice, it's a long way to 'Joe The Plumbers' cesspool. HA HA HA
SPARKS
In a day when investors went shopping and the market was up 10%+, INTC was up as well what happened to AMD. Its a couple bucks and change and should be ripe for a bounce and yet the investors avoided it while bidding everything else, whats up with
Arab Managed Devices
Would it have been more difficult for INTC to design ATOM (perhaps in the future) with an IMC? Further, why didn’t they just cut to chase, and build one in initially?
Moorestown, due out next year, does just that. It is projected to cut active platform power by 2x (compared to the Menlow platform, which does not use the 945 chipset) and reduce idle power by 10x.
Moorestown is what Atom is really about as others here have said. Silverthorne was just to prime the pump. No one expected it to be this big, this fast.
If Intel is presenting at IEDM, it would appear 32nm is relatively on schedule. First tapeouts/Si for logic maybe Q1-Q2'09?
I think you're off by about half a year. =)
"No one expected it to be this big, this fast."
I beg to differ, you did.
Besides, take some credit, it beats getting your teeth kicked in,--------and------- that happens to the best of us around here
SPARKS
Hey, fella’s! What is that LOVELY piece of hardware in the picture? It looks like a cross between Woods Holes’ ALVIN and a huge pizza oven.
I think anonymous was right about “six months off =)”.
http://news.softpedia.com/news/In-2009-We-Will-Have-Chips-Made-Using-The-32-nm-Process-14022.shtml
SPARKS
Ran across this comparison of K8 to Core2 and thought I would post the link here.
Sparks, not exactly sure what that beast is. It looks kind of like a CVD tool, but it doesn't look quite right. That is my best guess anyway. It certainly isn't one that I walk past on a regular basis.
Sparks - that Softepedia article is filled with misinformation, so I wouldn't put too much stock into it. Hard to tell the tool in the picture, I agree with ITK - looks a little "one off", might be a CVD tool.
Intel Officials have announced that the 32 nm production technology is essential for the company, and that they want it to be mass adopted beginning 2009.
Mass adoption of 32nm in beg 09? Must mean beg'10? (with initial product in late '09)
the 32 nanometer production technology uses very thin structures and needs special tools for manufacturing, like the extreme ultraviolet lithography.
First I like the phraing "the" extreme ultraviolet litho (kind of like "the" Ohio State University). Secondly, 32nm will use immersion litho, not EUV. Third, most new tools require some sort of new technology/tools - the article makes it sound like 32nm is something different than any other tech node.
Oops, I meant to say 32nm will use the immersion litho, not the EUV. :)
Speaking of the immersion litho, here is an article on double patterning.
I have to offer a couple of caveats.
First, I've only had time to skim the article.
Second, I don't do litho.
In fact I've worked hard to avoid any association with that process, so I can't speak to the validity of the information in the article.
Nehalem benchmarks should be coming soon?
So will it be a trick or treat?
Somehow I think the men in green will be in bigger trouble very soon
"not the EUV. :)"
I knew the EUV part wouldn't slip past you.
From ITK”s link:
“EUV is absorbed by everything: glass, oxygen molecules, water—you name it. Not only do you need optical assemblies made out of something far more exotic (read: expensive) than glass, you also need to carry out the entire process in a vacuum. The ultrasmall wavelength of EUV makes extremely small defects on optical surfaces and the photomask critical—and both hard to detect and eliminate. EUV sources are also too weak to produce chips at the rates needed for commercial success.”
“G”- Did you write this??? If not, this guy is more skeptical of EUV than you are! (Maybe)
ITK- Excellent link! (Robots, et al) It will keep me reading for quite some time. Better yet, it's written so you don’t need to be a Stephen Hawking to understand what the hell their talking about.
It was interesting to learn the optical focal length and feature size limits of immersion. It seems double patterning (double exposure?) has significantly reduced these limits. (I still don’t get the “pitch” part) I imagine it has to do with the vertical angle of each feature, 90 degrees or perpendicular to the substrate being the ideal.
Besides adding additional cost, couldn't they give the wafers another pass? Triple patterning, perhaps? They got excellent result with two passes, why not three?
SPARKS
still don’t get the “pitch” part
Pitch is just the relative spacing between 2 lines (or features). It is not just the space, but also includes the feature as well. So if you had a 5 foot wide line, with a 5 foot wide space and then another 5 foot wide line - the pitch would be 10ft. (you can think of it as the middle of one feature to the middle of the next adjacent feature)
Triple patterning, perhaps?
I think folks have explored this, but keep in mind you have a certain budget (or error) when aligning the first pass to the 2nd pass. With each pass you introduce more error. Suppose your alignment is off 2nm on the 2nd pass and then another 2nm on the next pass. You now have potentially 4nm error on features in the ~20nm range.
I don't know where the limit is and it is also related to how good the alignment is - but there's definitely a practical limit even if you can theoretically print it. (and of course as you mention there is the cost issue)
“You now have potentially 4nm error on features in the ~20nm range.”
Got it, 10% ‘budget’ apposed to a 20% to 25% ‘budget’ on a third pass, too much error, too much money, and not enough yield.
Is this misalignment common to the entire overlay, sort of like an incorrectly aligned head gasket? Or, is it like a Hollywood plastic surgeon, laser in hand, removing age spots, who is afflicted with palsy? (Uniform error (budget) as apposed to random error?)
This is one the more interesting thing about this whole setup, making accurate little pictures on an atomic level, one on top of another. Do you have any links that shows how these (heads, lenses?) throw down the images while compensating for the refractory index of ultra pure water (immersion)?
Or is this some kind of stuff no one talks about outside the bedroom? (Talking about how their tools work)
BTW: Thanks for the new terms ‘Budget’ and ‘pitch’.
SPARKS
“So if you had a 5 foot wide line, with a 5 foot wide space and then another 5 foot wide line - the pitch would be 10ft. (you can think of it as the middle of one feature to the middle of the next adjacent feature)”
Is the space always relative to the feature width? What if I wanted a smaller feature and a larger space so device ‘A’s electrical properties wouldn’t screw up device ‘B’s electrical properties?
SPARKS
In general space and width are *not* related. However...
1. From an area perspective, you don't get the scaling you want if you only shrink the line and not the space. No area scaling = no cost advantage = why bother?
2. From a patterning perspective, it is *always* advantageous to have equal line + equal space (or close to it). There is such a thing called a "process window" -- where you try to account for litho process errors -- such as an out-of-focus and/or out-of-dose shot. (focus is a tool AND topological issue, dose is mostly a tool issue). If you have 1:1 lines and spaces, the pattern's response to a change in focus in minimized.
As an aside, you are also trying to maximize contrast -- such that an error in the mask does not impact the final pattern as much. Maximizing contrast actually means small lines + large spaces are preferred. 1:1 are somewhat less so, and a 1x line + 2x space (1.5 pitch -- or somewhere in between) is generally worst case. So, yes, you trade off focus variability for mask variability. It's a balance, too be sure.
That being said, as you decrease the pitch (ie, from one node to another), your contrast gets MUCH worse, and you run into a fundamental capability. Unless you find a way to counteract it. Hence:
1. I-line to 248nm to 193nm sources. As the wavelength of the source decreases, you can resolve smaller pitches.
2. Better lenses. From diffraction theory, you need a lens to collect large and larger angles of incident light to resolve a particular pitch. If your lens can't be made well, it usually shows up at the edges with lots of non-idealities. Fix those and you can pattern better. The end result is an increase is something called the "numerical aperature" (allowable acceptance angle). In the past, we have shifted from 0.85 to 0.93 to 1.2 to 1.35NA tools.
2a. The theoretical maximum NA is 1.0 (a 90 degree collection angle (sin(90) = 1). How do they get > 1? Use a higher index of refraction interface between diffraction pattern and silicon. Hence, immersion lithography. (nb, 1.2NA ~ 0.85 "dry" and 1.35 ~ 0.93 "dry" in terms of lens capabilities.)
3. More agressive sources. If you really want a given structure to print well, you can shape the illumination to contain ONLY those frequencies that correspond to the pitch you want. Works well for a single pitch, but if you try to print virtually anything else, you run into serious capability issues.
4. Don't shrink the patterned pitch. How then do you get the area scaling? Simple -- do a double pattern. Interlace one set lines a 200 pitch with another set of lines at 200 pitch and you get a 100 pitch pattern. This is a situation where you *would* have a different line+space combination. (ie, 50L150P double patterned would give you 50L50S.) This suffers from the focus issues, above. YOu can read about these details in that Mack article referenced above.
Well there it is. More complicated variables to add to the process stew. Depth of field, focal plane, wave lengths, refractory indexes, aperture settings and angles, and I’m sure there are plenty more. It makes me wonder if there isn’t a bit of ‘McIver’ in all you guys.
Thanks for the peek (and the essay).
Hmmm, it also makes me wonder how can you dial all this stuff in if you ain’t got a FAB to test minor variations in the stew???? One subtle variation in the lens (es) could have a group unwittingly pissing in the wind for months!
SPARKS
Hmmm, it also makes me wonder how can you dial all this stuff in if you ain’t got a FAB to test minor variations in the stew????
The designer's work from the stated capabilities of the foundry. If I say I can give you a 15nm pitch (for example), you (the designer) don't worry about how I do it. You just expect me to deliver features with a 15nm pitch.
You take the all the foundry's stated capabilities (pitch, Ion, Ioff, ldsat, etc.) and design your product within those limitations. Then you send your baby off to the foundry to produce revA-0. Based on how revA-0 behaves, you will make changes to revA-1, or if major changes are needed to revB-0.
For the record, this series of revisions is how everyone develops their products. So what is the big deal about having your design and production under one roof? It is the ability to sit your circuit designers and your process designers down in one room and have them agree to the same limitations at the end of the day.
That can't happen with the foundry model. The foundry provides the manufacturing process and the designers provide the design independent of each other.
As an example, take line density. From the designers point of view, it may be best for electrical timing purposes to have some metal lines isolated while others are densely packed. But this causes polish issues in manufacturing, because dense areas and isolated areas polish/etch at different rates. By having the design groups and manufacturing groups sit down together and reach an agreement on what is manufacturable, you can get good yields as well as a good design. This is what "restrictive" DFM is all about, and it should be the first step of the design process.
It is also the step that is largely missing in the foundry model.
Again, my direct experience with this process is from the PCB industry where in-house design is almost non-existent, so realize that there will be subtle differences in semis, but the basics should hold.
Hmmm, it also makes me wonder how can you dial all this stuff in if you ain’t got a FAB to test minor variations in the stew???? One subtle variation in the lens (es) could have a group unwittingly pissing in the wind for months!
As ITK said -- it's the FABs job to deliver stuff to spec. If it's out of spec, the fab has to fix it. It's part of the overall process capability which leads to design rules. If the design rules are too aggressive, then the fab is going to have a very hard time making a viable part. (ie, you end up with a lot of tricores.. =) If the design rules are too conservative, you end up with VERY unhappy designers who take their business elsewhere.
However, if the *spec* was set on a tool (the first tool) with crappy tool aberations, then you *do* run into these kinds of issues. Tool matching is a beast of a problem.
And, as to your comment, welcome to my world. Part of what I do is to overcome that kind of stuff.
“Then you send your baby off to the foundry to produce revA-0. Based on how revA-0 behaves, you will make changes to revA-1”
“However, if the *spec* was set on a tool (the first tool) with crappy tool aberrations, then you *do* run into these kinds of issues. Tool matching is a beast of a problem.”
Well, there’s the difference in a nut. Take the revisions made to Barcelona, basically peanuts, after successive peanuts. In contrast, take INTC’s (GO) stepping, holy mother of revisions, what a pay off! The absolute overclocking champion of all time for miserable 250 bucks! And the thing is still selling like there’s no tomorrow. In fact, they extended the life of the thing unlike anything they ever made.
Were they better? Was there better communication between groups? Was C2D a better foundation to refine? Did they have better control of the tools? Were they manufactured with a closer eye on tolerances and performance?
It’s probably all of this and more. I’ll tell you what, if I’m paying 5 Million for a litho tool, and I need a dozen or so, it better print the entire Human Genome on a flea’s rectum, clearly around the entire donut, without flaw. And, I’d have my best boys on it every step of the way.
Arab Micro Devices will have no such clout.
“Part of what I do is to overcome that kind of stuff.”
Kudos for dealing with black art/voodoo science of lens manufacturing, alone, not to mention the other parameters/issues---- (“focus is a tool AND topological issue, dose is mostly a tool issue”).
SPARKS
“In order to enable these tight pitches by use of low-cost 0.92NA 193nm dry patterning, innovative processes were developed to produce robust patterning. This is demonstrated by the fidelity of the poly lines in Figure 6. The gate patterning process uses a double patterning scheme.”
“G”, ITK, ---- From this:
http://www.intel.com/technology/itj/2008/v12i2/1-transistors/4-designrules.htm
Thank you, you’ve given me a clearer understanding, and filled in a few gaps. (Forgive the pun). It’s getting easier, once the terms are dialed in.
BTW: The article is chock full of all the previous lessons such as—
Idsat
Idlin
Ioff
high-k first, metal gate-last
The article begins here:
http://www.intel.com/technology/itj/2008/v12i2/1-transistors/1-abstract.htm
Thanks again.
SPARKS
i7 Benchmarks are out
Game over for the boys with no fabs in green.
The reviews indicate a 5-10% advantage in mostly single-threaded benchmarks, 15-30% (or more) in most of the high-end tests. Seems to overclock well.
Anand seemed to enjoy reminding us each time the $284 i920 outran the $1400 Q9770. Nehalem looks to be pretty much what Intel would hope for, a clear improvement over Penryn.
HardOCP gave Nehalem a glowing review (more of a preview really) and added that "all of this was done inside power envelopes that we are already accustomed to. 130w TDP power envelopes are overstated. Even with our i7-965EE showing to be fully 100% loaded we never saw full system power pull over 280w at the wall, and this was with a GeForce GTX 280 running idle."
I know that there are a certain group of enthusiasts who were hoping for a much less impressive showing from a slow and overheated Nehalem. It seems that they will be sorely disappointed.
Sorry for the triple post, but from this review there is an indication that i7 might draw less power than both Yorkfield and even (giggle) Phenom.
The next few days should be interesting as the arguments start to rage. Hopefully there is lots more data, particularly on power consumption and overclocking. I have a lot of reading to do, it seems!
I know that there are a certain group of enthusiasts who were hoping for a much less impressive showing from a slow and overheated Nehalem. It seems that they will be sorely disappointed.
Yup, feel free to go over the fanboyZone side to have some laugh :) they started acting defensive, calling ppl stupid, selectively shows 2 graph out of 3 graph, out of more graphs from the multiple review sites, and the FUD kings Abin continue to spread FUD, and Sci continue to be MIA since the day AssetSmart revealed itself
The question is, which reviews can they point to?
Tom's Hardware? "The last time we saw a performance improvement of this magnitude was when Intel moved from the Pentium 4/D line to the Core 2 architecture. The new integrated memory controller offers much higher throughput and is even superior to AMD’s solution on the desktop.
The Core i7 is going to leave Intel’s rival AMD lagging even further behind. Put bluntly, you’d need two and a half Phenom X4 processors to compete with Intel’s current Core i7 flagship model."
AnandTech? "The Core i7's general purpose performance is solid, you're looking at a 5 - 10% increase in general application performance at the same clock speeds as Penryn. Where Nehalem really succeeds however is in anything involving video encoding or 3D rendering, the performance gains there are easily in the 20 - 40% range. Part of the performance boost here is due to Hyper Threading, but the on-die memory controller and architectural tweaks are just as responsible for driving Intel's performance through the roof."
Granted, the above will be dismissed as Intel "paid pumper" sites. But what do other sites say?
Hot Hardware: "Looking back through the results, it is obvious Intel has raised the desktop CPU performance bar yet again and in a big way. The Core i7 920, 940 and Core i7 Extreme 965 put up some of the best benchmark scores we have seen to date. In a few instances, the Core 2 Extreme is able to come close clock-for-clock, but overall there is no denying the Core i7 represents a significant step up in performance."
PC Perspective: "We basically know what we can expect from HyperThreading based on our discussion above: if the application is capable of taking advantage of more than 4 threads then there will be performance gains on Nehalem ranging from 5-35% on average."
(They do give some hope to the people who feel that single-threaded performance is the most important metric, admitting that Nehalem is around a 5% improvement in single-threaded performance at the same speed.)
Tech Gage: "I think it's safe to say that Intel successfully brought back the same spirit that came with the original Core 2 launch. That particular launch unveiled products that were substantially faster than the previous-generation, and the same thing can be said again today."
Tech Report: "The Core i7-965 Extreme is, by far, the fastest processor we've ever tested, and it seems clear the Core i7 architecture brings with it a general performance increase over the 45nm Core 2 processors it succeeds. We've seen that increase in everyday desktop applications, including the WorldBench suite and several of the latest games."
Maximum PC Magazine: "If we had to describe the Core i7 in one word, it would be monster. The CPU is to benchmarks as Godzilla is to downtown Tokyo.
Take, for example, the Core i7 Extreme 965 versus the Phenom X4 9950 Black Edition. It’s no surprise that the Core i7 throws the Phenom X4 through a couple of concrete walls and right into a telephone pole. We witnessed performance differences of 87 percent, 95 percent, and even 133 percent over the fastest Phenom X4 part."
NeoSeeker (31 page review? 31???): "In all but a couple of tests, the 2.66GHz Core i7 920 handily beats the 3.2GHz Core 2 Quad QX9770 - and in some tests by an embarrasing margin. Needless to say, the Core i7 965 tends to wipe the floor with the QX9770 in memory or processor bound applications, and I think that AMD is in for a nasty surprise on the server side"
And so on and so on... I don't consider many of the web review sites to be the end-all/be-all. Some of them definitely put the "enthusiast" into "hardware enthusiast" and it shows in their reviews. But the consensus seems to be that Nehalem is a small upgrade in single-threaded apps over Penryn, and a pretty large upgrade in multi-threaded apps, and potentially a massive upgrade in apps that benefit from the new features (IMC, Hyperthreading, etc). Now we wait on Shanghai to see what it brings to the table.
I have to admit, the single-threaded performance doesn't bother me. Most software that requires lots of processing power already support multiple cores and threads. How many single-threaded apps are out there that cannot be converted to support parallel processing? How many of those are having trouble running fast enough on 2.0-3.2GHz processors? I suspect that this is a non-issue.
I have to admit, the single-threaded performance doesn't bother me.
actually, if there is any small improvement with the single threading, Turbo mode would help tp pull it further. The fanboyZone kept hold to an Intel foil which stated that the i7 would be at Torbo mode as long as 1 minutes ... however, not sure if those ppl in the fanboyZone thought about it or simply ignore it ... says .. in 10 moniutes, how many of the '1 minute' the i7 can enter the turbo mode given a singlethread apps is running, then what about in an hour? :)
I think Ed at OC had it right... Nehalem is designed to be a 'luxury line" CPU - it is designed for heavy workload and multithreaded aps. For those not needing this they can stick with Penryn for a while (and perhaps later migrate to integrated CPU/GPU version of Nehalem?). And for the true low end... Atom.
I do wonder what the strategy is on mobile though - it looks like Nehalem has done quite a bit on idle power, but a lot of the improvements on Nehalem are not that critical on mobile (at least mainstream mobile). I wonder if mobile just waits for the CPU/CPU version.
But this thing should be a beast on the server-side where HT and the bandwidth can truly be utilized.
It does seem like HW is outpacing SW these days and I wonder if the SW will catch up in an efficient manner, or just trend like Vista, where the SW just consumes more computing power, but doesn't do much with it. While I see the need for continued computing power on the server and high performance areas - I do wonder how much computing power is truly needed on a notebook or desktop and are we approaching the point of diminishing returns?
Pointer,
Turbo mode seems to be dependent on how much cooling you can provide the CPU. Anandtech said:
http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3448&p=10
Now these tests were conducted on an open-air testbench with an aftermarket cooler by Thermalright, we wondered what would happen if we used a retail Intel HSF and stuck the Core i7 in a system with a Radeon HD 4870 and a 1200W PSU. The CPU actually ran a lot warmer and Turbo Mode never engaged, pretty much as expected.
So if the cooling is inadequate (like the retail HSF according to Anandtech), turbo mode may never engage. Considering this is a quad core capable of 8 threads, chances are there will be more than one load on it.
Sparks, you're sure to snap one of these up ASAP, perhaps you could shed some light on this with your experience.
Turbo mode seems a neat feature none the less, provided one can cool the CPU adequately.
“Anand seemed to enjoy reminding us each time the $284 i920 outran the $1400 Q9770.”
Ah, yes---and no. Perhaps, at stock speeds it will. However, it won’t outrun a QX9770 at 4 gig. :)) That’s why you pay the extra rubles in the first place.
“Sparks, you're sure to snap one of these up ASAP, perhaps you could shed some light on this with your experience.”
True, true, I’d love to get my greedy fingers on one. I’m rather apprehensive about life without FSB. Besides, a Disney Trip in two weeks put the kibosh on that purchase------for the time being.
With QX9770’s excellent performance (and remarkable stability), crossing over to 965 will not be so easy. My appreciation for this chip surpasses my original old flame i486-33. Remember how well it ran Win3.1? It’s that good, and so much more. QX9770 is still a wonderful mistress, and thanks to you guys I know her inner most secrets.
I hate to say this, but I’m satisfied….
……for now.
SPARKS
Tick Tock Tick Tock
Is that HighK Metal Gate, QPI, and IMC that I hear coming.
The winds are looking like they are brewing for a nasty storm for Shanghai. Isn't it appropriate that AMD should name their next processor Shanghai. It is going to get Shanghia'd, LOL
Looks like the server space is going to be ruled by INTEL as they own everything else already.
Scientia and all them other fanbois are off in their mommies basement in a big timeout I think.
Was I not there telling you all that years ago?
Deep pockets, superior silicon, and good engineering can overcome the worst strategic mistakes.
Bad technology even with Arab money ain't going to catch them.
anon: "Isn't it appropriate that AMD should name their next processor Shanghai. It is going to get Shanghia'd, LOL"
Good thing they didn't name it Bangkok. :)
"the worst strategic mistakes."
I beg to differ.
Now we just can't say INTC did anything wrong, it was just a bit short on doing things right,.....ahem.
SPARKS
While I see the need for continued computing power on the server and high performance areas - I do wonder how much computing power is truly needed on a notebook or desktop and are we approaching the point of diminishing returns?
That all depends on whether or not the software wizards can actually find one of computing's holy grails, voice recognition. If they can make it work reliably and quickly it will consume plenty of cpu cycles. And the demand for more power will go on for a while longer.
Any intelligent fool can make things bigger and more complex... It takes a touch of genius --- and a lot of courage to move in the opposite direction. (Einstein)
SW (and perhaps HW?) is moving in the wrong direction. With computing power so 'cheap' why struggle to write an efficient ten thousand lines of code when you can do it in a million lines?
Excess anything (computing power, money, resources) tends to breed complacency and inefficiencies and outside a few select niches, that seems to be where things are headed.
MS is finding this out the hardway with Vista.
BEGIN RANT:
Unfortunately, the US gov't has never seemed to learn this lesson. In years of excess, instead of addressing long term issues (debt, true reform of critical programs), the gov't spends like drunken sailors, and in times of distress the gov't spends more to "help". Heads we spend more money, tails we spend more money - who says politics isn't easy?!
That is ultimately why increasing taxes (whether it's on the rich, the poor, the evil companies or the successful companies) will never truly fix anything - it just feeds the power base in Washington as they can justify their existence (for re-election) with new programs created and the amount of increased benefits.
End political rant, as I prepare to go off and vote tomorrow and count the # of times I hear "americans have spoken" and "change". I would turn it into a drinking game, but I don't want to die from alcohol poisoning!
“G”, if I may be so bold. We’re in a rut. Humanity has turned in on itself. Physics has run the gambit of every ridiculous theory from string theory to cold fusion. And yet the idiots, for all their brilliance cannot come to terms with magnetism and gravity.
What’s the problem? We are stuck on this rock, in an enclosed biosphere, leveraging weapons, food, land, and energy against one another. We’ve turned inward as apposed to turning outward. Our science is good enough and impressive enough to kill ourselves 100 times over, but we still can’t get off this bloody rock, en mass, and turn outwards to do some real human exploration.
We’ve stopped exploring. Humanity is looking up there own asses as apposed to looking up at the stars
Take our vehicles, state of the art rockets, if you will. Nothing really has changed since the Chinese lit the first fuse! Paper tubes gave way to Titanium and gunpowder gave to LOX, LH2. No mater how you cut it it’s the same old Newtonian Physics, big, ugly, expensive, dangerous, and inefficient.
Our electronics for all its magnificent achievements and atomic miniaturization are still basic on/off relays, a very sophisticated compilation of resistors, capacitors, and active devices, no doubt. But it’s still the same physics Faraday stumbled into in his lab, power on, power off. (Throw in a few magnetic field lines for shits and giggles. That was the real bonanza.)
Why do I take these two examples? To breakout and resume the human quest we need to do this:
Electron > Propulsion
and/or
Proton and Neutron > Propulsion
Not this:
Electrons > guidance
Protons and Neutrons > Warhead
But we have, and what a waste of valuable resources and energy.
“God may be mysterious, but never malevolent” (Einstein)
For now, unfortunately, if I may quote a lyric from a song,
“With, without, and after all, that’s what the fighting’s all about.”
Currently, overall, this is the human condition and we need a technological revolution.
Personally, I have plenty of toys, and none are weapons of any sort, to keep me occupied, until that time.
So let them take from the rich to feed the poor. We, thank GOD, have our intellect, and I am proud to know you.
SPARKS
Hey, JumpingJack! Ed at overclockers deleted your excellent comment! You've hit the nail on the head, and the chicken shit deleted it.
Ed's stock just dropped to zero. However, it wasn't that high to begin with, peh.
SPARKS
Hey folks, it's been a while. I've been lurking, just haven't had time to post, and probably won't for a while either. Things are busy if you know what I mean. Much to my surprise, I've been informed that I'll be doing a tour of duty in Oregon some time next year, so I have the exciting task of training my replacement.
Anyway, looks like the curtain has finally dropped on Nehalem. For the most part it's looking very nice. I've seen a lot of bickering and whining on all the tech forum's, but I guess that's to be expected when something is as hyped as this. Overall, I think expectations were very high, and on some fronts, Nehalem met them very well, but on other's it was just OK, but I guess it's just gamer's that want to bitch the most ;)
Reviews seem to be at a consensus that it is a very compelling platform, but may not off much unless you're a power user with some money to burn. I guess that's just the space we're in now in computing, We're nearing the point where improvements in single threaded IPC are plateuing and we need software to catch up to the HW as other's have alluded too.
I honestly don't understand the concern over gaming. In the gaming benchmarks, the video resolution had to be dropped significantly in most games (as low as 1024 x 768, versus 1920 x 1200 or higher) in order to remove the video cards from the equation.
If you have to drop the screen resolution almost in half in order to measure the CPU's effect on a game, how important is the CPU at that point? Oh no! Nehalem is only 5% faster than Penryn for gaming! I'll only get 357 FPS instead of 349! *cut cut cut*
It's like complaining about synthetic benchmarks and then insisting that the only legitimate way to gauge CPU stability is by running Prime95. :)
Much to my surprise, I've been informed that I'll be doing a tour of duty in Oregon some time next year....
So you landed that spot on 32nm you were looking for. Congrats!
Hey, JumpingJack! Ed at overclockers deleted your excellent comment! You've hit the nail on the head....
Sparks, you just gonna leave us hanging like that. You're a cruel man.
Blogger Tonus said...
I honestly don't understand the concern over gaming. In the gaming benchmarks, the video resolution had to be dropped significantly in most games (as low as 1024 x 768, versus 1920 x 1200 or higher) in order to remove the video cards from the equation.
well, look for your answer here when GPU is not much of the bottleneck:
http://www.guru3d.com/article/core-i7-multigpu-sli-crossfire-game-performance-review/
Good heavens, that is just sick.
How ironic, that Nehalem may be the answer to NVIDIA's prayers. :)
Blogger InTheKnow said...
Hey, JumpingJack! Ed at overclockers deleted your excellent comment! You've hit the nail on the head....
Sparks, you just gonna leave us hanging like that. You're a cruel man.
Yes, I'm interested too. Please post it if possible
Ed is an interesting fellow. He's like a guy always yelling fire in a crowded theater. I usually get a good laugh from him. His previous editorial on Bloomfield OCability was hilarious. He's whining that the 920 (w/ 2.66Ghz stock clock) will be limited to OC's of 4Ghz due to only being able adjust BCLK and not multipliers. I don't know about you, but when is a 1.333Ghz overclock something to be bummed about on a sub $300 part.
I think anything that doesn't cost under $100 and overclock 200% is going to upset Ed. :) His viewpoint is centered around getting as much for his dollar as possible, and Nehalem is a bad fit with Penryn around.
At least he seems to have some grasp of market dynamics and business goals and he understands why Intel pricing Nehalem the way it does isn't some sort of victory for AMD. Watching some of the fanboys trying to wring a silver lining out of the cloud Nehalem is casting over AMD is... fascinating.
PS- good to see you posting again, Ortho. Your comments are usually interesting to read and both practical and educational.
Whoops, I dropped the ball. I incorrectly assumed that most folks here browse over to Ed’s once and a while. I was wrong. Ouch. That said, obviously, JJ did, and wasn’t pretty
The short story is Ed was moaning and pissing about Nehalem’s inability to clock over 4 gig. JJ jumped all over him and basically asked what hell he was looking for, and wasn’t a 1.3 gig over clock enough. I agree. Ed’s standards are set too high, especially on the bad boy’s first time out, and I don’t think he’s too happy about AMD.
He called my beloved INTC a “monster”. I would have preferred a big successful monster.
Sorry, here’s the link. Let’s hope JJ can fill in the details on his next post.
http://www.overclockers.com/index.php?option=com_content&view=article&id=4259:bloomfield-intels-next-flop&catid=57:processors&Itemid=4263
SPARKS
Orthogonal! Lassie, come home! Have you been reading about my exploits and wonderful affair with QX9770/X48? What a darling! 3.84 Gig for a day to day, bread and butter chores. 4.27 for the bragging.
But that’s nothing compared to the instantaneous experience. Imagine, you’ve just finished playing Crysis, and you’ve discover you’ve done a virus scan and defragged without a single glitch!
(nice job with the FSB, 1800 native)
DA BOMB!
SPARKS
Tonus said...
Watching some of the fanboys trying to wring a silver lining out of the cloud Nehalem is casting over AMD is... fascinating.
Thank you Mr. Spock :) I must say that UAEZone commentary seems fairly muted at this point, as if they are in a bit of shock. Guess they weren't expecting the onslaught of reviews showing Nehalem slaughtering the Phenom, and now it's mainly "wait until Shanghai reviews come out"...
However, an Intel engineer posting over on XCPUS.com says the latest reports he has seen show abysmal yields for AMD's 45nm, in the range of 25 - 30%. So I guess it'll be a paper launch. Oddly enough, some of the more extreme doofussers at UAEZone claim Nehalem will be a papier-mache launch on Nov. 17th :)
At least the "paid pumper" comments have all but disappeared for now. All in all, the recent spate of bad AMD news (lost market & revenue share, fab spinoff, no Bulldozer, no Atom competitor, Puma falling flat, etc) must have seriously depressed the fanbois in green.
Personally I'd wait a few months until the platform stabilizes and maybe even a new stepping appears, before buying - some of the reviews show performance very dependent on the mobo used, indicating BIOS or other issues. And larger L2 and L3 would be nice although we probably won't get that until Westmere at 32 nm..
I think that the "paid pumper" complaint has been replaced by what amounts to "no one knows how to properly benchmark a CPU anymore."
There does seem to be a bit of acceptance of the current situation, though I think that it is driven by a mistaken assumption that the high price of Nehalem represents an opportunity for AMD. I think that what it represents is Intel acknowledging that the situation is progressing as they would hope-- instead of Nehalem replacing Penryn, it becomes a high-end family of CPUs for a while, bringing in extra profits by occupying a niche that they have all to themselves.
If Deneb or Shanghai give Intel reason to pause, they can then shift Nehalem downward as needed in order to cover any price/performance categories that they want to maintain a lead in. I think they would only really sweat if one of those new architectures surprises them with performance that surpasses their best. But that seems unlikely. Intel appears to be sandbagging a small bit with Penryn and Nehalem, while AMD is (if I am not mistaken) planning on introducing Deneb at 2.8 and 3.0 GHz.
I don't think that will make Intel sweat. It may not even make them blink.
According to Fudzilla Shanghai may be out in 10 days from now. So I expect some comparisons to i7 in a couple of weeks. Although it seems AMD doesn't respect AMDZone with it's 5 readers enough to send them any test rigs :).
We all knew i7 was going to kick serious butt. It was a forgone conclusion.
As for me, I'm waiting for the real benchmarks, the ones that Nehalem was design for, scaling and HPC.
2009 the great server assault.
SPARKS
While Nehalem looks pretty good, realistically it will have very little impact on Intel's bottom line over the next year or so.
It should make the greatest inroads in server (eventually) - but that effect won't truly be seen until 2P, 4P and 4P+ come out and even then, these are still relatively small volumes (compared to desktop/notebook). In desktop you are talking about this chip pretty much taking the place of high end Penryn's - a market where AMD doesn't really compete these days. And given the Nehalem vs MCM Penryn approach, this may actually have a slightly negative impact on Intel's bottom line short term. It only helps if it pushes Penryn further down the foodchain and eats into the "budget quad" market.
Now before folks jump down my throat - I think there are 2 key things here.
1) Intel's implementation of IMC/QPI appears to be solid - there is not going to be a long learning curve over several product cycles as some AMD fans assumed/hoped.
2) Nehalem sets the stage for the integrated CPU/GPU - this in my view is the true key in the mobile space which is where the money is today (and will only continue to grow vs desktop and server). It also may be significant in the corporate desktop space (if atom doesn't grow into that market)
The folks on other sites, thinking "I expected more" need to keep in mind that Core2 HAD to be good because Intel was behind. Nehalem HAD to avoid being a complete failure on the key architecture changes and set up the future - at first glance it appears to have done this.
And if you look at the idle power gains Intel has made vs Core2 - it appears well positioned for the mobile space if they can implement CPU/GPU well.
Over the years I have come to respect Vince Freeman’s product reviews. His unbiased objectivity has always impressed me, and this review is no exception.
At $1k i7 965 is shaping up to be a bargain for an XE chip these days. Vince said Core i7’s launch is a resounding success, from the numbers and his analysis, I tend to agree.
http://www.sharkyextreme.com/hardware/cpu/article.php/3782516
SPARKS
Looks like Ed had to eat a bit of crow on his latest article, which was a frantic call to overclockers to make Intel pay for putting in an effective limiter on overclocking. It turns out that the option to throttle speeds down based on voltage can be turned off in the BIOS, and Intel made sure to point this out.
He didn't completely give in, grousing that this is still bad because the feature is there. Fair enough, but anyone who saw the content and tone of his original article has to stifle a giggle. He seemed to consider it a matter of such great import, that he implied that people who chose not to take a stand might want to take a look in the mirror and ask themselves some hard questions.
The embarrassing part is that the link he used as a lead-in to his rant was from Tom's Hardware... the same site where the information about the BIOS switch was also located. Had he simply done some additional reading at the same site, he would've had a much different article to write. I wonder if he'll step back and look at the situation objectively, or if he'll go off on another rant, effectively biting off his nose to spite his face.
HardOCP slams 'some people/sites' (gee, wonder who that could be!) for overreacting and spreading doom and gloom about the chip's ability to overclock properly.
They took an i7 920 (2.66GHz) and overclocked it to 3.8GHz, then ran the cores (and HT links) at 100% for more than 10 minutes without having the CPU throttle down. This was, of course, with the turbo boost option turned off.
They called it one of the easiest overclocks they ever got. They bumped the vcore to 1.35 and the RAM speed to 1520MHz and presto... a 1.14GHz OC on a $284 chip. Heck, they had to ramp down the fan speed on the heatsink in order to get the temperatures up to 85/87C so that could show that it wouldn't throttle down the speed.
ITK- You know buddy, I can’t stress enough how right you were about these little monsters.
Kudos
SPARKS
http://www.xbitlabs.com/news/cpu/display/20081103075533_Intel_Atom_Helps_to_Boost_Microprocessor_Shipments_to_Record_Levels.html
Tick Tock Tick Tick Tick TOCK!
Yes we will hear more complaints about some guy named Lex from some anonymous weenie.
Here is what we see on Nehalem on 45nm with HighK/Metal Gate. With a year lead and a far superior process INTEL is now competing with itself. I found these juicy and lovely quotes from Tom’s Hardware a well know INTEL fanboi site.
“The performance comparison with long-time rival AMD’s offerings is nothing short of painful.”
Painful, how so? It is fully what I expected 4 years ago. INTEL is right on track with my predictions and expectations.
“ The fastest Core i7, the 965 Extreme, is 64 % faster than AMD’s current flagship CPU, the Phenom X4 9950 BE. Across our benchmark suite, the AMD processors never placed better than towards the lower middle of the field, tending instead to fill the lower spots”
“Meanwhile, AMD falls even further behind in the x86 performance rankings, leaving price as its only counter-argument.”
Realistically AMD with Shanghai and whatever comes after that is not going to close any gap. Matter of fact they will lose the high end come 09 and they got NOTHING that can help. They simply have NOTHING. Of course INTEL could still trip and fall on their face at 32nm. But I suspect the taste of victory is too sweet and this time around the management is watching execution like a hawk, game over!
Can’t wait to see how Shanghai does and after that the spin that comes about the next thing from AMD. As any earlier poster noted. AMD should have named their processor Bangkok, LOL AMD at least has a sugar daddy letting it to continue to flush wads of cash down the toilet. The sugar daddy would be advised that it would get more ROI building a dog food factory then a chip factory in NY, but alas its Arab money and nothing makes me happier to see it spent creating jobs in NY as they need a little employment as they got a lot of unemployed shell game experts over there, perhaps they can move wafers for Arab Micro Devices in Luther Forrest. What is that rumor I hear about AMD yields on 45nm. But they are very happy and its right where they expect it to be is what their executives say. What, only a year behind and completely missed the boat being old slow and power hungry SOI with SION.
Saw an interesting article on EE times about AMD 45nm and had this thought on sugar daddy foundry ambitions. “ The competiveness of the Foundry Company will be based on technology excellence” Throwing money at this while TSMC, UMC, and Charter are seeing the bottom drop out of their leading edge business is simply to funny. People who buy this smoke and mirrors I say open your eyes! Already the foundry share so much similar base from the IBM club that I fail to see how Arab Micro Devices offer any competitive advantage to the Qualcomm, TI, Nokia’s of the world over TSMC. To think TSMC is lagging on 45nm and AMD has some advantage is to completely lose sight of why TSMC hasn’t yet ramped 45nm. Lastly, Arab Micro Devices, if you have the whore’s track record to consider isn’t exactly the resume you’d want to feel comfortable picking the Arabs Terrorists versus the Chinks.
Gents- I posted this at the INQ in response to this:
http://www.theinquirer.net/gb/inquirer/news/2008/11/05/amnesty-aready-making-demands
“Ah, yes the Sun never sets on the British Empire. No single country in history has left more atrocities in its political, imperialistic wake than Great Britain. The entire continent of Africa, along with India, Pakistan, Indonesia, and your greatest work, Iran are just a few examples of your penchant for Manifest Destiny, and arrogant English superiority. England was the model of white supremacy in which the after effects are felt worldwide to this day.
I’m sure Amnesty International UK is quite good at spotting so called racial “abuses”. You did, after all, invent Apartheid which I’m sure our new President Is well aware of, as well as the risk the terrorists currently in custody present to the security of the UNITED SEATE OF AMERICA, Great Britain’s single most LOYAL ally.
Keep this is mind as you try to deal with your own Moslem extremist within your population presently, while we deal with the aftermath of the 911 attacks to our nation. God forbid a commercial jet should fly into Parliament and you have to deal with the loss of friends, family, or coworkers. I mean that with all sincerity. This is still a powerful threat to this day.
You are a computer industry rag, and not a political forum vested with journalistic entitlements. Your inherent expertise in the industry is marginal, at best. Your combined experience and knowledge in international relations is three tiers below the likes of Susan Saradon which our new President was extremely wise not to embrace during his successful campaign.
Stay away from trite political commentaries. You’re simply not that educated and not that smart, obviously. Stick with industry trash, at least it’s entertaining, especially when you’re wrong. You won’t be ‘dancing in the isles’ of a hijacked 737, that for sure.
SPARKS”
I guess they can’t beat up INTC, so they’ve found a new venue.
SPARKS
I guess they can’t beat up INTC, so they’ve found a new venue.
At the risk of dating myself, I was in England at the time Reagan was elected for his second term. There was a serious concern among the British that his victory almost guaranteed WWIII. As it turned out, his policies brought about the collapse of the Soviet Union, not another war.
So don't take anything the British say about a president elect too seriously. For all that we speak the same language, there is a huge cultural gulf there that leads to a stunning failure of the British to really understand how America works. In fairness, I suspect that this cuts both ways.
Nehalem sets the stage for the integrated CPU/GPU....
I agree with this assessment completely. In fact, I think it goes even further than that. Nehalem has been described as a "modular" architecture. It was designed to allow processor "blocks" to be swapped out for "blocks" that offer different functionality.
I think that this is a key advantage that Intel's approach offers over AMD's fuzion. As I understand it, fuzion is specifically focused on CPU+GPU. Nehalem's architecture offers this option plus other possibilities.
Intheknow said...
think that this is a key advantage that Intel's approach offers over AMD's fuzion. As I understand it, fuzion is specifically focused on CPU+GPU. Nehalem's architecture offers this option plus other possibilities.
Hmm, now I'm wondering if there will be some sort of huge performance increase if Intel replaces an i7 core or two with Larrabee and connect it via QPI. I.e., the old "whole is greater than the sum of the parts" gestalt thing. Wouldn't surprise me if the geniuses at Intel had another ace up their sleeves. And no, I mean geniuses, not some Bozo from Apple's Genius Bar :)
Nehalem has been described as a "modular" architecture. It was designed to allow processor "blocks" to be swapped out for "blocks" that offer different functionality.
This would seemingly be the next evolution - put some fast cores with some workhorse cores and some graphics cores and some other specialized cores. Intel has spent a lot of time and research on high core counts between Larabee and some of the pure research terascale activities.
In fairness AMD is thinking the same way - they have talked about similar ideas, but it seems like while they were talking about it, Intel was working on the same (and just not publicizing it). No doubt you will hear 'Intel copied this' cries from the loons but A) There is no way of knowing who started what first and B) Who cares?
I think AMD's fusion will be similar - the last I had heard on it was it would start with an MCM approach and then go fully integrated on the next gen. Of course dates on this are nebulous (I think end 09 was the last I heard on the MCM design). I also found it funny on "another" forum there is no talk about MCM for this being "glued" or and inelegant approach! (For the record I have no problems with AMD's approach if it works)
I also found it funny on "another" forum there is no talk about MCM for this being "glued" or and inelegant approach! (For the record I have no problems with AMD's approach if it works)
What, no "doublecheeseburger" jokes?? :)
And here I thought the big bonus was the IMC on the integrated die... At least, that is what the AMD Genius Bar bozos said over at UAEZone.
“AMD is boasting unbeatable price/performance per watt for its new 45nm server parts.”
Please, someone, tell me that I don’t some unexplored chemical imbalance; you know the Deja Vous feeling you get when you hear or see something uncannily familiar. Was it a dream or a lack of vital nutrients?
Fudzilla is pimp pumping AMD on the eve of the November 17 Nehalem launch.
I have a formula/equation for all you industry experts. It’s called Sparks Law of Diminishing Hype……..
Price and performance per watt= ((Dog x horseshit)*(Pi*(J)) ^2)
Where (Pi*(J)) = Circle Jerk
http://www.fudzilla.com/index.php?option=com_content&task=view&id=10327&Itemid=1
And, this is after a layoff of 500.
SPARKS
I can't help but think of a boxer who has endured a horrible beating for 12 rounds and after the fight is over and he has lost a unanimous decision, he boasts to everyone about the fact that he wasn't knocked down a single time.
His price/performance per watt was off the charts. :)
Post a Comment