"AMD today reported first quarter 2008 revenue of $1.505 billion, a net loss of $358 million, or $0.59 per share, and an operating loss of $264 million. These results include an impact of $50 million, or $0.08 per share, from ATI acquisition-related charges. First quarter revenue decreased 15 percent compared to the fourth quarter of 2007 and increased 22 percent compared to the (abysmal) first quarter of 2007".
“A seasonally weak first quarter (natural trend therefore beyond AMD's control) was amplified by a challenging economic environment for consumers (people don't have money to buy PCs) and lower than expected (meaning awful) revenues of previous generation products (meaning obsolete), resulting in lower than expected revenues in all business segments. However, we are encouraged (interesting choice of words, why not speak in volumes like "overwhelmed") by the market acceptance of our Quad-Core AMD Opteron™ server processors as well as our new chipset and graphics offerings,” said Robert J. Rivet, AMD’s Chief Financial officer. “We remain committed to achieve operating profitability in the second half of the year, driven by our portfolio of new products and platforms (which NEW product?) and aggressive restructuring programs (meaning layoffs).
It appears as if AMD is back to square one. Not surprisingly, a product line-up of buggy, handicapped and obsolete products guarantees nothing else but massive losses. AMD may pretend all they want that they're giving what the customers want, the latest financial statement says otherwise. It is shocking that they can claim to be having a difficult market environment when their direct competition is thriving. I would like to think that they're lying and know what the hell is going because it sounds as if they're not living in reality.
What is surprising is that AMD admitted to the server share loses and the massive decline in overall unit shipment. Other alarming trends in their numbers include margins going south from 44% last quarter to 42% in Q1'08. While revenue declined $265M compared to last quarter together with unit sales, MG&A increased by $20M! The graphics division which was supposed to have a strong product lineup (at least relative to Computing) also made a loss.
To fix all their problems Hector once again pulled out the old (one year old to be exact) black rocket called "asset-smart". While I could have sworn hearing laughter through the muted telephones, at least one analyst was too pissed to mention that he's getting tired of waiting for the "wait-and-see-if-we-make-it-to-break-even-asset-smart" strategy. Of course, like always Hector said he will make an announcement "soon". With the way AMD puts emphasis on asset-smart whenever they're in trouble and less when they're doing better, it is beginning to look like asset-smart really is filing chapter 11. People who think Hector isn't smart don't know what they're talking about.
4.17.2008
AMD's Q1 2008 Earnings (or lack of) Report
Subscribe to:
Post Comments (Atom)
126 comments:
"We remain committed to achieve operating profitability in the second half of the year " ...
When did second quarter (analyst day meeting) become second half?
Sorry had to take a double-take ...
from the frontpage ticker:
http://ap.google.com/article/ALeqM5gXvPINxnEZQB5TDcsdvrrmfnPtRQD903UK9O0
Headline: "AMD posts 1Q loss, meets Wall Street's lowered expectations"
Wait, wait... concensus on the street was between 0.51 to 0.48 loss per share (depending on the news you read) before earnings ... AMD lost 0.59... that appears short to me.
Of course, 'strip out the one time charges'... ok, so that is a good argument.
However, strip out Intel's onetime charges on Q1 and the actually beat Q1 07.... why isn't that reported in such a manner?
First off - people need to throw out the YoY comparisons as AMD had all sorts of shenanigans between the Q4'06 dumping and the Q1'07 glut... so the comparisons look favorable because AMD front loaded that quarter's revenue into Q4'06 to help with the financing they were doing at the time. So a 22% increase is a bit misleading.
It is humorous how Q2'08 (operating) break even has now morphed into H2'08 and noone (except the poster above) seems to have focused on this - keep in mind this is a mere 4 months after their forecast at analyst day.
What this tells me as an investor is replace whatever folks are saying with 'blah, blah, blah... we're just winging it and will provide a more realistic update 1-2 weeks before the next earnings call as if we didn't do that we would potentially face stockholder lawsuits.
What is curious is that apparently gross margins improved significantly, yet AMD saw almost nothing on the bottom line - that is real disconcerting.
I do see a giant financial pyramid scheme about to happen... it started with the swap of the Morgan Stanley loan with convertible notes that will never convert to stock (I think it needs to get over $25, maybe even $40 share price for the conversions to kick in) - man I wonder about the people who bought those notes are thinking now? This swap let them accept private money from Dubai and not worry about having to use it to pay down debt (as the MS loan terms required)...
Now I see another shakedown of Dubai (hey you want to get some of your inital investment back? you need to spend more). AMD will then use their influence in NY to make sure this doesn't get snagged by Congress (remember the Dubai port outrage?). Then AMD may play the monopoly card and say if they don't get the money they will fail so the US has to let it go. They may also try to issue new stock on the spinoff (but AMD majority controlled) foundry - this would be further dilution for the current stockholders, but hey that doesn't seem to be Ruiz' priority right now!
The difference between Spansion and this new foundry plan (if true) is that AMD was allowed to eventually sell their ownership stake in Spansion and cut their losses (and they did so at a significant loss). They won't be able to do this with the x86 license hanging over their heads. Also I'm not seeing how the foundry will be able to get ANY money for the needed fab expansion/conversions - keep in mind this thing will be operating at a loss for quite some time.
So this will be interesting - AMD will now make the argument to Wall St and investors that by cutting AMD in half, it is worth twice as much! Once thr foundry assets are sold off though I wonder what the real value of 'AMD design' will be - is their any scenario where it could be worth more after shedding virtually all of their tangible assets (outside of design IP)?
And if they dump all the debt onto the foundry this will basically offset most of it's actual assets, so how will they entice people to pay more money for the 49% of this they plan to sell?
In a sane world this type of a deal would be dead as soon as it was suggested.... however I actually see someone taking the hook (yet again) on this and sinking money in. And if this becomes a flaming disaster in a year (which it will - because why is the fab after operating at a loss for 6 quarters going to magically become profitable?), then I see AMD crying their way out of the outsourcing restrictions on the x86 license and selling their stake. The only question is when does Ruiz bail out, and how big will his parachute be?
And as a former NY'er, if AMD does somehow build in NY, will Schumer and Clinton be as outraged with Ruiz' eventual parachute and call for Senate hearings on it like they did with other CEO's? (Sparks, this was a rhetorical question - I think everyone knows the answer)
“And as a former NY'er, if AMD does somehow build in NY, will Schumer and Clinton be as outraged with Ruiz' eventual parachute and call for Senate hearings on it like they did with other CEO's?”
Go ahead, keep it up, you sadist. You didn’t even have to type my name; I knew who you were addressing.
Unfortunately, in these maters the outrage always comes a little too late, after all the taxpayer money has been spent. Given AMD’s balance sheet which shows only 1.9B in cash and they are warning of a 2nd quarter loss, I hope to god someone in power in this state can read a financial report or the Wall Street Journal.
Those morons in the state legislation are too focused, too shallow, too myopic, and too consumed with looking good after spending tens of millions to back out now. This is the scariest part, and this is especially true after a 15% layoff from now to September of this year! I mean really, how many people from the Malta area are qualified to run this new facility? Do they think AMD is going to crank out Twinkies?
“then I see AMD crying their way out of the outsourcing restrictions on the x86 license and selling their stake.”
I swear by all that’s holy, if this ever happened, I’d sell my shares in INTC in NEW YORK SECOND! Besides, I really don’t think the Feds would let this happen. We simply cannot sell away our last remaining exclusive market strong hold in manufacturing for the likes of those charlatans at AMD. Christ, we don’t even make a TV set in this country. We sold out the steel industry. The auto industry is falling on its ass. Now, are we supposed to give away INTC, to say, Samsung!?!?! In a pig’s ass, the Feds gave MS a free pass and INTC will get one too.
I seriously hope INTC keeps the pressure on AMD till 2009 so they can’t commit to the New York deal. (Please dear Lord, hear my prayers)
“So this will be interesting - AMD will now make the argument to Wall St and investors that by cutting AMD in half, it is worth twice as much!”
Hold on there cowboy, easy wrangler! Let me get this straight. AMD paid twice the price for ATI, combined they are worth half the amount, they still owe 5B in the long term ($25 a share strike price, by the way), and now we split them up, and they are worth twice the price again. But this is still half the amount they both were worth before the ATI deal was signed, and they still owe the 5B! “Lunacy,… Lunacy, I say”
GURU, stop, I want to sleep tonight. You’re torture is killing me. I’m bleeding from the eyes.
SPARKS
The conference call was a disaster. They opened by touting their "strong" product lineup consisting of great "innovative" products like the triple core.
Almost every time Hector was asked a question, Meyer answered rather than Hector. Looks to me like a sign that Hector is on the way out. Meyer sure sounded more like the CEO, he was definitely more confident and articulate than Hector.
The best part was when Hector announced "asset smart". Don't worry guys, "asset lite" is still in the super secret plans, Hector just can't reveal the details yet. It's only been one year. Oh yeah, we can't reveal any "asset smart" plans either just yet. Here's a direct quote from the transcript:
"Okay, and I guess the final question on the restructuring would be I believe we’ve just hit the one-year anniversary of hearing about the asset smart or fab light strategy and really haven’t seen much to do with it. I know it’s a difficult and complex strategy. Is there any reason to believe that that scrutiny of your core and non-core businesses would happen any sooner than that sort of one-year anniversary we’ve already run into on the fab light side?"
Just admit it Hector, THERE IS NO ASSET LITE.
One more thing, I hear Vegas is taking bets on which we'll see sooner, AMD's "asset-lite" or Duke Nukem Forever.
"First off - people need to throw out the YoY comparisons as AMD had all sorts of shenanigans between the Q4'06 dumping and the Q1'07 glut... so the comparisons look favorable because AMD front loaded that quarter's revenue into Q4'06 to help with the financing they were doing at the time. So a 22% increase is a bit misleading."
Analysts are too stupid to understand this....
Here's another thing that's wrong with AMD mgmt - there are times to spin and time to swallow your pride (from the earning call):
(CFO)"In our microprocessor business, our innovative triple-core technology is gaining traction for its unique performance and value proposition."
Is here seriously calling tri (I refuse to call it tripple) core innovative? Fusing a non-working call is 'innovative'... ok I can see the value proposition, but innovative? Come on man, do you think folks are still buying this spin?
"All told, it was a quarter characterized by poor top line performance but a strong new product refresh in all of our major businesses that should pay solid dividends as the year progresses"
Solid dividends? That would be CFO - speak for I have not a clue on how to predict/forecast future revenue & earnings.... so we'll call it 'solid dividends'
"Our world-class manufacturing facility in Dresden continues to set benchmarks for operating performance, including yields and cycle types."
Set benchmarks? cycle types? does he mean cycle time? Should the CFO be commenting on this?
"Acquisition related charges are expected to be approximately $50 million" (referring to Q2'08)
Just how long are they going to incur charges for ATI acquisition?!?
Then Ruin-z took over:
"We need to intensely scrutinize all of our businesses in order to ensure that our core X86 and graphics products are on a healthy path to leadership and profitability."
Translation - remember when I told you our plan to be breaking even in terms of operating income in Q2'08? about that... not so much...
"We also need to intensely scrutinize our non-core businesses and revisit their strategic fit into our plans and their path to growth and profitability. Absent these, we will exit those businesses."
Translation - we will be selling anything that isn't nailed down. No more Tahoe junkets for fake launches either?!?
"We have made significant progress in our asset smart strategy and I am personally driving this effort intensely and I am very hopeful that we will be able to communicate details of this rather complex effort in the near future"
I know I'm speaking fast here - you could just re-print what I said a year ago in this area (when the plan was due out shortly?)
"We’ll start the production ramp in the summertime and start to ship products in volume in Q4."
I think this puts Dementia's ramblings to rest - basically AMD has managed to not close the tech gap with Intel at all in terms of schedule and HAS FALLEN BEHIND as this will be a lower performing process.
Full earnings calll transcript:
http://seekingalpha.com/article/72812-amd-q1-2008-earnings-call-transcript?source=yahoo
"We’ll start the production ramp in the summertime and start to ship products in volume in Q4."
Wait.... wait...
From Q3 07 conference call ...
"He said they're on track for production ramp in the first half of 2008. AMD's "first offense" for 45nm products will be server and desktop spaces, though Meyer said he would provide additional guidance in December. Basic yields are being attained right now at 45nm using existing materials. "
http://www.tgdaily.com/content/view/34439/135/
http://www.xbitlabs.com/news/other/display/20071022211442.html
Wholly molly ... these men are chronic liars....
Did I say it or did I say it a years ago when AMD was riding high and Core2 was but in A-step that AMD was finished?
They simply don't have what it takes. They pissed away what little advantage they had with arrogance and silly aquisitions versus running like hell pushing technology and design like they were behind, now they are really finished..
I told Rick, I told Sharikou and I told Scientia..
Oh and best question of all in the transcript:
"you talked about the asset light strategy and having details on that soon. I am trying to reconcile that with your excellent performance in manufacturing..."
So basically you guys are telling us how world class you are from a manufacturing perspective, and then you turn around and tell us that you are looking to offload it?
Hector's response was ridiculous and conveniently dodged the question - hs said because it is so goo it gives them opportunities 'to do things crreative"... but he failed to answer if it is so good why get rid of it?
"I am not really trying to be evasive but I think we are truly tremendous progress in this area and I -- I do not want to be flippant but to me, near future is any time in the next 90 days to the rest of the year." (re = when Hector would provide details on Hecto, um, asset-lite)
3-9 months is near term? Man the analysts have a lot of patience after all the crap Hector talked about this LAST YEAR...
RIVET: "Clearly that’s why we are stepping up to deal with excess employees and those kind of things"
I'm sure the 10% will be happy to know they were 'excess' employees... I can;t believe the CFO just referred to some of his co-workers as 'excess' as if they simply needed to be cut off like some 'excess' fat from a steak.
RIVET:"Again, our consumer exposure makes that a little tough but our goal is to hold share, not lose anymore share than we have and if anything, try to gain some of it back that we lost, particularly in the server space."
So the goals is to hold share and try to get some back - thank you captain obvious... this guy needs to go when Hector goes... If you read between the lines when he says 'consumer exposure' - that means Intel is working us over there and we have no clue what will happen in Q2. This guy wasn't asking about overall seasonality but specifically share vs Intel - AMD is using 'expected seasonality' being down to potentially hide forecasting potentially more losses in market share.
And the obvious nugget (RUIZ):
"First of all, I don’t think that the ownership of the capacity is necessarily a reflection of the leadership in technology needed."
so AMD will be selling a minority share in their manufacturing, the only questions now are how much and to whom.
Man these guys are ruthless (asked about CTO's departure)
"it wouldn’t be appropriate for me to comment on any specific person’s departure but I can tell you we are focusing on elevating our game and fixing some of our execution problems and in so doing, we’re going to bring into the company people with better talent and better experience, and you’ll see more."
So I won't comment...but we are elevating our game... real subtle Dirk... so when you say we are bringing in better talent... that's not a shot at the exiting CTO?
This is really unprofessional and rather poor for someone in Meyer's position.
And this is as close to a "I'm really pissed off" question you will get from an analyst:
"...I believe we’ve just hit the one-year anniversary of hearing about the asset smart or fab light strategy and really haven’t seen much to do with it. I know it’s a difficult and complex strategy. Is there any reason to believe that that scrutiny of your core and non-core businesses would happen any sooner than that sort of one-year anniversary we’ve already run into on the fab light side?"
Translation - this is a polite way of saying - you snowed us on this asset lite crap and now you are shovelling the same shit about cost reductions - is there any reason to believe you this time?
HECTOR "I can assure you that it will happen a lot sooner."
(well, gee that must make the analyst feel better... an assurance from Hector...)
"And are you currently running Fab 36 flat out? Are you 100% utilization?"
RIVET: No (so I guess we would no longer be in the we're selling evrything we can make mode?)
"The Bulldozer core is in development in 45-nanometer technology and we’ll be sampling that in 2009."
translation - 2010 best case for a new core (i.e. 45nm K10 will need to hold up against Nehalem)
“seriously calling tri (I refuse to call it tripple) core innovative?”
Wait, ‘tripple’ is not that bad, especially the way you typed it, and it goes well with cripple. Hence, Tripple Cripple! Good show!
So it’s got fused core. Hey, one mans garbage is another mans treasure!
How’s about AMD looking at the glass as 3 quarters full as apposed being one quarter empty?
“Just how long are they going to incur charges for ATI acquisition?!?”
Seriously, when I read the financials, the same thought occurred to me. They were going to milk (and trash ATI in the process) the acquisition for as much as they possibly can to help the bottom line. Read: smoke and mirrors.
Really, do you know what the worst is? INTC was down on the day (2 cents), while AMD was up 12 cents! It makes you wonder what the hell is going on here.
SPARKS
"And just one final one -- in the December quarter, you said you shipped about 400,000 quad-core processors. Can you give us the number for the March quarter for quad-cores?"
Rivet: "In excess of 0.5 million"
So I understand with record ramp rates and world class manufacturing, they've managed to go from 400K to 500K in a quarter (notice the use of 0.5 million to make it sound better). Also notice the use of "in excess" of 0.5Mil... if it was 600K don't you think he would have stated it?
Either the analysts just don't care anymore and figure they're not going to give any real info anyway or they have become patsies... hard to say which. I think some have thrown in the towel by now and some are just truly idiots - I guess I was hoping one of them would have had the cajones to ask the 'how long do you expect to be CE?O' question.
“we are focusing on elevating our game and fixing some of our execution problems and in so doing, we’re going to bring into the company people with better talent and better experience”
Man! What a scumbag! The seat in Hestors office chair is still warm and they are already burying him!
It official, he IS the fall guy.
SPARKS
"INTC was down on the day (2 cents), while AMD was up 12 cents! It makes you wonder what the hell is going on here."
AMD was up after hours too... fairly predictable - this is why folks pre-announce; you generally take a hit on the pre-announcement, but not probably as much as you should (one would think a 10% shortfall might trim ~10% off the price? or at least 5%?). Then when you get to the call they say... ah they met the lowered expectation so let's bump up the price a little.
What you'll see now is the price start to drift down as traders take a step back and say huh? Did they really just give no actual details on this grand asset smart plan, announce simple seasonality which implies they don't expect to gain market share back despite touting Q1 as a quarter where they did major product refreshes which are taking hold? With 'strong' new products on both the CPU and GPU side, shouldn't they be gaining share?
I would think a year after the acquisition they would not still be incurring acquisition charges (and they forecasted another for next quarter!)- this would have to be the slowest assimilation I've seen in some time! I wonder if they will soon be able to write off Hector's bonus as a one time charge!?! (I say VERY soon)
So before AMD announced the sales miss after hours on the 4/7 the stock was at $6.34 now the stock after the dip and subsequent recovery is at 6.23 after hours.
So a 10% revenue miss lead to an overall $0.11 drop in price (or a 1.7% decline), I wonder how long it will take the market to correct this one? They couldn't have liked the job cuts that much, could they? It's not like AMD's outlook was gangbusters either...
Since the ATI acquisition, the GPU business group has returned exactly -138 million. Put another way, AMD invested 5.4 billion and has to date (on just the GPU part) has lost -138 million.
Let's repeat that ... AMD invested 5.4 billion dollars for a -138 million return (for GPU only), after Q406 AMD lumps chipsets in with computing solutions, therefore the actual return could be lower (or higher).
http://www.amd.com/us-en/Corporate/InvestorRelations/0,,51_306_643,00.html
-27 million Q406
-35 million Q107
-50 million Q207
-3 million Q307
-12 million Q407
-11 million Q108
Actually, this is not atypical of ATI before acuqistion, they bounced between black and red for most of the last 5 or 6 years (cumulatively in the black).
This is just staggering ... 5.4 billion investment and, to date, 138 million lost.
Does anyone happen to know the market share change for Q1 on GPUs? Many have said that new products and lowered prices made AMD fare much better than it did before and I'd like to see how they actually did.
A few observations:
* I can understand why Ruiz continues to move the "Asset Smart" plan into the future. They are missing a couple of important components in that plan as of yet. Mainly, they are missing "Asset" and "Smart". They seem to have plenty of "Lite", however.
* When Rivet mentions "excess employees", you kind of wish he was referring to Ruiz.
* And there is this observation by Roborat: "Other alarming trends in their numbers include margins going south from 44% last quarter to 42% in Q1'08."
I recall from one of Scientia's blog posts that AMD's hope was to maintain 44% through the first half and increase margins to 50% in the second half of 2008. Having them drop instead of hold steady is a bad start on that plan, and nothing they said should make anyone think that they'll be averaging anywhere near 50% in the second half. Does anyone think that they can maintain 42% in the second half of 2008?
* The whole analysts call reminds me of those short films that the NFL makes for each team, as a "year in review" souvenir for fans. Even the worst team makes one, and for the bad teams the film usually has to try to put some polish on a turd, even though they know that you know that the team is just awful. So you get stuff like "even though the Saints lost, they played hard." They don't mention that they lost 55-3 and played horribly.
That is what this conference call was-- a highlight film for a team that went 2-14 on the season and has very few options available to them. If they finish the following season 3-13, they will celebrate their "year to year improvement." But that doesn't sell tickets, does it?
"Put another way, AMD invested 5.4 billion and has to date (on just the GPU part) has lost -138 million. "
Actually it's worse than that when you consider the interest expense AMD incurs on the debt related to the acquisition.
AMD claimed they were not going to get involved in a # of core race (ala the MHz race) right? Well....
http://www.dailytech.com/Dodecacore+The+Megahertz+Race+is+Now+Officially+the+Multicore+Race/article11531.htm
"Dodeca-core: The Megahertz Race is Now Officially the Multi-core Race"
This is a 'glued' package of 2 six cores.
So I understand:
1) we are not going to get involved in a core race
2) Native is better
were just convenient statements at the time? I'm waiting for the who's copying who (on both 6 core and MCM) comments from the AMD fanboy peanut gallery... I don't suppose I'll see them though.
OK 12 cores is just ABSURD... I even wonder about Intel's 6 core product (though perhaps for server this could be utilized)
"So a 10% revenue miss lead to an overall $0.11 drop in price (or a 1.7% decline), I wonder how long it will take the market to correct this one?"
AMD down $0.18 today(just under 3%) after the ridiculous ride up yesterday and the after hours misinterpertation of the results.
The market corrected this one rather fast (and in a day where tech stocks generally did well).
"* And there is this observation by Roborat: "Other alarming trends in their numbers include margins going south from 44% last quarter to 42% in Q1'08."
I recall from one of Scientia's blog posts that AMD's hope was to maintain 44% through the first half and increase margins to 50% in the second half of 2008. Having them drop instead of hold steady is a bad start on that plan, and nothing they said should make anyone think that they'll be averaging anywhere near 50% in the second half. Does anyone think that they can maintain 42% in the second half of 2008?"
We'll have to wait for them to file their 10Q with the SEC to know for sure, but I wouldn't be surprised if much of this drop is that they are running out of money from the German Saxony subsidy to apply towards their cost of operations that they have used as a crutch for margins the last few years. (Ya, they use gov't subsidy money as a credit to cost of operations...)
"So before AMD announced the sales miss after hours on the 4/7 the stock was at $6.34 now the stock after the dip and subsequent recovery is at 6.23 after hours.
So a 10% revenue miss lead to an overall $0.11 drop in price (or a 1.7% decline), I wonder how long it will take the market to correct this one? They couldn't have liked the job cuts that much, could they? It's not like AMD's outlook was gangbusters either..."
Eh, I'm not surprised, I think most casual investors have already bailed on AMD. After a 60%+ drop in the last year, most, if not all of the bad news has already been priced in. I don't think AMD's stock price will drop much more ($5 floor) unless BK occurs.
"Since the ATI acquisition, the GPU business group has returned exactly -138 million. Put another way, AMD invested 5.4 billion and has to date (on just the GPU part) has lost -138 million."
Phew, now I don't feel as bad about the equity I've lost in my house the last couple years.
I think this puts Dementia's ramblings to rest - basically AMD has managed to not close the tech gap with Intel at all in terms of schedule and HAS FALLEN BEHIND as this will be a lower performing process...
he doesn't think so.
when asked for his opinion:
[D. Meyer: We’ll start the production ramp in the summertime and start to ship products in volume in Q4.]
Enumae:
Whats you take on this?
Scientia from AMDZone said...
enumae,
I have no idea from that description. It sounds similar to what was said before but nothing specific. By that statement he could mean release in Q3 with volume in Q4 or release and volume in Q4. It could also mean either early Q4 or late Q4. It's still too vague.
Vague? this clearly means 2 terrible things are about to happen:
1) AMD 45nm is gets to the consumers more than a year behind Intel which he once insisted that the gap is just 6 months
2) Now we're positive we'll see Nehalem before Shanghai.
You can't be any more clear than that! "It's still too vague"... yeah right.
Depressing state of AMD's Phenom
“What you'll see now is the price start to drift down as traders take a step back and say huh?”
Kudos, brother, spot on.
The market took off today, spurred by INTC outstanding 1Q (typically seasonal low) financials, earlier in the week. Let’s hope this isn’t a dead cat bounce in the long term and for the market overall. Maybe, we’re out of this banking/lending mess. That said one company DID NOT bounce back, as did INTC/APPLE, etc.
Anonymous, called it.
http://moneycentral.msn.com/msn/stock_quote?Symbol=amd
Flush twice. AMD cannot float this much longer.
Sparks
In a previous thread Chukula said...
I hope it IS possible to put 6 DIMMS on an ATX board, or else Intel is going to have to abandon the ATX form factor for Nehalem (3 memory channels and 2 DIMMS/channel).
The top of the line Nehalem part for desktop will be bloomfield according to X-bit.
Intel wants to go even further and scale the number of memory controller channels. According to PC Watch web-site, the top Nehalem processor code-named Bloomfield with four cores will have triple-channel DDR3 memory controller, whereas slightly less advanced may have less channels.
And Xtreme Systems has a picture of a bloomfield prototype motherboard here.
Note there are not 6 memory slots on the board.
And as to my comment about traces, just look at all the area between the CPU socket and the memory (the dark green area). Those traces take up between 1/2 and 3/4 of the area of the memory slots. You would need to double that area to add another 3 memory slots.
So my original point stands, you have to do a lot more than just look at a board and say "can I cram another component in this space".
Hey I have a question for all.
Hmmm, ---- nice, so what do we have?
-27 million Q406
-35 million Q107
-50 million Q207
-3 million Q307
-12 million Q407
-11 million Q108
“they bounced between black and red for most of the last 5 or 6 years (cumulatively in the black).”
So, let’s say you’re ‘Big Paulie’.
Ya got the other Boss on the run.
Ya got your soldiers on the street (Q6600) plus their underling assassins (E8x000) putting pressure on the entire neighborhood, they work cheap, never complain, and are fast enough to take care of things. You’re making money.
The Family’s happy.
Now, your Lieutenants and Captains are really bringing home the bacon. Although they cost more to take care of business, in pinch, they could hit the street and really clean things up. But, it’s gonna cost ya, if you cut their pay. (45nM)
The Family’s wants a high cut.
Do you tell the family “look if we take a hit on the money now, I’ve got a captain, waiting to take over and put a big hit on Don Wrectum” (Nehalem)
The Family says, “Look, Big Paulie, you do watch ya gotta do, aright? When ya gotta bring ‘im out, bring ‘im out, but make sure dis guy does rite ting, aright?”
So what do you do? Do you bring Nehalem out early or and really put the squeeze on? Is he going work well with your other guys on the street? How long can the older guys on the street take care of business? Can they keep making money for the Family (higher margins)? Will the new guy keep the pressure on? Will it bring more pressure on Don Wrectum?
Or do you hold him back, refine his skills and pull him out, late, only when you have to?
Yo, I’m part of the Family. I want my cut, too.
SPARKS
InTheKnow said ...
And Xtreme Systems has a picture of a bloomfield prototype motherboard here.
Note there are not 6 memory slots on the board.
And as to my comment about traces, just look at all the area between the CPU socket and the memory (the dark green area). Those traces take up between 1/2 and 3/4 of the area of the memory slots. You would need to double that area to add another 3 memory slots.
So my original point stands, you have to do a lot more than just look at a board and say "can I cram another component in this space".
while you analysis is correct, but there is nothing technically prohibitive to have such a board with 6 memory slots. I just took a look at the Bloomfield board that you linked, what I would say is that there is a still a possibility of a few variance of Bloomfield boards.
Pointer, the prohibition is space. If you do not increase the size of the board there is no room for the additional memory and associated tracing. So you must either increase the layer count to allow for the additional tracing, or increase the board size.
My contention wasn't that bloomfield boards couldn't support the memory. My point was that there is more to adding components to a board than looking to see if there is a spot where you can squeeze the physical component. You have to allow for the tracing as well, and in the case of memory to feed the Nehalem beast, that is a lot of tracing.
Chuckula responded saying that if you couldn't fit 6 memory slots on a board the ATX form factor was dead for Nehalem. This board would seem to indicate 6 slots aren't needed.
You guys seem to be splitting hairs. What both of you, rather, what’s not clear here is whether or not the future triple channel will be interleaved the way DDR2 and DDR3 are.
If it is dual channel interleaved then there would need to be six slots. Perhaps there is some new configuration.
In The Know, perhaps the triple channel is triple leaved or some other newer technology we currently aren’t aware of, perhaps? I’m guessing, but if that prototype is configured with 3 Blue slots it would seem to be the case, maybe? Got any links?
SPARKS
I'm somewhat confused at this talk of not being able to place six memory slots on a standard ATX motherboard.
Take a look at this MSI board:
http://global.msi.com.tw/index.php?func=proddesc&prod_no=1250&maincat_no=1&cat2_no=170#
It has six slots, four for DDR2 and two for DDR3. Now obviously all six slots can't be in use at once as the memory controller only supports running a single type of memory at once.
But I see that as a limitation of the chipset itself, and not of the motherboard. This particular board doesn't cost a fortune either.
I purchased the ASUS P5E3 Premium and SuperTalent 1800 DDR3 a few weeks back. Obviously, I take my time overclocking, a tweak here and there, play a few games, and tweak, some more. If can’t swing an entire game of Crysis, I’ll back it down.
Mind you, these numbers represent real life, 100% stability, relatively low temps, 24/7 on air-----so far.
Everything is out of a retail box, I have no factory ringers, I’m not that important, or hooked up. I don’t turn my babies off, booting is too dramatic.
If there’s any doubt about INTC’s Q6600’s ability to BLAZE let’s get this horseshit out of the way once and for all! It’s a goddamned rocket.
And, those who said X48 wouldn’t mean a hit of crap, think again, RATED FOR 1600 MHz FSB, enough said.
I kicked down the multiplier from stock 9 to 8 and cranked the FSB to 421.
CPU Z is reporting 421.1 X 8 = 3368.7 MHz
FSB @ 1684.2
Vcore=1.3875
The bread and butter $240 Q6600 is clocking A FULL GIG over spec!
That Pheromone DOG can’t clock a couple of hundred Megs!
Orthogonal, don’t even try to tell me INTC wasn’t sandbagging these puppies. I’m sure INTC ENGINEER GENIUSES were saying HOLY SHIT!!!
Temps, you ask? Try 45 degrees, BFD. Zalman 9700
SUPER PI = 1M in 14 sec.
Sandra 2007 Processor Multi - SSE4= 370323 SSE2=200717
Mem. Band = 8452, 8278
Cache + Mem = 62474 MB/sec Combined Index!
Mem Latency = 65ns
Processor ALU 62,226 MIPS and SSE3 42866 mflops
Again this is NOT the limit! I have gone there----yet!
Real World:
The memory is COOL to the touch
The 6600 is slightly warm.
ALL the regulator heat sinks are cool!
The X48 is slightly lukewarm.
The Raptors got a boost too! 113MB sustained, 114 MB burst, 121 MB peak.
The whole magilla is tuned at the FSB, they can SHOVE that IMC.
In The Know, what ever these guys did to these “traces”, they got ‘em playing like a violin! Boy, oh boy, did INTC get off it’s ass during the past two years!
Next stop, the big fella QX9770!
HOO YA!
Clock till ya rock!
SPARKS
Giant, your link took me to an ethernet switch. Not quite what you intended to show I'm sure. :)
Oh yeah, a couple of things I left out.
Memory Timings 7-7-7-21 2T @ 2.0V
Everything opens quick, real quick.
It just feels crisper, no mater what you do, or how you do it.
No hafnium. Heh, heh, heh, heh, heh.
SPARKS
Here it is:
http://global.msi.com.tw/index.php?
func=proddesc&prod_no=1250
&maincat_no=1
SPARKS
Ouch.
I guess you can only thumb your nose at analysts and investors for so long before they begin to resent it. Even with all of the leeway that AMD has gotten, it seems that patience is running very thin. I guess when you get the benefit of the doubt well past the point where you deserved it, getting yanked back into reality can be especially painful.
2008 is looking to be especially painful...
They might have a long wait. After dodging persistent questions on its asset-smart/fab-lite manufacturing strategy, Ruiz could only promise last week that the company would provide details in the "near future."
Those details might not be known for some time based on Ruiz's definition of near future. "I am not really trying to be evasive, but I think we are truly making tremendous progress in this area. I do not want to be flippant, but, to me, near future is any time in the next 90 days to the rest of the year," Ruiz said.
It may not be clear to Mr. Ruiz, but he has been both evasive with his answers to questions on the asset smart manufacturing strategy over the last four quarters and repeatedly flippant with investor concerns about the company's future.
While AMD's management is taking its time working out details of the asset-smart manufacturing strategy, the company's stock is getting clobbered and its costs remain uncompetitive compared to Intel's.
...
Here's an exchange between Ruiz and Ross Seymore of Deutsche Bank on the issue of AMD's newest plan to "scrutinize our non-core businesses and revisit their strategic fit."
Seymore: "I believe we've just hit the one-year anniversary of hearing about the asset-smart or fab-lite strategy and really haven't seen much to do with it. I know it's a difficult and complex strategy. Is there any reason to believe that that scrutiny of your core and non-core businesses would happen any sooner than that sort of one-year anniversary we've already run into on the fab-lite side?"
Just when you think AMD couldn’t be hurting any more, guess again. INTC is slashing prices.
This has gotta hurt. INTC is really turning the screws. They’re going to clear inventory, move very quickly into 45nM (higher margins), and cut AMD’s throat, all in one fell swoop.
The lower margins on 65nM, now, will be made up by larger margin 45’s at the end of the quarter. (Very Smart)
http://www.extremetech.com/article2/0,1697,2285978,00.asp
SPARKS
LOL tick tock tick tock is that not what one company keeps talking?
65nm fire sale
45nm how many factories ? 3 or was it 4
how many 65nm did they have 3 or was that 4?
They shrunk chips, double the cache and have the same number of factories? NO, they have 45nm and they got 65nm.
Damm that is a lot of factories!
Didn't AMD have problems competing with the 65nm and now 65nm is garage sale price and 45nm is ramped.
AMD BK in 2009 I say
Looks like Ah-ben-Stoopid over at AMDzone has finally wised up to Dementia's tricks:
"I really don't know when to believe you when you talk about things that I'm not sure of, because you spoke of things that I actually know in a wrong but very surely way. We've seen enough of such misinformation from all the Intel fan-sites over the internet, and here you're performing the same anti-technical acts in favor of AMD. This is wrong."
That, coupled with hyccup's telling Sci that he doesn't know much about coding, seems to be terminally tarnishing the tin god of tech :)
FYI -- Mercury Research report was published yesterday, MSS swing to Intel was massive. AMD lost share to Intel in all segments, the most was in server followed by mobile, then DT.
I suspect some news headlines will start up in the next day or two.
Jack
JumpingJack,
Nice to have you back.
Do you have a link on that Mercury Research post?
SPARKS
Sparks ...
No, I have access to the report and read the summary information... the total swing in marketshare was on the order of 2.0%. Intel now holds over 79% and AMD just over 20%.
Both Intel and Via increased their %share, AMD dropped. All companies showed a drop in processor shipments from last quarter... both Via and Intel showed percentage drops that were seasonal, AMD of course performed much worse than seasonal.
Ironic since it was Intel that guided to below seasonal, and AMD that guided at seaonal in Q4 conference calls.
I suspect after the report has a chance to digest some, a few headlines will make their way to the surface.
jack
Jack
2 percent, down to 20! So much for Wrectors 2007 slash and burn. GURU said it wouldn't fly. Now what can they do?
INTC's recent price cuts will be the Coup De Gras. I see INTC gaining more market share this next quarter.
Tally Ho, back to the more traditional 80's.
Leap Ahead
Thanks
SPARKS
"NTC's recent price cuts will be the Coup De Gras. I see INTC gaining more market share this next quarter."
I don't think the price cuts will matter much in terms of market share... just AMD's profitability... at this point AMD has no point but to price to whatever it takes to sell. The only thing worse than selling low margin parts is having an idle factory while really expensive equipment depreciates rapidly.
Much like AMD's theory of they need over 30% to 'break the monopoly', if they fall below a certain MS (I would guess about 12-15%) things would snowball - OEM's might start bailing, no chance of any idiot putting up money, and much more serious free cash shortfalls (leading to more layoff leading to more execution issues, etc). At this point, because of Hector's lunacy (or ego?), they have no choice but to price to whatever it takes to maintain ~20%... it is their only shot at getting cash from some luny investor. And no, Dementia's ridiculous arguments of AMD's been in worse condition won't fly because their debt load was not so severe and they were running a 200mm fab (the economics between 200 and 300mm are significantly different). Also the complexity of bring a new tech is a lot more difficult and costly these days.
What really could be the death blow (and I think Robo first suggested this?) is if atom takes hold. There's simply no way AMD could compete price-wise with these with K10 (or K8) and some of the 'high end' atoms are approaching AMD's new 'good enough' campaign from below, leaving AMD dual cores nowhere to hide. At this point either AMD needs to have a ridiculous jump in clocks on K10 or hope for Intel to fall down on multiple items (some combination of 45nm ramp, Nehalem, atom uptake, chipset issues) or hope for some more EU socialism to take hold (or start buttering up the US gov't).
On a side note is anyone else amused at the Microsoft - Yahoo takeover attempts?
First this is the same MS, that has faced all sorts of monopoly issues with the OS, and apparently people seem to have no issue with them acquiring Yahoo.
Second, Yang seems only second to Ruiz in terms of idiocy:
A) He said Microsoft was not offering fair value - AFTER THEY OFFERED 62% ABOVE MARKET PRICE. Hey, Jerry, how else you getting that stock up 62% in our lifetimes.
B) He is now being laughed at by most analysts after jacking up his forecasted revenue in Q2 and the rest of 2008 - he transparently did this in an effort to force MS to jack up their offer price. The market saw right through this - a 25% or more increase in the forecast led to NO CHANGE in the stock price! He could and should go to jail for this - though it will be impossible to prove. It is clear he did this as he knows he will never be around to report those Q2 #'s (he will either be removed in a hostle proxy fight or give in and accept whatever MS's final offer is). He might as well have forecasted 5 Trillion in revenue! (or did something like a 10% bump which wouldn't have seemed like such an obvious lie)
I tell you what - if I were Ballmer right now... I'd threaten to walk away and say 'wow, if he can get such great performance, perhaps we shouldn't take the company oover and then when the stock crashes to the pre-takeover level and crashes again when Yang missed his ridiculous Q2 numbers, Ballmer should step back in and lowball an offer and say "how do you like me now?!?"
My point for this digression - Don't CEO's like Ruiz and Yang have a fudiciary responsibility to their stakeholders? It seems like they are run by their egos and stockholders get left holding the bag. If I were a Yahoo shareholder right now I would seriously sue and try to get a class action suit started for effectively STEALING money from stockholders. Ruiz is not quite in the same boat as in his case it is more complete incompetence.
And what are the board of directors doing at either of these companies?
“Don't CEO's like Ruiz and Yang have a fiduciary responsibility to their stakeholders?”
I think Wrector is in a league all his own. In contrast, Yang has something to sell, like a solvent company, perhaps? Anyway, he hasn’t made fatal errors in judgment,------so far.
Wrectors moves were too expensive, myopic, ill contrived, and ill timed. As far as the board is concerned, going along with him, I think we’ve beat this to death before. They at least had to be asleep at the wheel.
I don’t know, I’m a working class slob, I’ll never understand the political dynamics of corporate ‘yes’ men and ass kissers. Besides, I’ve got a union to protect me from the good old boy club, especially when one these jackasses are way out of line and in my way on one of my jobs. Once, I had a VP complain to the President of our company about my “irascible nature”, the boss told him “that pain in the ass turns a 40% profit on his jobs, do you still want me fire him?”
I read somewhere that AMD insiders felt Phil Hester was more show than go. Does that mean they kept him on because he was a 23 year veteran peacock at IBM? Was he directly responsible for the “bad process and bad architecture” concerning Barcelona’s relative failure? Wrector and Dork couldn’t wait to piss on him, as if they were absolving themselves of any responsibility for the company’s position today. Therefore, does that mean Hester running the company while Dork, Wrecktor, and the BOD were out bicycle riding?
Horseshit, when the head of the fish stinks, the whole thing stinks. When I compare ‘Big Paulie’ to Wrecktor, I know who I would want running the ship. I know who I’d like to work for.
“I think you'd be misreading the market if you think people care about the packaging," - Paul Otellini
In other words, you don’t know what the hell you’re talking about. Read: Q6600 “the Killer”
Think of it, Atom and QX9770, all on 45. That’s quite a spread, brilliant, simply brilliant.
Big Paulie is a class act, Wrecktor and his minions are a bunch of jackasses.
SPARKS
"Both Intel and Via increased their %share, AMD dropped."
Did Via see these gains as a result of the Isiah CPU? There were some news stories when they announced it and then silence, for the most part.
Another head bounces out the door.
http://www.fudzilla.com/index.php?option=com_content&task=view&id=6995&Itemid=1
SPARKS
Anonymous said...
"Dodeca-core: The Megahertz Race is Now Officially the Multi-core Race"
"This is a 'glued' package of 2 six cores. "
Just think - if they fuse off 2 of the cores (like with the tricore Phenom) then we would have a DODO-core :)
Think I'll go register that as a trademark before AMD marketing thinks of it...
Tonus -- I did not see mention of Isaiah with respect to revenue material for VIA, but also did not read the entire report (just summary section 3.1 :) )
I am sure there was some movement due to Isaiah. However, the MSS gains for Via were quite small, they 0.1 or 0.2 % or so, they are still under 1% for the market at 0.9%. However, at the size of the market, these tenth percentage movements are material revenue especially to a company the size of Via.
Jack
“(the economics between 200 and 300mm are significantly different)”
and----------
“having an idle factory while really expensive equipment depreciates rapidly.”
I wondered about these transitions for quite some time. INTC and AMD construct these huge and complex buildings. They have all the tools in whatever sequential order it takes to get the chips sliced, diced and out the door. Why do they have to build an entirely new structure whenever they move to a different wafer size or process type/size? In other words can’t they just “gut the floor” and use the buildings existing infrastructure and replace older tools with a newer lineup. Why is it necessary to start from scratch or from the foundation to move foward?
Further, AMD has apparently mothballed a FAB. The process and its tools obviously depreciate over time. What do they do with the multimillion dollar tools and/or the buildings, especially the equipment that proved, shall we say, less than successful? It’s not like you can sell this stuff on ebay?!
What does INTC do with a dated FAB and the (obsolete?) tools that are in it? I get the picture these things are peppered all over the globe like some extravagant testament to past technological glory. (In AMD's case failed vision.)
Sparks
"Sparks said...
I wondered about these transitions for quite some time. INTC and AMD construct these huge and complex buildings. They have all the tools in whatever sequential order it takes to get the chips sliced, diced and out the door. Why do they have to build an entirely new structure whenever they move to a different wafer size or process type/size? In other words can’t they just “gut the floor” and use the buildings existing infrastructure and replace older tools with a newer lineup. Why is it necessary to start from scratch or from the foundation to move foward?"
That's a good observation, but it isn't as simple as that. It takes considerable foresight/planning to predict how a 200mm would upgrade to a 300mm fab. From process to process, there are only so many tools that can move along with it. Obviously, with a wafer size increase you have to gut everything. It takes a lot of planning on the part of the industrial engineers to make sure adequate sub-facilitlies are in place (Gas/Chems etc...). There may be new specs or standards for material delivery systems. Changes in quantities and depending on how many tools you need to supply there may not be enough floor space in designated to accomodate existing facilities. Process flow may change enough that other systems are completely useless and need to be removed. Whether or not the new ones fit into those areas is also a concern.
Further, There may be significant structural changes to accomodate new automation systems, stockers for large 300mm FOUP's (Wafer boats) etc... This list goes on and on.
"Why is it necessary to start from scratch or from the foundation to move foward?"
Not necessarily starting from scratch, but more or less athe fab has to be gutted and there are many reasons for this... some of the big ones:
1) Wafer handling - on 200mm this was largely done by hand (either carrying a 'lot' of wafers or pushing or cart around with the lots). With 300mm the weight of a lot became an ergonomic issue, and an automated material handling system (essentially an overhead rail) was implemented ("AMHS"). This is fairly significant hardware automation system.
2) The facility requirements were different btw 200mm and 300mm for the equipment. FLA's (something Sparks may know a bit about) were different, requiring different electricals, water and waste requirements and subsequent plumbing, weight of the tools (some older fabs would need structural reinforcement on the floors) due to the increased weight, exhaust requirements, and simple height of the tools in some case were all potential issues.
Granted this stuff could TECHNICALLY be done while still ramping down production on 200mm but it is a logistical nightmare and likely incurs more expense.
On the economic side, the big change is the relative ratio of the capital (tool cost) to overall wafer production cost change - capital is now a more significant % of the overall wafer production costs. This means more cash is required up front.
Finally consider the impact of this in the development phase, where you generally like to have at least 2 tools per tool type (one for handling normal development and one for more risky or non-standard upgrades you are testing). This means there is also now a much larger up front development cost.
So now after you have incurred more up front expenses and you are running at less than 100% utilization of the fab, you have a much larger % of your production cost depreciating while idling (not a good thing). This makes planning a lot more dicey as there is a lot more risk if you overestimate capacity needed. I seem to recall AMD talking about the ability to add capacity in terms of 9-12months based on equipment leadtimes (and subsequent installation and qualification) in their last conference call. This means a much more conservative approach - if demand is there, order more tooling, but means much slower responsiveness to the demand and potential lost opportunities. I'm sure they do some risk ordering - but given their current financial situation it has to be a lot more conservative(look no further than the change to the F38 conversion and ramp)
"Further, AMD has apparently mothballed a FAB. The process and its tools obviously depreciate over time."
I don't think this fab (Fab30/38) was ever really loaded with that many 300mm tools - I suspect some have been installed, but certainly the full capacity has not been 'built out' yet. The 200mm tools are being sold off and AMD is probably getting ~30 cents on the dollar for it (this is typical of the equipment reuse market). I suspect AMD has done a lot of the structural 300mm work (facilities, basic AMHS work, etc) but probably has pushed out much of the equipment purchases as this is what will depreciate very rapidly.
The tools are typically depreciated over 4 years... so if you are talking about a $4mil tool (fairly typical), this is $250K of depreciation for a typical tool every quarter. If you want to get an idea of how quickly this adds up take a look at AMD's quarterly depreciation expense... for Q1'08 this was $267Mil (and probably most of this was just the single F36 factory).
So let's say you accidently thought you needed 1/4 of a fab's capacity 6 months prior to when you actually did need it... you are talking about potentially a ~$267Mil * 1/4 of a fab * 2qtrs or a ~$140mil error. In reality tools and tool orders can be pushed out, cancelled and creatively accounted for (in a legal fashion) but still you incur all sorts of collateral costs.
Now Intel has similar risks, but they have an additional degree of flexibility as tools can be re-directed from one fab to another, if say 1 fab's ramp changes. AMD does not have this luxury. Also when you buy a more significant amount of capital, equipment suppliers tend to be a little nicer about cancellation fees and changes in tool forecasts.
"The tools are typically depreciated over 4 years... so if you are talking about a $4mil tool (fairly typical), this is $250K of depreciation for a typical tool every quarter."
4 years is typical, the concept is that you want to appropriately account for the 'upfront' captital costs fairly, but also depreciate it as quickly as possible while turning in good margins.
As mentioned, 4 years is typical ... pretty much standard, however, in 2006 AMD announced that they stretched their depreciation schedule to 5 years then later to 6 ... the effect is that it makes their gross margins look slightly better earlier on... but they retain the effective capital costs for longer.
Add: Link describing AMD's depreciation schedule:
http://www.fabtech.org/content/view/1419/
"in 2006 AMD announced that they stretched their depreciation schedule to 5 years then later to 6"
Good link - thanks for the info. This is a bit of a head scratcher though as if the tool is no longer needed after say 3 or 4 years you then have a nice depreciation expense to eat at the end. (Though granted in AMD's situation they obviously are eager to defer any near term expenses and probably aren't too worried about 4-5 years into the future).
Given that some tools don't see more than 2 technology cycles and each cycle is ~2 years, AMD is potentially facing some "ballon payments" in terms of having to eat the final couple of years of depreciation. Think of the 90nm equipment that AMD purchased in 2005/2006 for F36 while a good portion of that can be reused for 65nm and probably some for 45nm, the rest of that equipment will now have 2-3 years of depreciation left as that fab gets converted to 45nm and all of the remaining depreciation will suddenly come due. Either that or they are planning slower technology migration cycles and fab conversions in the future.
Good for the short term #'s but if they do survive they will potentially have to pay the piper later on. It does not seem like a sound business decision - even in 2006 when it was done and the cash and market situation was better. This must have been when Hector was still in the, 'aww heck, by the time these expenses come back to bite me I'll have >30% market share and I'll be printing money!" phase.
I guess we could be glad Ruiz doesn't run a financial institution as his mortgage the future mentality would have crushed him in that arena.
Whew! So, let me get this straight. You guys look at the buildings themselves as merely a package, a shell, albeit a big one, whose life span is determined by the proprietary equipment, built into the structure? This gives a new meaning to planned obsolescence.
It’s not the jar; it’s the truffles inside, and their losing their flavor, real fast. Therefore, it doesn’t pay to air express the jar back to France for a refill, you just make ‘em a new jar.
It seems the only standard I see here is Orthogonal’s FOUP’s , (Front Opening Universal Pods) Why do I get the feeling he used to hump the 200mm variety around the building until the 300mm jobs were put on a track?
It’s really too bad they can’t make the tools modular in some way where the size of a unit could be integrated in a standardized container/track system. Over sizing electrical conduits with flexible interconnects, standardizing pluming and network infra structure would be relatively cheap. Slide the old line out, slide the new line in, throw in a couple of spacer units here or there, reprogram the FOUP tracks, done! (I can hear you guys now, “what an idiot”)
Ah, just pipe dreams. But, here’s the thing, and there’s no getting around this, where and how the hell is AMD going to “ramp” 45nM by the end of this year?!?!? They don’t have a pot piss in or a window to throw it out of, let alone an elaborate (and expensive) 45nM FAB mentioned in the previous posts.
Speaking of which:
http://news.moneycentral.msn.com/ticker
/article.aspx?Feed=ACBJ&Date=20080424&ID
=8535425&Symbol=AMD
Another 8.5 million, there ya go, GURU, death by 10,000 paper cuts.
SPARKS
Some rumors on Puma (courtesy of THG):
http://www.tomshardware.com/news/amd-puma-intel,5225.html
'Regardless Intel’s 45 nm design has the edge and and consumes less power. However, the current situation apparently prompted AMD to adjust its marketing message, stating that it is focused on delivering “balanced” platforms. An AMD source told us that "Griffin will not be able to touch Core 2 Duo", which is quite a bombastic statement coming from a senior source.'
Prepare for 'platformance' metrics to be made up accordingly. Either that or some marketing on how notebooks are all about gaming and graphics!
This is long and a bit disjointed, so apologies in advance.
Sparks said...
Whew! So, let me get this straight. You guys look at the buildings themselves as merely a package, a shell, albeit a big one, whose life span is determined by the proprietary equipment, built into the structure? This gives a new meaning to planned obsolescence.
Not quite. The equipment is interchangeable. The issue is whether or not the equipment will fit in the fab. If you were to set an 8" tool next to a 12" tool, you would think the 8" tool was a tinkertoy.
By way of background, a Fab has 3 main levels. The subfab (actually built at ground level), the main fab and the fan deck. There is also an interstitial space between the fan deck and the main fab, but this is really only half as high as the subfab or main fab, so no one ever seems to treat this like a level.
The sub fab is not a clean space. It is where all the air handling equipment, pumps, water chillers, etc. lives. If you look at vacuum systems for example, it is pretty obvious that if you are going to double the size of the area you are pulling a vacuum on, you need a bigger pump. You also need a higher capacity chilling system for cooling water. The air handling system is a function of floor space, so all other things being equal, you don't need to change that to move from 8" to 12". So your subfab space is more heavily utilized in a 12" fab than an 8" fab.
The main fab is where all the automation systems and process tools are. One big consequence of moving to 12" from 8" was the need to incorporate automatic handling systems into the fabs due to the weight of the wafers.
Anyone loading trucks at UPS would laugh at this idea, but these aren't Grandma's goodie boxes being tossed around. Gentle handling is a plus and the automated handling system (AMHS) provides that.
Since a big part of the fixed cost in operating a fab is in the air handling system and the air handling system is function of floor space, the logical thing is to put the AMHS above or below the tools. Above is easier, so that is where it went.
Combine the extra height of the 12" tools with the room needed for the AMHS and you get to the real crux of the problem. You need to raise the ceiling. Now to convert your old 8" fab, you have to raise the ceiling. And not just the ceiling, but the entire roof since you still need the interstitial space for all your exhaust plumbing, etc.
So to reuse the old 8" building you have to raise the both the floor and the ceiling of the 3rd floor on a 3 story building. It can be (and has been) done, but it is seldom considered to be cost effective.
anonymous said...
Prepare for 'platformance' metrics to be made up accordingly. Either that or some marketing on how notebooks are all about gaming and graphics!
This is the option I'm putting my money on. I think I remember seeing the first inklings of this in an article I read on the web somewhere.
It makes perfect sense to me though. I'm sure that my company bought me that laptop so I could play Crysis during my lunch hour. (sarcasm intended)
“This is long and a bit disjointed, so apologies in advance.”
Ahhh----NEVER! I wish I could get into one of these places, let alone work in the sub FAB!
“The sub fab is not a clean space. It is where all the air handling equipment, pumps, water chillers, etc. lives.”
These are the places I live!
Noisy, dirty, BIG, power (read: BIG POWER!), sounds like heaven to me. Keeping all the jewels upstairs in tip top shape, cranking along with ultra clean UPS? Transfer switches, line filters, reactors, line conditioners, motors, starters, VFD’S, triple redundant uninterruptible power, networked BMS, lighting, and power distribution, oh the joy!
Bro, it’s better than overclocking.
I appreciate your generous time.
By the way, playing Crysis on a laptop? Someone’s pulling someone’s chain. It ain’t gonna happen, not at high resolutions and frame rates, (35+). Unless, of course, you want to loose blood flow to your legs from the weight, and sterilize yourself from the heat. Oh, yeah, you’ll throw your back out from the weight of the extra batteries in your backpack.:)
Thanks much,
SPARKS
InTheKnow said...
So to reuse the old 8" building you have to raise the both the floor and the ceiling of the 3rd floor on a 3 story building. It can be (and has been) done, but it is seldom considered to be cost effective.
You're telling me, that's exactly what had to happen to convert Fab12 to 12". It was a significant undertaking and the cost was large enough to be marginally better than building a new fab from scratch.
Fortunately, when Fab22 was built, it was built with 12" in mind. It is structurally built for 12" tools and the AMHS track is even installed in the ceiling. When it converts eventually (Who knows when, they keep jumping around on dates internally), it should be a lot easier than Fab12 and much faster to ramp.
Well, they (we, I) were right; it is a three legged DOG!
http://www.amdzone.com/index.php/reviews
/60/9457-phenom-x3-8750-review?start=1
That said, even a busted clock is right twice a day.
SPARKS
http://www.eweek.com/c/a/Desktops-and-Notebooks/Intel-Weathers-First-Quarter-Downturn/
Only news I have seen to hit the net.... my memory was a bit off, I thought it was 79%, 78.5% ... went back and checked, sure enough.. 78.5%, AMD 20.6%, and Via 0.9%
Jack
Ok, so we have a $3.5B plant. It’s obviously incredibly sophisticated. Then you have all the nice guys on this site who work so hard for ‘the company’. So, I ask all you boys. What’s the best drop of nectar you could squeeze out of this thing?
A brand new shiny, ever loving, FANBOY pimping, juicy, fat, stupid expensive,…..
QX9770
Yes, dear process/manufacturing genius buds o’ mine, I just dropped Da Bomb.
Giant, I kept it under the radar with a NewEgg Preferred account; the WAF (wife acceptance factor) was extremely low on this purchase.
Orthogonal, the aluminum side panel of my Lian Li case is out at the paint shop, your well endowed goddess is taking form.
Doc, I thought of this as a celebration of sorts, as this is nearly my first anniversary on your wonderful site. With the extremely talented and knowledgeable people here, I thought I might share my ‘Extreme’ exploits with you all. (I’m surprised my extreme attitude didn’t get my ass thrown off.)
Jack, we are definitely going to put your motherboard trace pattern theories to the test on an X48, P5E3 Prem., ASUS board. We ARE shooting for 2G FSB (well at least1950). By the way, next month, 79.5% or 80% will not mater, it will be higher anyway.
GURU, I will be testing gate leakage through Hi-K on a more practical, real world, level. My PC Power + Cooling 1KW PS is very much up to the task, HOO YA!
In The Know, we are going to find out, first hand why it is absolutely necessary to build a new 4B plant for the likes of this bad boy.
Tonus, I’ll keep you posted on the unlocked aspects, perhaps you’ll reach for your wallet, too, as did I.
Ho Ho, I’m going to give you numbers you can beat Dementia and the Share-A-Coo-Coo to death with.
Thanks to all for a very good year, (and a nice dividend check).
I hope I can share my excitement with you all, without being too pompous. After all, this is what you all worked so hard for, all in one little LGA775 package. It is the most powerful processor on the planet.
LEAP AHEAD
(oooh, I’m so excited) :) :) :)
SPARKS
Sparks ...
Yep, unfortunately for AMD the Phenom (be it quad or tri core) is still too weak to make up for the delta necessary to give AMD any kind of market leverage. I agree, AMD will be below 20% MSS by Q3 or Q4 this year. This is assuming AMD doesn't go into kamakazi mode and drop prices 30% across the board... only then could they rationally stop the migration.
When your competitor can pin you down on your top bin with their lowest bin (of a product launched over a year ago and in the process of retirement), the hope of gaining any pricing leverage pretty much evaporates.
If you think about it, over the past 6 quarters AMD has materially lost about 440 million/quarter (on average), this is roughly 34 million a week, or 4.9 million per day.... so Hector/Meyer wake up, go to work and then punch out with the thought "I just pissed away almost 5 million dollars today" ...
This company is so mismanaged it is painful, and frustrating at times to watch.
I don't like Microsoft not because I don't like the software, nor they way they do business (I do not care), but because they are the only player... their product is too expensive and too slow to progress (it took 7 years for them to release Vista POS) ... if AMD evaporates, my fear is that x86 CPU will slow to the same pace... I don't want that.
Jack
Jack,
“This is assuming AMD doesn't go into kamakazi mode and drop prices 30% across the board”
Do you think it could get worse? I mean really, I thought Wrecktor was in Kamikaze mode for the last five quarters!
There’s something else going on here, I’m guessing, but I believe there are a couple factors that aren’t so obvious. At least, that’s what I’m getting from my reading. One, I don’t believe the channel ever forgot how they were dumped by AMD in lieu of the OEM’s back in 2006, retribution reaps it own rewards. Two, I really don’t think their lineup is attractive at any price. Nobody wants the B2’s.
Further, as a possible third, there’s a lot of OEM’s and major companies, that were hurt waiting for the Barcelona that never materialized. I can recall the way AMD spun things when CRAY took a MAJOR hit. When Barcelona did materialize it had the TLB which could have only made maters worse for vendors who were geared up and ready to rock based on a years worth of PowerPoint launches.
Look, if you and I are building computers based on a specific processor, and we get badly burned, do we wait (twice), or do we find another way? Conjecture, I’m sure, however, even if 25% did find another ‘more reliable solution’, they aren’t going back, not at any price. What’s even uglier, with AMD’s current performance metric, why bother? The OEMs are probably patting themselves on the back for making the, excuse the expression, smarter choice.
30% of 200 bucks, BFD. Hey, I just got a deal. I saved 60 bucks at NewEgg. They dropped the price of QX9770 from $1530.00, to 1470.00, 60 bucks. I don’t care about the sixty bucks , not when I'm HP competing with DELL for the server market.
The important margins AMD lost will stay lost, even with 30% price cut. In the short term, they may pickup a few points in the low end ghetto. However, it will only make matters worse in the long term. This is gospel.
I don’t know how these f**king guys sleep at night, while collecting millions.
SPARKS
'so Hector/Meyer wake up, go to work and then punch out with the thought "I just pissed away almost 5 million dollars today" ...'
Jeesh, don't exaggerate... with inflation it's more like 4.8mil/day. (See it ain't so bad!)
And I think Hector's punching out with the thought "I just 'earned' $56,000 today while the company lost $5mil!" (~14mil last year @ 5 days/wk * 50 weeks). He earns in a single day what many families hope to earn in a year and he gets that money while wrecking a company!
"I just 'earned' $56,000 today while the company"
Come on, give the guy some credit. The math is wrong. It's $28000 a day, while wrecking the company. Such a bargain!
SPARKS
"Come on, give the guy some credit. The math is wrong. It's $28000 a day, while wrecking the company"
14mil / 250 days = 56K/day (I'm assuming 5 days a week and 50 weeks..though I'm gonna go out on a limb and say he may take more than 2 weeks of vacation).
Sparks... you bring up some good points.
First, thanks for spelling Kamikaze right, I thought to myself I should go to dicitonary.com and check... but didn't :)
The wounds AMD has inflicted upon itself with the reent spate of quality and product delays will not heal easily nor quickly. AMD's history is riddled with these type of execution problems. The term 'flawless execution' has made a very obvious exit from their lexicon.
The worst part for AMD is, as you state, the channel will not easily forget ... and it was the channel that basically made them and kept them going during the other period of dark days in 2001-2002.
AMD is self imploding, and 10% reductions and fancy talk of 'Asset Smart'(tm) is not going to stop it... They need to do something vere radical -- perhaps the Asset Smart(tm) plan is still in the works... one thing is certain, whatever Asset Smart(tm) is and what it will do is not known to anyone but AMD.
AMD's latest CC's may be smoke-and-mirrors spin-fest's by flippant management, but when it comes to reporting to the SEC, it's nothing but the cold hard truth.
From AMD's Q4'07 10K filing. It's still a month or so away before we will see Q1'08 filing.
http://www.sec.gov/Archives/edgar/data/2488/000119312508038588/d10k.htm
If we cannot generate sufficient revenues and operating cash flow or obtain external financing, we may face a cash shortfall and be unable to make all of our planned capital expenditures.
In 2008, we plan to make approximately $1.1 billion of capital expenditures. Our ability to fund capital expenditures in accordance with our business plan depends on generating sufficient cash flow from operations and the availability of external financing, if necessary.
Our capital expenditures, together with ongoing operating expenses, will be a substantial drain on our cash flow and may decrease our cash balances. As of December 29, 2007, we had $1.9 billion in cash, cash equivalents and marketable securities. During 2007, we incurred substantial losses that have had a negative impact on cash balances. During 2007, net cash used in operating activities was $310 million and net cash used in investing activities was $1.7 billion.
The timing and amount of our capital requirements cannot be precisely determined at this time and will depend on a number of factors including future demand for products, product mix, changes in semiconductor industry conditions and market competition. We regularly assess markets for external financing opportunities, including debt and equity financing. Additional debt or equity financing may not be available when needed or, if available, may not be available on satisfactory terms. Our inability to obtain needed financing or to generate sufficient cash from operations may require us to abandon projects or curtail capital expenditures. If we curtail capital expenditures or abandon projects, we could be materially adversely affected....
We may not be able to generate sufficient cash to service our debt obligations.
Our ability to make payments on and to refinance our debt, or our guarantees of other parties’ debts, will depend on our financial and operating performance, which may fluctuate significantly from quarter to quarter, and is subject to prevailing economic conditions and financial, business and other factors, many of which are beyond our control. We cannot assure you that we will be able to generate sufficient cash flow or that we will be able to borrow funds in amounts sufficient to enable us to service our debt or to meet our working capital and capital expenditure requirements. If we are not able to generate sufficient cash flow from operations or to borrow sufficient funds to service our debt, we may be required to sell assets or equity, reduce capital expenditures, refinance all or a portion of our existing debt or obtain additional financing. We cannot assure you that we will be able to refinance our debt, sell assets or equity or borrow more funds on terms acceptable to us, if at all.
Ouch, the truth hurts
JACK
“and it was the channel that basically made them and kept them going during the other period of dark days in 2001-2002.”
AMD fulfilled a nice little niche for itself. Don’t, by any stretch, assume I’m being condescending by referring to the low end, white box market. There was the other end of the market, HPC, ironically that AMD held. They covered both ends quite well, especially in HPC where their products scale excellently, simply by architectural design. (I don’t know which one of you guys straightened me out on this scaling thing last year, In The Know, I think). In fact, as you are well aware, they are still tenuously holding on to that market to this very day.
I can’t say INTC was happy about this arrangement, which would be ridiculously presumptive. However, from all evidence, AMD was at least tolerated by INTC. Additionally, as someone else on this site once pointed out, the amount of pure profit AMD made over the years, didn’t really amount to much collectively, until Opteron. With Opteron’s performance, and INTC sitting on its’ then fat ass, they were making some real money. AMD was taken for a serious player, rightfully so. (2004-2006)
Here is where the crossroads diverge. The correct paths lead to a higher performing processor, WITH AN OPTIONAL LOW END COUNTERPART. It was indeed essential to maintain the inroads they worked so hard to obtain. An esoteric native architecture, exceedingly large, quad core solution, with a process that was approaching its limits, was just plain stupid. Basically, too much gamble, not enough options, as demonstrated their present 3 core conundrum. They had a lot of cash, a substantial lead, good margins, and as a side benefit, market exposure. Opteron =opportunity.
What did they do? They got cocky. Right out of left field, their answer was ATI. They sold the ‘platform’ purchase to the entire industry (as well as the market) on both the merits and benefits of this acquisition. Many bought it. I didn’t, not for a second, not with NVDA looking up their asses. (Just look at the money NVDA made in 2007!) AMD made the ATI shareholders an “offer they couldn’t refuse.”
The value of both companies together halved overnight.
It was a bad move, a rookie move by an arrogant management team, riding high on the wave, which wasn’t ready for primetime, big time. The ship got larger right underneath their feet as soon as they signed on the dotted line. The ship wasn’t as easy to turn as used to be, once they set it on a course. This was a classic case of the Peter Principal that states an individual in their own time will seek their own level of incompetence.
Wrector, Dork, et al.
Literally at the same time, fundamentally, it was hideously stupid and myopic to believe small things like shortening a pipeline, staying with a proven process, and adding one more IPC to the equation could have such a radical effect on performance and to the entire industry, practically, over night.
Enter CORE.
The maverick engineer geniuses in Israel saw it, quite clearly. In fact, insiders at INTC who didn’t like Centrino, nicknamed it Latrino, like the toilet! Big Paulie put a stop to that shit. He cut losses, chopped heads, squeezed out the fat, hence, CORE, CORE2, and CORE2 MCM.
“I think you'd be misreading the market if you think people care about the packaging," - Paul Otellini
Fast and flexible, this approach is precisely what could have kept AMD in the lead easily by coving both the high and low end simultaneously. AMD incredibly MOCKED the idea, publicly! Incidentally, INTC saw this high end low end thing so well; they came up with two separate solutions to cap the market, ATOM and Nehalem, the niche AMD had come to occupy to begin with. Presently, in perfect 20/20 hindsight, AMD has no where to turn later this year.
A ridiculously expensive 5.4B graphic card company squander/purchase and a ‘Hail Mary’ processor design/process (a La GURU) was not the answer, holy Christ! I knew it then and I know it now. These kinds of mistakes will never get a pass from me. AMD wanted a fair fight with the big guy, that’s exactly what they got. They botched the opportunity in spades back in 2006 when they had no debt and over 2B in cash. The CORE answer was the key to survival. Holding on to a lucrative customer base was mandatory.
I believe INTC is hell bent on absolutely crushing AMD, and they’re going to do it. INTC had the gun. AMD gave them the ammunition, and they waltzed right into the line of fire.
SPARKS
I think one of the main problems on the AMD front was their K10 rollout strategy... yeah they needed something to compete with Intel's quads, but for desktop at this point in time who cares (if you are in AMD's position)? They definitely needed server parts, but why so much focus on quad and tri-core desktop parts? They could conceivably attempt to get away with the energy efficient/value crap in server land... but how many quadcore desktop folks care about an extra $20 at this point or "drop-in" replacements if there is a significantly better performing part out there?
AMD has taken a weak performing part and tried to force it into a high end market, only to be forced to cut it's price which collaterally hurts all of AMD's other downstream desktop parts.
After shooting one foot they then turn the gun to their other foot and make matters worse by INSERTING another price segment between quads and duals with this tricore ridiculousness! Sure they save a dime now by selling scrap - but what happens when they (theoretically) fix the yields? They are then forced to abandon the product (which will probably make OEM's real happy again) or start disabling good quads and hurt the margins.
And then consider the long term impact this has to the high volume dual core market that will be felt so long as tri-cores exist. So they have managed to fix a short term problem (yields on quad core parts) into a long term nightmare (depressed pricing on volume dual core parts)
So why no dual core parts? It's been 3 quarters since K10 was theoretically launched and no dual core parts?
Forget about competitiveness with Core2, I'm going to continue to speculate that the K10 dual core part barely exceeds K8 dual cores and the only hope is to raise the clocks. If they can't then I guess they might as well continue to crank out K8's and dribble out K10 quads and tris.
I've heard the "they got cocky" argument for a while and I'm not sure I agree with that... I think they've had to have been working on K10 all along, and I don't think they slowed down or coasted on the development of this - I think they simply misread (and mis-executed on)a few critical pieces. I think they misread the manufacturing intricacies on such a large die. I think they misread how big an impact Core would have over P4. I think they misread and mis-executed on the transition to 300mm and also to 65nm (and soon 45nm) technology. Recall AMD's original strategy was to start up 300mm on 65nm - they either had significant issues on 65nm (which I think was the case) or made a painfully bad short term decision to pump up the 90nm output at the expense of slowing the 65nm transition. They will soon have a bunch of tools if/when they migrate to 45nm with a lot of depreciation left (thank to JJ for the 6 year link) that they will suddenly have to eat.
And obviously they misread the value of ATI. I think they saw Intel's platform margins (esp Centrino) and thought we can do that too. However the size of the platform market is pretty much the size of the CPU market share, so if you don't grow the CPU share (which requires better cores), why would you think simply having a platform would fix this?
AMD naively thought that the platform would drive the market share, but didn't realize the CPU drives the platform share. Centrino was successful because it had a real strong mobile core that linked up to the other chips. Put an underperforming CPU in there and Centrino's success doesn't happen.
BTW - Intel similarly mis-read the platform strategy with the Viiv (as did AMD with their "Live") on the desktop world.
“I've heard the "they got cocky" argument for a while and I'm not sure I agree with that”
“I think they simply misread (and mis-executed on)a few critical pieces.”
“ I think they misread the manufacturing intricacies on such a large die.”
“I think they misread how big an impact Core would have over P4.”
I don’t wish to sound trite or facetious, are you sure you’re not agreeing with me here?
The points you sight are solid, I basically said that. The issue here, if I may, is the word “cocky”.
When you combine the three above misreads,
Throw in a dash of ‘picnic’ flyby skywriting.
Pepper Manhattan with the ‘Total power wasted by INTC Processor’ billboards,
Kick in a bit of “screw the channel, we want DELL” and,
Pipe in ‘Screw Consumers and OEM’s with 939, we’re doing fine”
What else could they have done before the misread light went off and the cocky light went on?
SPARKS
jack: "When your competitor can pin you down on your top bin with their lowest bin (of a product launched over a year ago and in the process of retirement), the hope of gaining any pricing leverage pretty much evaporates."
This is what comes to mind when I see people boasting how AMD has managed to get their CPUs into a very attractive price-performance niche. They will tell you that an AMD Phenom XXXX compares very well to an Intel XXXX CPU and it also costs less, sometimes a fair amount less (say $40, $50). They remind you, with a sneer, that Intel dare not drop its prices any further!
Well, I agree that this becomes a great situation for the consumer-- you can get a quad-core CPU with good performance at an attractive price point, and you might even squeeze a few extra hundred MHz out of it. It's like having your cake and eating it, too. But this isn't good for AMD. AMD doesn't want its strategy to be "let's undercut Intel on the low end of the scale" when it doesn't have any other products to offer.
Being price-competitive at the bottom end of the scale leads to situations where Intel is recording record quarterly earnings and AMD is trying to figure out a way to manage a financial situation that brings to mind the term "arterial spray." By all means enjoy the excellent price/performance ratios that are available, but don't think that the current situation is something that AMD would look at as a positive.
"Being price-competitive at the bottom end of the scale leads to situations"
On top of this keep in mind the 2 products are not cost competitive! Intel's production cost of 2 dual core 65nm vs AMD's monlolithic quad? Now factor in the 45nm dual core shrink (and obvious impact on 45nm quad MCM's).
The problem with AMD's strategy of trying to price their way to market share (or at least competitiveness) is that Intel has both more manufacturing capacity anf a lower production cost. The 0MB L3 cache 45nm product may help but even then it'll be competing against a relatively cheap to produce MCM of 2 fairly small 45nm dual cores.
"What else could they have done before the misread light went off and the cocky light went on?"
I see the misreads as management INCOMPETENCE not cockiness (the dual core for dummies book and various PR stunts notwithstanding)
Or to put it this way - if they weren't cocky, would they be in a vastly different position today (would K10 be a better product? would 65nm be any healthier? Would bulldozer not be a paper plan at this point? Would they have been able to implement high K at the start of 45nm?)
I think the only real questionable action in terms of cockiness is the decision between MCM and monolithic approaches. Had they not been so cocky about K10's performance, perhaps they would have had an MCM backup plan (but again this could be more ego and/or incompetence than cockiness)
I got only one thing to say
Sharikou and Scientia I told you so you stupid f***s
You never believed me when I told you money and capability would beat stupid luck.
“Or to put it this way - if they weren't cocky, would they be in a vastly different position today (would K10 be a better product? would 65nm be any healthier?”
Knowing your expertise in the industry, I wouldn’t dare argue the point, it’s yours. However, exercising prudence before the acquisition with two powerful adversaries, that’s quite a different scenario. I’ll take a shot at that.
Railroading the company and it’s assets into the buyout with horse blinders on while diverting a lot of time, money, and resources towards the acquisition became a goal in itself. All I’m saying is the energy directed toward that goal should have been redirected toward their CORE product. At the very least they wouldn’t be burdened by the monumental debt they are today.
“Had they not been so cocky about K10's performance, perhaps they would have had an MCM backup plan.”
This is the point. They might have in the interim developed another more competitive/lucrative solution, maybe.
Granted, the process was shot, the course was set, but the stupid bastards wouldn’t be facing BK.
Potatoes, potaatoes, I call it cocky and arrogant. Besides, I think we're spliting hairs here.
SPARKS
Roborat
The self named "Intel Fanboy Review" has shut down. I had linked to his site for the purely Intel perspective but he stopped posting in January; Sharikou180 hasn't posted since last December. Since you have the only Intel perspective blog still in operation I'm replacing his link with yours. Sharikou was already linked as the purely AMD perspective.
"Sharikou was already linked as the purely AMD perspective."
That's rich - as if Dementia from AMDzone is not a pure AMD fanboy blog. He should link to his own site and see if people notice!
"All I’m saying is the energy directed toward that goal should have been redirected toward their CORE product."
I think we are saying the same thing, just differing on the "why" part. Perhaps it was a combination of imcompetence AND cockiness/ego?
And quite frankly for AMD's sake the more Ruiz focuses on non critical things, the better off they probably are as he is less likely to mess the critical things up - the only issue is if he starts siphoning away resources (like cash in the case of the ATI acquisition).
It is frustrating that the board has not done something yet. The board should not micromanage every decision or push someone out based on a single mistake - but I don't think anyone would dispute there has been a trend of bad decisions, execution issues (with no backup plans) and most importantly bottom line performance (the buck should stop at the CEO's desk, no?). You can't plan to the best case scenario and just blisfully assume everything will go according to plan and make excuses.
My frustration is not over AMD's downturn (hey, that's capitalism) - it's the impact to the shareholders. Normally you can say well people should just pull their money out - but AMD is ~78% institutionally owned, so most folks probably don't realize they are paying for mgm't mistakes via mutual funds, 401K's, etc...
“Intel perspective blog”
Uhhh? I think most of us on this site know more about AMD”s process, manufacturing, financials, and market position than any AMD fanboy he’s ever known.
Maybe, he should read Khorgano’s SEC post on this page to get a true AMD perspective on AMD.
SPARKS
Goodness... the latest misinformation on Dementia's blog just gets better. So I understand someone with no manufacturing and fab experience is explaining fab manufacturing? Someone who states Asset Smart has yet to be defined, knows what it is?
And then there's these gems:
"Faster cycle time means more chips with the same tooling."
This shows a lack of understanding between cycle time and tool throughput! One of the (not so great) ways of reducing cycle time is to have an excess of tools. I'm not saying this is the case with AMD, but it is a way). You could also buy 25 tools, split the lot up into 25 separate lots and get much better cycle time, though the capital costs would not be so good.
Lower cycle time CAN mean more chips, but it doesn't have to! If Scientia knew some of the key contributors to cycle time, he'd realize the folly of his conclusion.
"Reducing batch size would allow AMD to catch problems sooner and allow much easier manufacturing of smaller volume chips like server chips."
Hmmmm... wasn't APM 10.0 (or whatever version # you want to put after it to make it sound 'official') supposed to take care of this and proactively catch and adjust for problems? This is important when you run a process on a cliff and you are constantly falling off (or having the safety line catch you). If you design a process not near the cliff you do not have this issue. It's the difference between engineering a process vs engineering a manufacturable process.
And what is most humorous it that batch tools are typically FAR better in terms of wafer to wafer variability. Yes there are batch to batch issues (if you don't know manufacturing) - but try to grow a 12A film on 100 wafers - do you have a better shot doing this in one batch in a vertical diffusion furnace or attempting 100 growths on separate batches? I've spent some time studying this you have all sorts of process chamber memory and first wafer effects with single wafer tools - these can be dealt with but are not trivial. In fact today (as many of the fab guys on the site can probably attest to), one of the ways this is dealt with on a brute force method is to run 'warmup' wafers prior to a production lot.
Oh and anyone want to guess the economics between single wafer vs batch? The capital cost to process the same # of wafers is a wee bit different (hint batch is typically FAR cheaper from a capital perspective AND takes up less fab space).
Scientia makes it sound like AMD is breaking new ground with all of this - these type of discussions have been going on for years, if not decades in the semiconductor industry. AMD is pursuing this because they see the writing on the wall - the only known way to significantly reduce cost/cm2 of processed Si has been to increase the wafer size. In fact Sematech has a program called 300mm prime to do this (though I suppose 'next generation fab' sounds much cooler from a PR perspective) to appease the folks who were crying about 450mm.
With costs of things like lithography increasing at the pace it has been, unless people figure out how to dramatically reduce 300mm cost through things like cycle time, 450mm is inevitable for large manufacturers. The outlays for this up front for an HVM fab and the development costs will be enormous and AMD realizes they would be at a ridiculous disadvantage if 450mm happens - in their current state they simply couldn't afford it.
And the topper, is that he spits out a bunch of impressive sounding statistics by AMD, but fails to put them in ANY perspective! (i.e. how did the overall industry fare on these same metrics over the same time period).
It never ceases to amaze me how he can just look at #'s he really doesn't have any fundamental understanding of and draw conclusions so easily.
So here's my theory:
Asset Light was an "oh shit, we have to tell Wall St something"... they realize we don't have the cash to build out the capacity we originally planned to, so we'll tell them we plan to do more with less!
Asset Smart is... well we told them Asset light last year and really did NOTHING, so we need to change up the name to make it sound different and like we're making process.
So Hector you are going to try to do more with less! FAN-FREAKIN-TASTIC! Thank you captain obvious! Here I was thinking you were going to try to do less with more resources! All he needs to do is weave in some 'out of the box', 'shifting to a new paradigm' catch-phrases and it'll really sound impressive! I'm sure he'll get a bonus for executing the asset smart plan.
Some good points in the last post.
Scientia is one stupid dude, or he takes others to be stupid and he like Hector is trying to hoodwink everyone.
Scientia you are one DUMB F**K
Cyle time or mask layers/day only tells you how fast you move wafers thru the factory. It helps you learn as mentioned. But if you don't increase the wafers processing capability it doesn't really do much for you.
BS like smaller batches may get a wafer out faster, but won't really many more chips. The reason is the tool it self isn't capable of producing more or faster. You are limited by the inherent capacity of the furnace, stepper or whatever. You get more output by getting more steppers, more furnaces. Things like SPC, APC, Mask layers/day are just BS.
Also so what if the wafers get out faster. It clearly looks like all AMD got out was crappy defective chips faster. It still took them a good 9 more months. All those fancy words didn't work in the past, why should we believe it will help them at 45nm. They were in deep doo doo already and they couldn't do it. Being deeper in $hit now makes them more motivated? I don't think so. AMD is missing all the fundamentals. I told many people that 2 years ago. It was only time before they go BK, not if or when.
Hector and the AMD masturbators talk using fancy names about things that they hope to confuse you because they bottom line is they can't make money. So they invent complicated names for simply things to try and cover the fundamental things that keep them from profitablility.
What are they
1) Invest in technology from research. This is stuff for 8 to 10 years in the future and takes hundreds of millions of dollars a year. AMD doesn't have this money. That is why they were late to strained silicon, late to High K,
2) Dedicated process team to work with design team to formulate accurate SPICE models and DFM that accuratly allow the designer to simulate their circuits. AMD doesn't have this because IBM is doing their silicon and IBM has many customers from cheap low-power to their internal PowerPC teams to keep happy. They don't have a dedicated team to tweek and model to AMD's need. The result is slow suboptimal designs and surprise like recent 65nm that are no faster and have little power savings over 90nm. I'll wager the TLB is due to poor process to design characterization.
So, since AMD can't do the obvious they talk and make up fancy names for things everyone in industry already use. Things like SPC, APC, APM, LEAN, SMART are used by everyone. Every single manufacture from bolts to cars, to semiconductors use these concepts. It isn't the concepts that cause you to produce better products and profits. Its the underlying investment, technology and execution.
AMD has none of that and thus spews BS out its backside.
BK I say heah Sharikou... another sucker LOL
"My apologies to anyone who was hoping that Asset Smart meant selling off pieces of the company. Sorry."
He should appolagize for lying like Hector and his merry lot. Scientia is a smart person, but a blind propaganda machine for AMD. With this post it should be so obvious he is spinning what he knows is a no win , BK situation for AMD
Well, after reading an authoritative statement that everyone in the world but our friend on that other blog is wrong (with no supporting evidence), I've come to see the error of my ways. Clearly no-one on this site is smart enough to see the truth or draw correct conclusions. Oh yeah, I need to throw all the media hacks into that pile of idiots as well. What would it be like to be the only one in the world with a clue? I guess I'll never know.
Now that we have that out of the way, let's get down to brass tacks.
SPC in 1998? What a joke! SPC has been around since Demming introduced it in the late '40s/early 50's. As to reducing cycle time, I just about had a seizure when I read that. If anything, it slows down your cycle time because it requires you to stop processing and measure your wafers. So why do it? Because it is a huge yield enhancer. Might I recommend that someone read “Statistical Methods for Industrial Process Control” by David Drain to find out what SPC is and does? The book is a good text on the application of SPC in a manufacturing environment.
As to increasing cycle time (or if you prefer reducing throughput time) here is what Paul Otellini had to say about what increasing cycle time will buy you.
Our manufacturing organization is delivering faster factory throughput times, higher yields, and improved equipment utilization. As a result of their efficiencies, we are now able to lower Intel's 2007 capital spending forecast by about $600 million without compromising our ability to service a growing computing industry.
There are a number of benefits that we derive from our throughput time reductions. We have reduced Intel's pipeline inventory by several weeks, enabling us to meet customer demand with lower inventory on hand. We are able to iterate new products more quickly, decreasing the time to healthy silicon. Lastly, shorter throughput times allow us to start new wafers closer to the time our customers need the output, giving us truer demand signals and a better ability to respond to a changing mix.
You will note that nowhere in there did Intel say that they were able to start more wafers. Why doesn't increased cycle time help this? It is real simple. If a tool takes 15 seconds to process a wafer, it takes 15 seconds. Staging more wafers in front of the tool, or staging smaller lots doesn't change the 15 seconds. You still get 4 wafers a minute out of the tool. You can get more wafers out of the fab by either buying more tools as someone already mentioned, or by finding a way to shorten your process.
As Paul Otellini pointed out, what you do get is less inventory sitting around in the factory waiting to be processed.
And if AMD is just now figuring out they need to crank up cycle time, they are a bit late to the party. Intel published this (see slide 10) in Sep07 showing they had already reduced cycle time by 50%. Note that the time frame shows this trend beginning in Jul06. The following slide will show you the impact on inventory tied up in the factory.
And the idea of eliminating batched tools is classic too. The industry has only been trying to find a way to do this for the last 20 years. Of course now that AMD has decided to get serious about the issue it will all be resolved shortly. NOT!
Batched process exist because the process steps take a long time. It is not uncommon for the key processing step in a batched tool to take 30 minutes. If you processed a standard lot of 25 wafers one wafer at a time, you only need to apply basic arithmetic to see that each lot would take you 12.5 hours. If you load the entire lot (or better yet several lots) into the tool all at once and process them together then you get the whole thing processed in 30 minutes. So which would you choose, 12.5 hours or 30 minutes.
Now before someone bites my head off and spits it across the room, I know that batched tools take a lot longer than 30 minutes to complete the process. You have to load the wafers into the tool, ramp temperatures up , run the process, ramp temperatures back down, let the wafers cool, and unload the wafers.
Let's say this all takes 12.5 hours just for the sake of argument. So there is no net gain for running one lot. But if you put 4 lots on the tool, you have effectively cut the processing time for those wafers to an average of just a touch over 3 hours.
That is why you batch wafers and that is why going to small lots is bogus. The efficiency in batching comes from batching as many wafers as possible together. Running smaller lots puts fewer wafers in the batch and hurts throughput.
The basic problem is that Scientia has is mixing different concepts because he doesn't really understand them (other than what AMD publishes). SPC and to some extent APC are control schemes to make sure the process stays 'in control' (ie within acceptable variation) so you can continue to crank out production wafers. The lean and smart concepts (as far as I can tell) are how to get more output or more manufacturing flexibility out of a given amount of capital equipment. This can be accomplished a variety of ways - speeding up the tool, reducing the downtime/maintenance, improving tool sharing with other process steps or even other technology nodes, optimizing wafer flow and technician allocation to maximize the constraint tool utlization.
The problem, as I've written previously and intheknow also mentioned, is that these are not new concepts. You can dress them up and give them new names but these are pretty general and quite frankly obvious concepts to all manufacturing. In fact, other industries are often benchmarked as many of the learnings can be utilized across many different industries.
The problem, and the issue with Scientia's whole rationale, is that there is a cost to improved flexibility - and there is a cost to going to all single wafer (or mini-batch tools). It may be capital, it may be resources to maintain the additional tools, it may be more fab space required - but I can tell you it is not free (if it were everyone would be doing it).
So AMD can hope and wish that some miracle solutions arise that for some reason haven't been found in over 20 years or they need to be prepared to pay the piper for it. While it is good to challenge the status quo and ask has anything fundamentally changed that might lead us to the solution now - I think AMD is being quite a bit disingenuous by taking what occurred in one fab, on a different technology node, on a different wafer size, with a different automation system and say well this means we can do it on 300mm as well. They also do not comment on whether they reason they got such large improvements might be because the starting point was so poor. As they do not reference other data from other IC folks, it is impossible to conclude much from it.
BTW - what Scientia fails to mention with his rattling off of all the AMD improvements in F30 - those somehow were done WITH CONTINUED USE OF BATCH TOOLS IN THAT FAB! Once again this is a nice 3 card monty game - change up a bunch of variables, throw in a bunch of data on different baselines and attempt to compare them and then conclude (or at least infer) something that really can't be concluded.
Finally, while DML (days per mask layer) is a nice pretty benchmark statistic - what really matters in cycle time is start to finish... Each new generation has an increasing # of mask layers and sometime the average # of process steps between mask layers may go up or down as well.
So you can have identical DML between say 130nm and 65nm technolgies but the cycle time of that lot WILL BE VASTLY DIFFERENT! Again another nuance (actually this is more than just a nuance for those in manufacturing) that Scientia doesn't get!
They are just trying to feed the press some raw technology meat - AMD fully knows that the press will not be able to digest it and will simply just regurgitate it.
Of course I like to check my raw meat for things like Mad Cow or Ecoli, especially if it is coming from the AMD slaughterhouse.
I posted this on Dementias blog this morning. 1st time post actually, and he promptly deleted it. I suppose he claims to be unbiased, I think not. However, I kept the copy for your entertainment.
“Sir,
Please, by all means, keep up the good work. Your posts contribute greatly to another blog I’m sure you’re quite familiar with, Doctor (PH.D), Roborats, AiMeD. Please understand that I am just an enthusiast who is crazy enough to spend ridiculous amounts of money on only the best components. (I think I’m tolerated at the site purely from an outsiders/layman’s/end user’s perspective) That said, they have been very patient with my lack of knowledge in the field.
Frankly, as far as I can tell, with my extremely limited exposure (read: none) to the intricacies of chip manufacturing/process engineering, your posts initiate very lively discussions that ultimately contribute to my education without being a burden to the very experienced individuals that contribute to the blog. At times, very recently I might add, you have a way of stirring up a hornets next of activity, that allows me to research the terms and concepts discussed therein.
You see, in comparison, as they are zooming up in the stratosphere of process engineering, I can sit back from my modest lawn chair and enjoy the air show and fireworks. They are, by the way, brilliant displays! Thank you!
And please, stop by more often, as I am currently building a very fast machine with the newest Intel chipset X48 (ASUS P5E3 Prem,). The machine is equipped with very fast DDR3-1800, but with only the very modest 65nM Q6600 (G0). I’ve managed to run it at an extremely stable 3.38 GHz, 24/7 on air.
I recently purchased a VERY expensive QX9770, Hafnium processed, 45nM 3.2 GHz, 1600 FSB Intel Extreme Edition processor. I think I’ve generated some interest on the site, however, it’s hard to tell, as they all very modest (albeit brilliant) fellows. Frankly, I believe some of them contributed to its development, of course, they’ll never say.
Again, please stop by, or perhaps with your permission of course, I could stop by here to keep you posted on my progress on my newest build. It will be the very fastest machine to date, purely from a real world, retail, end user/enthusiasts perspective.
Yours Truly,
SPARKS”
Now he's a coward.
SPARKS
Scientia is completely off base on what Asset Smart means. Hector has categorically implied that Asset Smart involves something that AMD do not yet wish to reveal to the public for fear of Intel trying to interfere with its progress. As has been explained here, the proposed ISMI efficiency gains in 300-mm wafer processing are already common knowledge in industry and AMD would gain nothing by attempting to keep top secret their proposed manner of implementation of these techniques.
The only plausible explanation that would justify this level of secrecy is that Asset Smart involves a major deal with some other entity, and the identity of this other entity cannot be revealed until the time is right. If I had to wager, I would guess that the deal has something to with selling fab capacity but we will never know for sure until Hector chooses to spill the beans.
At least Scientia is now admitting that Hector will not explain this mysterious strategy. All last year Scientia was convinced that Hector never hid anything, everything was already fully explained. As usual, like nearly all of his predictions since mid 2006, he was grossly incorrect. As he will be again this time.
"As usual, like nearly all of his predictions since mid 2006, he was grossly incorrect. As he will be again this time."
Come on you must be forgetting his entire blog on how K10 was a good start... ok bad example...
Well there was this one about how AMD is closing the technology gap... shoot...
OK there was this on about how great DTX would be and leave nothing with Intel to compete against it... darn.
I'm sure if I look hard enough I'll find something (give me a few days)
Also it is pretty clear from AMD's insinuations during the past few earnings calls that:
1. Asset Light and Asset Smart are the same thing, but AMD has recently chosen to use the latter in order to avoid the implication that it involves the sale of fab capacity. This of course doesn't preclude the possibility of it involving the sale of fab capacity... As we know from the Barcelona propaganda experience in 2007, AMD are quite willing and able to spin and hype things one way (and even lie e.g. "outperform Clovertown by 40%") and the reality turns out completely different. I wouldn't be surprised to see Asset Smart turn out to be "Fab Light", though I agree that such a strategy would probably be to AMD's detriment in the long run despite the short term CAPEX (and possibly OPEX) relief.
2. Asset Light and Asset Smart likely have nothing at all to do with the LEAN and SMART manufacturing efficieny projects.
Scientia said about Asset Smart: Most theories seem to focus on either the idea of selling all or part of a FAB to raise cash or on the idea of making graphic products at Dresden...
Incorrect. The whole "asset smart" idea revolves around AMD running a business with minimal CAPEX for the sole purspose of generating a positive cash flow.
In Q1'07 Hector Ruiz knew AMD's predicament. He knew that AMD's business model is broken therefore uttered for the first time the words "asset-lite". This only later changed to "asset-smart". If Scientia only has a little more understanding of what "assets" mean in the business world, he wouldn't be talking about process improvements and instead focus on how AMD can run a money losing business with one Fab idle and the other under-utilized.
There is a common understanding between the AMD management and the analysts when they talk about "asset smart". I find it amusing to see that this has gone above Scientia's head.
"2.) Scientia must be wrong because he doesn't work in chip fabrication.
Again, these are AMD's documents and conclusions, not mine." (Scientia's blog)
Actually he has played quite the shell game here and to many readers on his blog which blindly accept what he says this may seem like what the presentation said - but of course it is Scientia who has chosen not read and analyze the data carefully.
- AMD said they improved cycle time in F30
- AMD said they increased output in F30
- AMD DID NOT SAY THAT THE DECREASE IN CYCLE TIME (or as Scientia likes to misname it, Asset Smart) was the CAUSE of the increased output!
And another example:
- AMD said they improved cost per wafer (and on a side note they oddly took out capital and depreciation cost, which is not a standard way of economically viewing this as they are by far the greatest component to production wafer cost)
- AMD said they improved cycle time
- AMD DID NOT SAY IT WAS THE IMPROVEMENT IN CYCLE TIME THAT LED TO THE WAFER COST IMPROVEMENTS!
Again Scientia takes a couple of separate bullet points and assumes they can be correlated to each other! And then has the arrogance to say AMD concluded it! No, they actually didn't say it.
And this is where his lack of background in manufacturing comes in - if he is concluding it, I now have to question his background in making the conclusion. There's a difference between correlation and causation. He is correlating.
And it is amusing to see him continually fall back on his crutch of well these are anonymous posts. If you don't like the message, the best tactic is to attack the messenger (abinstein and Sharikou have also tried this technique). I have tried to post (non-anonymously of course) on his site numerous times only to have him reject the post because he doesn't agree with it. Or he will selectively edit the post (or his classic I'll just cut/past things out of context in his own comment and respond that way and conveniently omit stuff he can't refute). And finally he resorts to his 'prove it', yet many of his 'facts' are actually opinions masquerading as facts - much like his latest blog where he linked to an official looking document and then drew bogus conclusions from it.
Do we have to go back to his Ioff power consumption remarks again? "I" generally representing current to most people with a high school background - of course he lifted the 'concept' when it was planted here to see what he would do and started drawing bogus conclusions around it.
I have no problem with opinions - but the stuff in his latest blog are not AMD conclusions, they are misinterpretations of an AMD presentation to fit what he WANTS Asset Smart to be.
As "axel" correctly pointed out, if Asset Smart is simply making things more efficient and improving cycle time, cost, productivities, etc... why is Hector so secretive about it? This blog article is going to blow up on Scientia big time - he better start planning his exit strategy on this.
axel Hector has categorically implied that Asset Smart involves something that AMD do not yet wish to reveal to the public for fear of Intel trying to interfere with its progress.
i doubt this has anything to do with Intel or with competition. in a way the secrecy is understandable. a lot of money can be lost if things are said while negotiations are ongoing.
Sparks, Scientia deleted your post, but he did respond to it. That was the surefire approach that got me to stop posting there, but I'm sure he doesn't miss me.
But just for kicks, let's go through his points and see if they apply to what I posted.
Roborat's blog is a great home for trolls which he fosters with his reliance on anonymous posts.
I have never posted anonymously here, or on his blog. So I must not be one of the trolls.
1.) Scientia must be wrong because no one knows what Asset Smart means.
I'll plead partially guilty to this one. Someone at AMD knows what asset smart is, but despite his assertions to the contrary, I don't think he knows what it is any better than I did.
The document he refers to doesn't convince me since it clearly refers to Manufacturing SMART, not ASSET SMART. Last I checked, assets were something I owned while manufacturing was something I do. They don't seem quite the same to me.
2.) Scientia must be wrong because he doesn't work in chip fabrication.
You wont find his SPC--> APC--> APM --> LEAN --> SMART progression anywhere in the documentation he linked to. This is his own creation and has nothing to do with what industry he works in.
So what I actually said is that Scientia is wrong because he doesn't know what SPC is or does. Lots of people outside of the semiconductor industry know and use SPC and have for decades. So in this case he is wrong because he is making stuff up. Working in the industry is not a necessary precondition to understanding SPC.
3.) These ideas won't work because they are not unique to AMD or because they don't apply to every company.
Nope, I never said that.
4.) The numbers are useless because they do not include industry averages.
I didn't say that either.
5.) Switching to smaller batch or single wafer tooling isn't free.
No, it is not free, but if he thinks that was my point, he needs to re-read what I said. I said changing batch size will decrease the number of wafers you get out with the same amount of tooling. So you either need more tooling, or you need to change the process to get around that detail.
For the record, I have gone through the documents that AMD provided. Nowhere do I see a direct contradiction to anything I posted.
AMD has proposed to get around the lot size issue by getting the equipment manufacturers to make tools designed to run batches of smaller lots (i.e. mini-batch tools). That is a process change in my book, so I stand by my assertion.
There are also simpler solutions to this problem. To look at AMD's presentation, you would think that lots are fixed blocks of wafers that can never be broken down. That isn't true at all. Lots are split and reassembled all the time. This can be done to help move things through batched processes.
Again, these are AMD's documents and conclusions, not mine. Anyone who can actually read can look through the ISMI presentation and see what AMD's conclusions are. AMD concludes that shorter cycle time will save money. I would have no way of verifying this.
I do believe I said that reduced cycle time was good and provided documentation showing exactly why. I have no trouble verifying that at all. I also said that if AMD was just waking up to this fact, they are way behind the curve.
This is intuitively obvious, by the way and I'm sure AMD does understand this and has been working on this all along. The presentation Scientia is linking to is just proposing changes that AMD would like to see the industry and equipment manufacturers adopt.
In conclusion, Scientia's 5 reasons posts on this site would be irrelevant and/or nonsense don't seem to apply to my post. He has yet to address a single issue that I have brought up. I doubt he can successful address them.
In The Know-
I responded to his reply. Not quite what he would expect. I was a bit more objective and less tongue in cheek. However, I invite you to read the post as it is a summary on my part conserning both blogs. I centainly hope he doesn't delete this one for the sake of postarity.
SPARKS
Axel said...
As we know from the Barcelona propaganda experience in 2007, AMD are quite willing and able to spin and hype things one way (and even lie e.g. "outperform Clovertown by 40%") and the reality turns out completely different.
AMD has certainly messed up in a lot of ways, but I don't think the 40% "hype" was a lie. The current SPECweb2005 result shows Barcelona is clock-for-clock 40% faster than Harpertown, and 50% clock-for-clock faster than Clovertown.
Clovertown
Harpertown
Barcelona
My own tests with OpenLDAP showed Barcelona with a 66% lead, clock-for-clock, over Clovertown.
OpenLDAP Scaling
It's entirely possible that, averaged over some (unspecified) mix of workloads, the overall advantage was just 40% at the time they published those claims. While there are certainly a lot of workloads where the advantage is less, or nonexistent, there are probably also other workloads where the advantage is even higher than 66%.
You can criticize them all you want for the botched launch, and failing to deliver faster clock speeds, but I think there's plenty of evidence that they weren't making these numbers up. Given that (a few) folks are now showing Phenoms overclocking to 3.5GHz, I think it's also reasonable to believe that AMD actually expected to be able to deliver 3.0GHz parts to the market. This is pretty significant; it would probably take a 5GHz Harpertown system to match the performance of a 3GHz Barcelona for my workloads. I shudder to think of the performance-per-watt metrics for such a 5GHz server.
axel: "1. Asset Light and Asset Smart are the same thing, but AMD has recently chosen to use the latter in order to avoid the implication that it involves the sale of fab capacity."
If they really are the same thing, I wonder if the name change was designed to make them appear separate and give AMD more time without drawing even more ire from investors. I recall that one of the pointed questions directed at Hector was that they'd gone a year without much detail on "asset lite", and would they see this process repeated with "asset smart."
Hector's reply was along the lines of providing more details "soon", after which he clarified that "soon" meant anywhere from 90 days to 9 months. So it could be a case of not wanting to reveal anything yet, but finding a way to keep investors from becoming even more agitated than they must already be.
I think that it's not going to be anything revolutionary, probably something pretty mundane. But Ruiz would rather roll out the details when AMD is in better financial health and the investors have put the pitchforks down. If AMD hasn't made some progress on the financial front by year's end I don't know if he can delay revealing the details behind asset lite/smart any longer.
hyc
It's entirely possible that, averaged over some (unspecified) mix of workloads, the overall advantage was just 40% at the time they published those claims.
The quoted claims were that Barcelona would "be 40% faster than Clovertown across a wide variety of workloads" and "would blow away Clovertown in every dimension". Those statements are not consistent with any scenario in which Barcelona beats Clovertown in only three or four benchmarks like SPEC or OpenLDAP. Those statements clearly imply a general level of performance superiority in the server space and that's exactly why so many were so excited and looking forward to the K10 launch. And it's also why everyone was so let down. IMO, AMD had deliberately fostered an expectation in the public that they knew would not be fulfilled.
Now you're free to give AMD the benefit of the doubt and believe what you wish regarding their veracity & honesty, but I believe (as do many others) that AMD essentially lied and deliberately misled the public with those statements. They have lost all credibility with me and I no longer trust anything they say. Not to mention a couple other huge claims they backpeddled on:
- Promised that Quad FX support would continue through the K10 generation. Now it has been EOLed.
- Demonstrating K10 at 3.0 GHz and one AMD rep later even referred back to the demo as evidence of healthy K10 clocks.
"Given that (a few) folks are now showing Phenoms overclocking to 3.5GHz,I think it's also reasonable to believe that AMD actually expected to be able to deliver 3.0GHz parts to the market"
I'm not sure I follow the logic here - these overclocks I can only assume have Vcore jacked way up and may have some aggressive cooling(?). AMD's process appears to have a leakage problem which would mean even slight changes to Vcore would mean large changes in TDP (and thus difficulty in getting higher clocks without exploding the thermals).
Overclocks are difficult to use as an indication of future stock performance - I think even Scientia correctly points this out on many occasions. It shows the architecture has the capability but it is not necessarily indicative of future products as you have to consider process and thermal and binsplit implications.
If I carry your logic and then look at the Intel 45nm overclocks, what clocks would you reasonably assume for Penryn?
From the early leaked roadmaps (I think this may have been from HKPEC?), most of them had 2.6 in Q4'07 or Q1'08, and only a modest increase to 2.8GHz in Q2'08 - based on this I would have to assume AMD knew what they could do and even then couldn't deliver on it.
“even slight changes to Vcore would mean large changes in TDP (and thus difficulty in getting higher clocks without exploding the thermals).”
I’ve wondered about this dynamic for quite some time.
Let’s exclude the lower binned chips, if I’m using the correct terminology, for the sake of argument. Take for example QX9770 rated for 136W as we increase the Vcore and the clocks, does the TDP rise on a linear level, or is there a rise on some esoteric exponential progression?
Additionally, why does removing heat enable one to raise the clock rate substantially? Given that the leakage is part of the chips static design parameter, why does cooling or super cooling enable one to clock faster at higher voltages, aside from lowing resistance as a function of the chip operation? Does it lower leakage? Or am I confusing the two functions as interrelated, when they aren’t.
Obviously, circuit pathways in the chips design play a role in limiting frequency. I would like to exclude this as a variable, if possible.
SPARKS
Jack, recall above, my example of AMD’s market share loss sighting CRAY as customer who will be lost indefinitely, perhaps I was lucky in my timing and analysis. Why do I get the feeling Nehalem is further along than most suspect.
http://www.theinquirer.net/gb/inquirer/news
/2008/04/28/cray-goes-intel-hpc
This is big.
SPARKS
I'm not sure I follow the logic here - these overclocks I can only assume have Vcore jacked way up and may have some aggressive cooling(?). AMD's process appears to have a leakage problem which would mean even slight changes to Vcore would mean large changes in TDP (and thus difficulty in getting higher clocks without exploding the thermals).
Overclocks are difficult to use as an indication of future stock performance - I think even Scientia correctly points this out on many occasions. It shows the architecture has the capability but it is not necessarily indicative of future products as you have to consider process and thermal and binsplit implications.
If I carry your logic and then look at the Intel 45nm overclocks, what clocks would you reasonably assume for Penryn?
From the early leaked roadmaps (I think this may have been from HKPEC?), most of them had 2.6 in Q4'07 or Q1'08, and only a modest increase to 2.8GHz in Q2'08 - based on this I would have to assume AMD knew what they could do and even then couldn't deliver on it.
I can't argue with your fundamental point. But in this case, I don't think 1.3V is out of line. Water cooling is certainly not commonplace yet, but it's also not massively extreme. E.g., it's not something ridiculous like liquid nitrogen...
This result seems a bit over the top...
http://pinoypureview.blogspot.com/2007/10/penryn-overclocked-to-6ghz.html
I would certainly not assume from these results that products in the 5GHz range are coming out any time soon.
These seem a bit more comparable
http://www.xtremesystems.org/forums/showthread.php?t=142430
Sorry, forgot to paste the first link
http://www.xtremesystems.org/forums/showthread.php?t=185909
3.5GHz @1.3V, water cooling.
"Take for example QX9770 rated for 136W as we increase the Vcore and the clocks, does the TDP rise on a linear level, or is there a rise on some esoteric exponential progression?"
Power is typically exponentially related to clock speed (and I believe also Vcore - someone correct me if I'm wrong here)
Heat has all sorts of impacts - it will impact carrier mobility negatively and exponentially (which in turn lowers overall speed), for those with a semiconductor physics background it also increases minority carriers, which can hurt some of the parasitic capacitances, I think (it's been awhile since I studied this) and of course you touched upon general resistances vs temperature issues.
HYC - you ignore the forest and focus on the trees... do you really think OC's represent (with some factoring down) what you should expect for future stock?
Yes water is no longer 'extreme'... but for AMD to release these chips commercially we are talking stock air cooling AND low enough TDP's to fit their 'drop-in' to existing MOBO's in the field marketing campaign.
I don't see even a relatively mild Vcore increase like to 1.3V (or 1.33V) fit AMD's stock TDP bins. Just look at the Vcores on their current 135Watt bin parts.
And by the way the link you provide was 1.328V not 1.3V (unless you were referring to the Vcore for 3.0GHz). I also have no idea if this is typical performance, or an outlier, but quite frankly I have not seen that many reports of AMD chips at 3.5GHz at 1.33Volts
Finally I think it would not hold up with an AMD stock cooler. Let's face it, if it could - AMD would be selling them at 3.0GHz and helping their margins, instead of letting folks try to get there on their own with 'black editions'.
My point on the Intel OC's (ignoring the LN2 results) is that I wouldn't expect those overclocks to translate directly to future stock parts either (but if you assume the AMD results are indicative of future gains, then why wouldn't you expect similar gains based on Intel OC's). I think NEITHER OC (minus some correction factor) should be considered a forecast of future clocks - it simply is indicative of the headrooom an architecture potentially has, ultimately the process performance will dictate how close you can get to that point. And at this point, esp on 65nm, AMD has not demonstrated that they could get there - nor was it a reasonable expectation a year ago... and they're own leaked roadmaps were reflecting that,
My point on the Intel OC's (ignoring the LN2 results) is that I wouldn't expect those overclocks to translate directly to future stock parts either (but if you assume the AMD results are indicative of future gains, then why wouldn't you expect similar gains based on Intel OC's). I think NEITHER OC (minus some correction factor) should be considered a forecast of future clocks - it simply is indicative of the headrooom an architecture potentially has, ultimately the process performance will dictate how close you can get to that point. And at this point, esp on 65nm, AMD has not demonstrated that they could get there - nor was it a reasonable expectation a year ago... and they're own leaked roadmaps were reflecting that,
I understand your point, and I'm not disagreeing. I linked to the Q6600 results simply for context (which really was largely for my own benefit). Intel lists the Q6600 as a 2.4GHz part http://www.intel.com/products/processor_number/chart/core2quad.htm and overclockers appear to be getting to about 3.3GHz with it on water. So 3.5GHz with a 2.5GHz 9850 seems like rough parity.
It seems if Intel had expected their 65nm process to get to 3GHz or more they would have released such a product already. Instead, they shifted to 45nm, to produce the 3.2 GHz QX9770. Everything you're saying makes sense.
Also looking at the photo of the AMD 3GHz demo, there's a huge bank of fans being used. http://www.theinquirer.net/en/inquirer/news/2007/07/26/amd-shows-off-30ghz-barcelona
So OK, I agree on all your points.
"and overclockers appear to be getting to about 3.3GHz with it on water. So 3.5GHz with a 2.5GHz 9850 seems like rough parity."
You're kidding right? You pull up a rare good OC on 9850 (granted there isn't a lot of chips AMD has made, but @3.5 there is not enough data to suggest this is an 'expected' overclock).
And then you say 3.3 is a comparable OC on the 6600? with water cooling? on a small population chips? (or on a good majority?)
You are taking THE MOST FAVORABLE data point on the AMD chip and comparing it to a common data point on the Intel chip and then say they are comparable?
"it seems if Intel had expected their 65nm process to get to 3GHz or more they would have released such a product already"
I know from your past comments you have a bias toward AMD, but please stop with the FUD. You won't get away with the spin on this site - I would like to also note you are comparing the OC on Intel's BOTTOM bin with the OC on AMD's top (best) bin.
No 3GHz on Intel 65nm? How about (mix of dual and quad cores):
http://www.newegg.com/Product/Product.aspx?Item=N82E16819115026
http://www.newegg.com/Product/Product.aspx?Item=N82E16819115028
http://www.newegg.com/Product/Product.aspx?Item=N82E16819117134
http://www.newegg.com/Product/Product.aspx?Item=N82E16819117133
http://www.newegg.com/Product/Product.aspx?Item=N82E16819117092
I think you're changing the subject; we weren't talking about dual cores.
Your point I was addressing was - does a X% overclock give you a fair indication on what future models will do. Can you reasonably assume from anything from the fact that a 65nm 2.4GHz Q6600 commonly overclocks to 3.3GHz with water cooling? Can you reasonably assume from this fact alone, that 3GHz parts in the same family will arrive down the road? Your answer was No. But in hindsight, those 3GHz parts were indeed introduced.
As for fair comparisons: 3.5GHz for the 9850 may turn out to be exceedingly rare, we don't know yet. (Though there are at least two different people reporting results in that range on xtremesystems right now, so it's not a completely isolated case.) But given that 3.5GHz has been demonstrated, it's logical to assume that 3.2, 3.3GHz will be achieved in greater numbers. So yes, I think it's fair to say that 3.5GHz over 2.5GHz implies that 3.3GHz will be a more common OC, and that puts it in a range that's pretty comparable to what is commonly seen with the Q6600.
“And then you say 3.3 is a comparable OC on the 6600? with water cooling? on a small population chips? (or on a good majority?)”
Ah----no way. I have a Q6600 @ 3.38 1.4125 Vcore it does SuperPi 1M in 15 seconds. It PLAYS stable all day. I could be sandbagging here, the X48 chipset has no problem with a 1600 FSB. Not many folks have such a board.
Here is a link that did have the Phenom X4 9850 clocked to 3.5 but it was unstable, “but it preserved stability only at 3.2 GHz”
At 3.5 it ran Pi 1m in 22 sec.
My Q6600 came out of a retail box; I seriously doubt the AMD chip did. Also, if I recall Giant has achieved the same result with his Q6600. Thats 2 for 2.
Both chips are on air, the Phenom is on water.
Obviously, the thermals are lower, and the architecture is more efficient, clock for clock. But, we already knew this.
The point is yours.
http://xtreview.com/index.php
SPARKS
I was not intentionally FUDing. I just missed the significance of one of your statements. I apologize for not reading more carefully.
Q6600 is the bottom bin, for both speed and thermals. 9850 is the top for current speed, but more importantly, thermal limits. So while the 9850 may have headroom for frequency, in its current state it'd be unlikely that any faster chip would fit in the thermal limits, and therefore a 3GHz product is unlikely.
Is that what you were trying to say?
By the way fellas, if need to know anything more about the Q6600, let me know I try it out, right here on this keyboard.
The big fella's coming soon, the Q6600 is coming out and going back to wifes machine.
THEN we'll see what the top bin badboy can do. HOO YAA!
SPARKS
hyc... you twist and twist these arguments when you start to realize you are wrong... so let's take a step back to your original post:
"Given that (a few) folks are now showing Phenoms overclocking to 3.5GHz, I think it's also reasonable to believe that AMD actually expected to be able to deliver 3.0GHz parts to the market."
you are the one who made a 'resonable' claim, I was merely pointing out that:
A) This is a real bad conclusion (use of an OC to forecast stock speed parts or expectations thereof)
B) It is impossible to tell whether the 3.5GHz overclock is the norm or an outlier... given the 1.328V Vcore and use of water cooling, and a very limited data set of folks who have achieved this... I'm skeptical and even if this were 'normal' (which there is no evidence one way or the other) go back to point A
It appears as though you are trying to excuse AMD for their false claims by drawing conclusions on an incomplete data set. While that may be the norm at Scientia's blog, you will get challenged here.
So do you stand by your above claim? Given that AMD themselves did not have a 3.0GHz part on their roadmap through Q2'08 BEFORE ALL THE TLB AND INITIAL RELEASE OF K10, I'm not sure how you can make the assertion that it's reasonable that AMD expected to have 3GHz parts, especially, again, when it wasn't on their own roadmaps.
So now you can go on and on about trying to discredit minor points to distract people from your initial erroneous and/or unsubstantiated claims - this seems remarkably familiar to a recent string of posts about 64bit and is getting tiring.
"...Is that what you were trying to say?"
No what I was trying to say is taking an OC on a part and using that to predict what you should be able to do in the future is not reasonable and to use that to defend a chip manufacturer (who would certainly know better) is also wrong.
Then to compound that by comparing an OC on a top bin part vs the OC on a bottom bin part is also not a very good comparison (typically bottom bin parts catch most of the crap that falls out of the upper bins).
It's frustrating that given the lack of data you want to interpret everything in the best possible light... I generally go by the point of view that until something is proven (or at least substantiated by significant data) you should go by a more conservative point of view.
We both agree there is 'limited' (to put it favorably) data on a 3.5GHz overclock... given the limited data I would not be using it to make arguments and excuse AMD and use it to support AMD 'reasonably' thinking they could get to 3GHz on 65nm. (This is the logic issue... then of course there are the technical issues which I beat to death about using an OC to predict stock performance)
My Q6600 came out of a retail box; I seriously doubt the AMD chip did. Also, if I recall Giant has achieved the same result with his Q6600. Thats 2 for 2.
That's correct. Furthermore, that is a year old B3 stepping Q6600 on air cooling. With a G0 stepping and a great motherboard and ram like Sparks has 3.6GHz should be possible on air cooling. My 3.3GHz OC is 24/7 100% stable on a near two year old P5B Deluxe board. The same board is good for 4.5GHz with my E8400 (though the nForce 790i board in my main machine is better at OCing, the Q6600 just becomes thermally limited past 3.3GHz, so it wouldn't be wise to take it any further)
Comparing a once off OC on a Phenom that wasn't entirely stable using water cooling to an OC that is easily achievable on most Q6600s using a high end air cooling is just stupid.
The big fella's coming soon, the Q6600 is coming out and going back to wifes machine.
Now that's we've been waiting for! Can't wait to see some OCing results on the QX9770! 4GHz should be a walk in the park for that monster!
hyc... you twist and twist these arguments when you start to realize you are wrong... so let's take a step back to your original post:
"Given that (a few) folks are now showing Phenoms overclocking to 3.5GHz, I think it's also reasonable to believe that AMD actually expected to be able to deliver 3.0GHz parts to the market."
you are the one who made a 'resonable' claim, I was merely pointing out that:
A) This is a real bad conclusion (use of an OC to forecast stock speed parts or expectations thereof)
B) It is impossible to tell whether the 3.5GHz overclock is the norm or an outlier... given the 1.328V Vcore and use of water cooling, and a very limited data set of folks who have achieved this... I'm skeptical and even if this were 'normal' (which there is no evidence one way or the other) go back to point A
It appears as though you are trying to excuse AMD for their false claims by drawing conclusions on an incomplete data set. While that may be the norm at Scientia's blog, you will get challenged here.
So do you stand by your above claim?
I thought I already made it clear that I conceded on your points and that I agreed that you were right. What part of "I'm not disagreeing" did you not understand?
So now you can go on and on about trying to discredit minor points to distract people from your initial erroneous and/or unsubstantiated claims - this seems remarkably familiar to a recent string of posts about 64bit and is getting tiring.
I think you may find it's tiring because you don't know to stop attacking after you've already won the point.
Giant, it’s good to hear from you.
Ah, Q6600 the new overclocking darling of the Twenty First Century. I’m afraid to take the sweetie higher than 3.4, incredible! Actually GURU scared me a bit when he gave me a new concept to read, carrier mobility. Whoa! It was like a surgeon, who specializes in carcinogenic abnormalities, suggesting some ‘Light Reading!” I imagined thermal runaway times a couple of hundred million transistors, not a happy day. I backed it off a bit to your similar 3.3 clocks. (I’ll wait for the anti carcinogenic hafnium and some water treatment on the big boy.) I OC’d with the 8X and the 9X multiplier, it’s all the same, the thing just cranks.
What make the Q6600 so versatile you can breeze it up to gaming speeds for a gaming rig, or drop it in a media machine (with limited cooling) and let the four cores do all the work @ stock speeds for rendering, etc The P35 seems to be ideal for this application.
There is some FUD running around the net that Q6600 will be sold till the third quarter, as if it were a bad thing. (INTC 45nM shortages, I think, frankly, I think they got caught with their pants down; its demand) Whatever the reason, it will be a sad day when Q6600 gets EOL’d.. Actually, 65nM (GO) is running so well they could keep it in the lineup for another quarter. That said, I know, it’s back to business, yields, DPW, margins, and all that, but man, what a ride.
Q9300 is out and from what I’m reading it cranks just as well, if not better, with lower thermals. F**K’N A BUBBA!
Wait! Wait! Is that the UPS truck?!? Ah, false alarm.
Stay in touch, Giant, that other guys has got GURU’s nuts twisted again. This time GURU came out swinging, and it hasn’t been pretty.
I’ll keep ya posted.
SPARKS
Hey GURU! It just dawned on me! I personally am going to put this 'low bin' (old)Q6600, 'top bin' QX9770 thing to the ultimate test!
Same motherboard, powersupply, memory, bios, RAM, basically the whole f**king magilla, out of a retail box. No spin, just numbers on air, then with some di hydrogen oxide.
Now we have top bin vs. top bin!
HOO YAA!
Any takers on Super Pi 1M @ 11 second?
BTW: We've got to go over this carrier mobility thing. Christ, those formulas scrambled my brain, or what's laft of it. Looks like my little buddies (electrons) start running into each other and don't where to go, from heat, no less!
SPARKS
"Actually GURU scared me a bit when he gave me a new concept to read, carrier mobility"
I wouldn't worry so much about mobility - if you are OC'ing you may have to eventually worry about electomigration though!
Sweet dreams while you think about that...
“You've hit 4GHz very easily. Are you increasing the CPU multiplier, or the FSB to OC at this stage?”
The 4 Gig run was done strictly by a 10X multiplier, with memory set at the board natively assigned DDR3-1333 bios parameter. Incidentally, also listed in those options are, DDR3- 1600 *DDR3-1600 O.C.*, and *DDR3-1800 O.C.* I had to manually assign this parameters, but the board SAW the 1600 native.
Subsequently, I keyed in the DDR3-1600 native and checked the latency, it went down to 57ns. That’s well within reach IMC.
There is an interesting option I have, frankly, never seen before. The frequency multiple can be increased or decreased by .5. I always felt that a full multiple was too much of a jump; ASUS has addressed this issue quite nicely.
“BTW, have you picked up an equally impressive video card do go with this monster CPU?”
Unfortunately, no, I haven’t. I am still using the 1900XTX Crossfire set up which really isn’t bad. The scores I got with the setup along with the Q6600 were 11,490. With this chip the scores increased to 12,857, not too bad for 2 year setup. They’ve got some new things on the horizon in the interim. I really would like a substantial increase.
The ATI purchase really turned the graphics industry sideways.
My next purchase will be that “electric cooler” we spoke. GURU’s Electromigration, and carrier mobility abstracts have me pissing my pants. The next thing you know I’ll be wearing a dress and high heels.
With that in mind, that E8400 is absolutely beautiful, spectacular, in fact. I thought that Q6600 was something irreplaceable and unique. Man, was I all wet, it was only the beginning.
I’ll keep you posted as I develop a relationship with the new chip. Next stop, 1800 FSB, than to 4Gig and beyond.
SPARKS
Post a Comment