9.16.2007

Raising the Dead

There is a rumour floating around that AMD is planning to ship triple core K10’s. Quality and Reliability engineers at AMD are scratching their heads while Intel’s Yield engineers are probably rolling on the floor laughing at this obvious desperate attempt to salvage revenue from traditionally defective products. If indeed true, I can make a guess that this is a marketing idea because no sane product engineer would want to be associated with such travesty.

Some may already be calling this a clever move since according to some, a gap exist in the market between dual and quad core CPUs. And on top of that, Intel’s MCM approach prevents it from creating their own backward stepping 3-core CPU. Let me just say that is a silly assumption. First of all, what’s stopping Intel (besides good yields and common sense) from disabling one core from the 2nd chip to create a triple-core? Secondly, while there may be a gap between the numbers of cores, you won’t find any gap in pricing. There is no niche segment to take advantage.

There is a known relationship between yield and reliability which is why most semiconductor companies do not ship products from a bad production lot even if some dies do past the test. To illustrate the real concern, medical (i,e., heart pacers) and automotive companies are very strict when it comes to yield and how maverick lots are handled. While microprocessors built for PCs and Servers do not require similar tight controls, consumer level quality requirements aren't that far off. What is alarming is AMD's willingness to sacrifice quality over revenue.

When a CPU is designed to run at 2.6GHz and the output maxes out only at 2.0GHz, there is clearly something wrong with the process and the product's physical make up. The farther the products are from the design target the greater the uncertainty is with the reliability. Salvaging the CPU by disabling one of the cores does not remove the risk that the product would perform normally over time or is free from time dependent defects. No matter how anyone would want to spin this, AMD's compromise with quality says a lot about how poor their yields are and how desperate they have become.

Update: 18th Sept 2007
Intel's official response to AMD's tri-core:
Asked in a press conference following his Intel Developer Forum keynote how Intel might respond to AMD's recent announcement of a planned three-core processor for early 2008, Otellini offered a brief but savage response: "We see a distinct advantage in having all the cores on our die work."

78 comments:

SPARKS said...

This is why I asked on the T.O.D. post. If the chip could run on three, two, or one core(s), they wouldn’t have to dump them. I don’t know anything about chip production, how the IMC would handle the loss of a core, or a core running slowly. That said, it had to be designed with this feature early on. (They would be REAL idiots if they didn’t.) I’m no chip guy, not by a long shot, but I did have a hunch.

As far a pricing, niche, and position, I believe it wouldn’t be that bad. Dual cores and quad cores are so cheep now, most people would probably pick a 4 over three anyway. Look, a Q6600 ASP is currently $285. An E6600 is about $230. This would mean they would price these so called 3 cores below E6600, say $200 and the 4 core at $275. We are talking about an $80 spread here. This is not to mention they would be competing against the X2 line. But, hey, at least their not in the garbage bin! The biggest factor would be out and out performance. Who knows if they can get the ‘rejects’ to work AND beat the Intel chips categorically, in raw performance, they might cover severe losses!

This was on my mind last week, as it would be INSANE to throw out 3 good cores, and not sell them. Then, again, believing in this pre INTC IDF rumor, I could be a crazy as the guys who buy into the stuff AMD pumps.

Make you wonder how these things run single threaded, and if they can pull it off, don’t it?

SPARKS

Anonymous said...

"I don’t know anything about chip production, how the IMC would handle the loss of a core, or a core running slowly."

Aren't all of the cores crosslinked with HT? If so, in theory you should be able to diasble (fuse) the HT links from the fused core. Now who knows if the architecture could handle this, but it would seem to be possible.

Now I could see disabling a core when taking a dual core to single core (especially when they were the only things on the market). It just seems ridiculous to think there is a market between 2 and 4 cores. What SW runs optimally with 3 cores? Is there a price disparity between 2 and 4 cores (as robo wonders) to justify an intermediate SKU. Of course this is a company that has THEIR ENTIRE ATHLON X2 product line in what a $170 range with what 5,6,7 SKU's?

Intel needs to right a marketing for dummies book to help AMD out.

Also, keep in mind the INQ on technical matters is somewhat suspect. Forgetting the reverse HT, the INQ also said the 90nm process was "designed" for P4 (of course it seemed to work reasonably well for the mobile products). They also had a pretty poor explanation of high K/metal gate when they weren't cutting and pasting from Intel's foils. Don't confuse a site which is a collection of technical links with a site that is technical in nature.

SPARKS said...

Even Anton Shilov doesn't get it.


http://www.xbitlabs.com/news/cpu/
display/20070914212726.html

SPARKS

Anonymous said...

I'd wait to see the rumor confirmed first, maybe this is just something that they were investigating at one time and decided not to do. It sure sounds like it would add a whole new layer of products and pricing, though, and I'm not sure that's a good thing for anyone.

Meanwhile, Sharikou compares Intel to Nazis. I'm sure that Andy Grove will appreciate that comparison. That's just shameful.

pointer said...

There is a known relationship between yield and reliability which is why most semiconductor companies do not ship products from a bad production lot even if some dies do past the test.

True and further explained. Yield needs to be under certain threshold to trigger the alarm of reliability issue. When it is lower than the mean, but higher than threshold, it is safe statistically to consider the passing units are ok. Even when the yield is below threshold, the engineer can insert an extra procedure (burn in) to screen out the reliability issue. It is when the yield is far below the threshold, the whole lot will be scrapped.

Thus, it is possible for AMD to disable bad core and release it as a lesser core product 2P, 1P or even the speculated 3P, without much reliability concern issue. Of course, its CPU design must have a clean way (design) to seperate the unwanted core from the rest of component (power, internal buses) at the first place.

Anonymous said...

while Intel’s Yield engineers are probably rolling on the floor laughing at this obvious desperate attempt to salvage revenue from traditionally defective products.

Do you know that Intel first Core 2 Duo where INTEL should already have very high mature yields due to past Core Duo and Pentium 4/D 65nm processors experience, released BROKEN Core 2 Duo (almost all the Intel line) 6300 and 6400 processors?

The Xeons, E6600, E6700 and X6800 account to 5% of all Intel production leaving the other 95% as broken selling products? Or as you call them defective products!!!

Anonymous said...

If indeed true, I can make a guess that this is a marketing idea because no sane product engineer would want to be associated with such travesty.

The problem Doc is that Intel engineers are doing this since the Celeron days!

Anonymous said...

First of all, what’s stopping Intel (besides good yields and common sense) from disabling one core from the 2nd chip to create a triple-core?

The cache moron! The stupid FSB! And Intel stupidity also helps here!

Anonymous said...

Secondly, while there may be a gap between the numbers of cores, you won’t find any gap in pricing. There is no niche segment to take advantage.

The fact that the cheapest quad core cost 266€, where dual cost 70€?
There is no gap. Please Doc go to the real Doc!

Anonymous said...

There is a known relationship between yield and reliability which is why most semiconductor companies do not ship products from a bad production lot even if some dies do past the test.

Well Intel sells everything doc! Including VT disable cores and so on! Any thing is a penny, maybe that’s why Intel is profitable Doc?
Selling every broken CPUs from Celeron, Pentium E2xxx, E4xxx, E6xxx, …

Anonymous said...

What is alarming is AMD's willingness to sacrifice quality over revenue.

That’s what Intel is doing Doc. Is it bad Doc?
Maybe this explains the Intel market share loss Doc?
Castrated CPUs, castrated chipsets, already build castrated IGPs, and so on...

Anonymous said...

When a CPU is designed to run at 2.6GHz and the output maxes out only at 2.0GHz, there is clearly something wrong with the process and the product's physical make up.

When products are shown that can be run at 10Ghz but never clocked more than 3.8Ghz with issues there is clearly something wrong with the process and the product's physical make up.

Anonymous said...

Salvaging the CPU by disabling one of the cores does not remove the risk that the product would perform normally over time or is free from time dependent defects.

Maybe this explains the failing Pentium 3, Pentium 4 and Pentium D and including the exploding Intel laptops CPUs! From Intel any CPU can have thermal defects!!! Beware!

Anonymous said...

"Do you know that Intel first Core 2 Duo where INTEL should already have very high mature yields due to past Core Duo and Pentium 4/D 65nm processors experience, released BROKEN Core 2 Duo (almost all the Intel line) 6300 and 6400 processors?"

Nice misinformation - these were parts that had the cache fuse from 4MB down to 2MB. In fact many of the hcips actually had the full 4MB working...

Why do this? well Intel had one set of masks for 6300, 6400, 6600, 6700, 6800 which reduces cost and streamlines manufacturing (no need to switch reticles between lots which lowers the utilization of the litho tool which are typicially one of the constraints in the factory. They also only had to rel test and certify one product.

Core Duo = 65nm, I thought that was only a 90nm product? I must be wrong, you seem like such a knowledgeable fellow.

Anonymous said...

"When products are shown that can be run at 10Ghz but never clocked more than 3.8Ghz with issues there is clearly something wrong with the process and the product's physical make up."

Please provide a link where INTEL EVER SHOWED a 10GZ PRODUCT!

Anonymous said...

He is so knowledgable that he can't figure out how to put all of his quotes and responses in one post.

Anonymous said...

Folks ignore the various anonymous comments from Sharikook/ oneexpert/whatever new name sharikou name uses to post comment on his own blog.

It is rather obvious the technical and logical errors in the posts so no need to feed this troll! I guess he is having a hard time making up new articles on his own blog so he feels the need to expand.

Anonymous said...

"I can't recall the last time that there was so much controversy surrounding the launch of a new product nor can I recall the last time that testing was so poor and analysis so far off the mark."
(Our pal Demetia from his new blog)

Controversy? Where? Talk about trying to hype up the hype? The launch was MUTED, not much controversy to show because there was little to see. And maybe testing was so poor because AMD schemed this? (knowing there promises of 40% better at LAUNCH were ringing just a bit hollow) When you give review sites samples the Friday before a Monday launch (the ones that were 'special' enough to get them, what do you expect?

AMD could have also fixed this 'problem' by DOING (which they probably did) and RELEASING (which they DIDN'T) a full set of benchmarks themselves.

In short, BLAME AMD FOR THE POOR SET OF BENCHMARKS AT LAUNCH.

Forgive me, I use the word "launch" liberally - to Scientia this was not a paper launch because it was shipped to OEM's - as opposed to launch meaning you can buy (and get) actual products. When all you can buy is paper (or an IOU and you'll get it in November) at launch time it is a paper launch. Don't pull this crap about it shipped to OEM's. If the OEM's aren't ready, wait until they are as opposed to launching something that the buying public can't get. Feed the OEM's, have them agree not to sell before date X so all OEM's and channel have a somewhat level playing field and then hey, here's a concept, do the launch on the day (or near) the time people can buy them! I'm assuming Phenom will be a bit different?!?

This from a compnay that only does "hard launches". What a joke!

"With K10's base architectural changes confirmed it should just be a matter of time before the reviews catch up"

Well that is if chips ever stop "shipping" and actually make it out into the market...Don't blame the reviewers for getting all of 2 days or some of them (non-Tierr1) being forced to sign NDA's which requires AMD to see and 'review' the review prior to publication.

"This should also mean that for AMD it is primarily a matter of increasing clock speed above 2.5 Ghz to be fully competitive with Intel."

The problem is AMD needs a chip that is more then "competitive" and AMD is facing a moving target. EVEN IF a theoretical 2.5GHz chip is competitive with a 3.0GHz Intel chip(which remains to be seen), there will likely be 3.33GHz and eventually higher clocked ones.

I'm sure Scientia's response would be well AMD has 45nm coming around too... the only problem with that (twisted) logic, is what happened to K8 clock speeds when AMD shrunk it from 90nm to 65nm. And the 65nm to 45nm transition for AMD has even less a quoted performance by AMD (20%) than AMD claimed on 90nm to 65nm (40%). Factor in the CTI approach - 45nm will not be at max until at least a year after "launch" (again I use launch liberally as who knows how the hell AMD defines this other than we'll tell you when it is). AMD better hope the performance the get on 65nm K10 holds up well, because when they migrate it to 45nm, all they will get is likely power improvements for the first year of 45nm

Anonymous said...

Well it looks like Tom's Hardware Guide drank the triple core AMD coolaid.

One "author" there even thinks AMD can use this as an advantage if they price it competitively against Intel's 45nm dual core....

So he thinks AMD is going to price a 283mm2 die competitively against a 45nm dual core - which is what 120mm2?

Oh and in other news apparently clock speeds will be competitive too :) As we have seen from AMD's progress on quad cores to date.

It's scary, people get paid to write such excellent "analysis".

Giant said...

The fact that the cheapest quad core cost 266€, where dual cost 70€?

Yes. We all know that Intel sells one dual core product for $70 then jumps straight to quad core. They couldn't possibly have any other products priced over $70 but less than $266 could they?!

Oh wait. Stepping back to reality, we have a full line up of dual core CPUs. The dual core 1.6GHz Pentium E2140 at 1.6Ghz all the way through to the E6850 at $266. The dual cores finish at $266 and the quad cores start at that price.

For anyone that can use the power of a quad, that price is an utter bargain.

SPARKS said...

“When products are shown that can be run at 10Ghz but never clocked more than 3.8Ghz with issues there is clearly something wrong with the process and the product's physical make up.”

“Maybe this explains the failing Pentium 3, Pentium 4 and Pentium D and including the exploding Intel laptops CPUs! From Intel any CPU can have thermal defects!!! Beware!”


Fella’s, Fella’s, FELLAS! Come on, nobody ever said 10 GHz! Give it break. They did say 5 though! And, what’s up with the exploding laptops? Wasn’t that a major battery f**k up? The only chip I ever saw burn up was an AMD chip (on You Tube), they ran the poor bastard without a heat sink.

http://www.youtube.com/watch?v=wSALep8QZ84


First, let’s put the battery thing to rest. Yes, Exploding Laptop.COM will give you gory details.

http://explodinglaptop.com/ibm-thinkpad.php


Now, let’s go to the “never clocked more than 3.8Ghz with issues” thing. Whoa, easy wrangler! I can tell you with out any reservations WHAT SO EVER, the so called issues start EXACTLY at 4.40 GHz. These issues are not sole province of the chip alone. The motherboard make and type will have a great deal to do with O.C. potential of the last of the late, great, deep stage pipeline, Preslers. Yes, it can run hot, 68 degrees C with a Zalman 9700, in a warm room. Keep the room between 70 F and 75 F, and it will play all day and all night 4.26 GHz, not 3.8. Hey, if it starts to throttle under FEAR @ 1600x1200, max settings (80 F room), simply back off the 16X multiplier to 15X for a nice cool 4.0 GHz.

You will need 1.45 at the core, an ASUS P5WDG2-WS (8-Phase P.S., so smooth), a really good P.S., and some nice DDR2-1066 memory. Sure, the setup is slow by today’s standards but back in April ’06, it was great! The benchies are still up there with the big boys, (E6600’s).


Now, if you never owned a really great chip, Intel or AMD, for that mater you shouldn’t post such nonsense. Not only do you not know what you’re talking about, it is very obvious you are talking out of your ass. If you did, you would respect BOTH companies’ best efforts.

The chip is D955EE (3.46) (default 13X multiplier), and, it is running as I type at ‘the standard’ 4.26. It will be VERY difficult to eventually give this machine to my daughter, when Penryn arrives. Damn, I’ve never had so much fun with a chip. Like an old friend, it has its faults, but it’s always been a good ole buddy.

THERE IS NOTHING LIKE AN EXTREME EDITION CHIP, NOTHING.

SPARKS

Axel said...

Copied for posterity again. Scientia's latest article claims that K10 SSE performance is in fact far better than shown by Anandtech. But his sole foundation for this claim is a flawed Techware article in which a Barcelona test system outperforms Woodcrest 5150 in Linpack by some 49% per clock. The reviewers express surprise at K10's SSE performance. But what they missed is that they had unwittingly compared an eight-core Barcelona system against a four-core Woodcrest system!!

Scientia

I know you did not get your configuration from the article because it isn't listed.

What's so difficult about this?

1. The configuration for the Barcelona test system is listed. It's a PSSC A2400 2P server box with two Opteron 2347 Barcelona processors. That's a total of eight cores.

2. The Xeon 5150 is a dual core Woodcrest processor designed only for DP systems. The only Merom-based Xeon processors to support 4P are the Xeon 7200 and 7300 CPUs.

4. Therefore the Xeon 5150 system in the Techware test was a DP system, with a total of four cores.

5. Therefore Techware committed an obvious blunder and their credibility is called into question. It's telling that AMD would pick some unknown tech site to review the "most anticipated" product of 2007 and ignore established sites like Hot Hardware, Hexus, Tom's Hardware, etc. They wanted to keep the disappointing per-clock performance hush hush, and they knew that these enthusiast sites would emphasize per-clock over per-watt anyday.

6. Therefore the entire foundation for your latest blog entry is flawed, and hence so are any conclusions you made.

7. And hopefully you've also learned that you cannot predict a processor's performance on a vendor-produced software optimization guide, which seemed to be the entire basis this year for your incorrect predictions of K10 performance. Never trust the vendor, only unbiased third parties. Your big oversight was that K10 is a radically different CPU than Conroe that will only run at its best when code is optimized for it (particularly with the SSE code paths), and most benchmarks and applications will be optimized for Conroe. The fact is that for most benchmarks and applications, K10 generally does not greatly outperform K8 in SSE per clock due to the fact that this code is optimized for Intel processors. In addition, its performance per watt per mm2 die is also too little too late vs Penryn to save AMD from the major restructuring that is coming.

So why did you delete our comments and thereby make it impossible for other readers to follow this illuminating discussion? So here I've provided valuable links above for other intelligent readers to clearly follow the irrefutable logic for themselves, but you'll now go and delete them, leaving only your links instead that you've already admitted to be questionable. This has the effect of making it appear like you're censoring dissenting posts that you're unable to refute, hiding the truth if you will. It's poor form and only harms your image, because in the time that it takes you to delete the post, dozens of people have already read it and know exactly what you did. You erased the proof positive.

In order to preserve these words, I'm once again resorting to the rather tiresome tactic of copying this post onto Roborat's free speech haven.

Giant said...

eep the room between 70 F and 75 F, and it will play all day and all night 4.26 GHz, not 3.8.

The dual cores ran alot hotter than the single core Presler CPUs. A friend of mine is still running the Pentium 4 661 (3.6Ghz/65nm/2MB L2) and it clocks fine at 4.5GHz on the same high end Zalman 9700. It runs fine in the cooler weather but during summer the thermal throttling kicks in due to thermal issues.

It would have been interesting to see how high a Pentium 4 could have clocked on the 45nm/HK process. The HK solves the leakage issues that were preventing large increases to the clockspeed.

SPARKS said...

You know G, that question continuously haunts me. In fact, I was going to pose the same question/senario to that moron, but I didn't think ANY one thought about it, would understand, or even cared. Boy, was I wrong.

Imagine the 20/someodd stage pipe line, with a tweek here and there with order execution, high metel K, and 45 nM?

Alas, I quess we will never, know. In any event, I can't tell you how well this bad boy has served me. I'm gonna miss it, come late this year. ("sniff,sniff,sigh")

SPARKS

Giant said...

Alas, I quess we will never, know. In any event, I can't tell you how well this bad boy has served me. I'm gonna miss it, come late this year.

I know that feeling. I still think about my old dual Celeron 300mhz A/ABIT BP6 system from back in the good old days. Those CPUs would run at 450mhz fine all day long.

Just as we get ready for Yorkfield, Intel is going to show off Nehalem to wet our appetite!
Link

Anonymous said...

Hmmm... Nehalem must be late if they are going to show working samples a year before schedule! Now it seems like it is just a matter of how many steppings are needed - Core2 went through these iterations very quick. Culd be another June/July type of release (limited qty) ala Core2.

(BTW - Remember way back when Scientia said Nehalem not before 2009? And of late Q4'08? He may have to change the magic timeline showing how AMD "is closing the gap")

Anyone want to bet whether Intel does more than a "task manager" demo of the type AMD did back in Dec to "show off" Barcelona? I'm thinking they won't provide any hard #'s (on either clockspeed or benchmark-side) but will show some applications running?

So Barcelona @3GHz sometime maybe mid/late H1'08? Nehalem - Q3'08, maybe? Huh...things could get rough - forget about competing with Penryn...though I don't see Nehalem as a big a jump as Core2 was to P4, with the exception of 4P server space that will benefit from CSI/IMC the most, it still should be good.

Of course there are 2 critical questions:

1) How does Nehalem stand up on the all important "average power consumption" metric! :)
2) Can it be made into the all important triple core! (Unfortunately I see Intel following AMD down this hole...with the exception that Intel will likely be fusing WORKING cores)

Hmmmm... AMD tryng to steal thunder by claiming triple core (or more accurately a crippled quad core chip)...Intel showing Nehalem off...I bet the press will give the triple core as much airtime simply because AMD remains the lovable underdog, fighting the good fight, blah, blah...blah.

Anonymous said...

Why is a 3 core chip not "tri-core"?

2 cores - dual core (not "double core")
4 cores - quad core ("no quadruple core")

3 cores - tri core? (not "triple core")

Giant said...

I'm thinking they won't provide any hard #'s (on either clockspeed or benchmark-side) but will show some applications running?

That's probably what they'll do. It's far too early to show off performance or clockspeeds for Nehalem.

Anonymous said...

Scientia is a biased AMD nut. Bitch got nerves to call others biased and nutty ROFLMFAO.

Giant said...

CONFIRMED! NEHALEM IS RUNNING!

http://www.tgdaily.com/content/view/33913/118/

8 cores! 32nm wafers shown off as well! Intel's execution has been totally flawless lately. While AMD stumbles around with 2Ghz CPUs and boasts about three core CPUs Intel goes and does this.

AMD is done for.

Anonymous said...

CONFIRMED! NEHALEM IS RUNNING!

The IMC worked?
The CSI worked?
The NB worked?
The 2-way 4-way worked?

Seams the same shit AMD does but this is from Intel.
Late Q4/2008 beginning Q1/2009.

CONFIRMED! K10 IS RUNNING!

Anonymous said...

The executive also announced that Intel's 45 nm processors and 65 nm chipsets would use halogen-free packaging technology beginning in 2008 in order to make these products more environment-friendly.

Intel the pollution company!
Intel the cancer company!

Anonymous said...

"We expect our Penryn processors to provide up to a 20 percent performance increase while improving energy efficiency," Otellini said.

3.0Ghz 65nm = 3.6Gh 45nm

Intel already pre-fragged by AMD 3.2Ghz 65nm quad core.

Anonymous said...

Folks while Nehalem is running it still has a bit of a way to go...the fact that it can boot some OS's (but not Vista) is a huge plus, but keep in mind there is much more complexity on this then the standard debug the logic architecture. There's the IMC and CSI links which will probably mean more time between functional silicon and product launch then past Intel architectures. Look at AMD's recent troubles and they had been doing IMC/HT for a 2nd (or 3rd?) generation.

Still, Intel's execution has been impressive and process technology remains second to none (despite the 'hopeful' claims that AMD is closing the gap by some folks). Apparently Intel will be launching both desktop and server Penryn chips in mid-Nov now, which is a bit of a surprise as it seemed this would be server only... Will Penryn desktop be out before Phenom?

I think Intel should do an AMD "hard launch" and just do the announcement when they ship the parts to OEM's and then wash their hands of the lack of availability! (Just kidding). After all launch isn't supposed to mean you can actually buy the product is it?

Of course Scientia on Nov 13 will claim there are only limited quantities of Penryn chips available for purchase and claim a paper launch, and conveniently forget at launch there were plenty of Barcy chips you could theoretically purchase, you just couldn't actually receive them for some time...

Anonymous said...

Again - ignore the other anonymous troll!

When one's favorite company is floundering and you can't be positive about the company's most important product launch ever, the only thing you can do is try to bash the competitor.

Anonymous said...

"We see a distinct advantage in having all the cores on our die work."

Otellini needs to be careful - while Intel clearly has the upper hand now, you never know if things will change. Confidence is a good thing, arrogance however?

While not quite the same as AMD dismissing MCM and focusing on "native" (I hate that description) designs, you never know what the public will believe or accept. I'm not saying tri-cores will be the next great thing but who knows...

That said it was a pretty good barb - AMD PR did a decent job of getting a lot of press to spin this as a unique capability of AMD and a means of supplying solution to a potential market NEED, as opposed to a means of SALVAGING non-functional quad cores.

It is pretty funny though to still read some AMD fans around the net thinking that this is actually a "native" 3 core design and not a quad chip with 1 core disabled. Yeah, makes a lot of sense - they did a parallel development of a three core design and just decided to hide it until the last minute!

Giant said...


Intel already pre-fragged by AMD 3.2Ghz 65nm quad core.


Except that AMD only has processors at 2Ghz quad core you idiot.

Intel also actually delivers products... aahhhh... on time! That's an odd concept for AMD aint it? Quad core CPUs: Ten months behind Intel. DX10 graphics: Seven months behind Nvidia.

Pathetic.

Giant said...

When Otelini says 20% faster he is referring to the small clock speed increase the launch Penryn CPUs have as well the IPC boost for a speed increase of 20%. The current rumors say that by the time Nehalem is ready (second half '07) they'll have Nehalem up to about 4GHz.

SPARKS said...

"Otellini needs to be careful - while Intel clearly has the upper hand now, you never know if things will change. Confidence is a good thing, arrogance however?"

http://www.tgdaily.com/content/view
/33913/135/

Arrogance? This goes beyond that, what you are seeing here is flawless engineering and execution. The way Intel is running their business is a perfect example of diametric opposition, in contrast to the way AMD produced and executed. Maybe, you thought High K Dielectric was a horse and pony show. Perhaps, it was a Power Point, catch phase. It seems WITHOUT BRAGING, they can launch early with BOTH server AND desktop products in less than FOUR, (4), IV, F-O-U-R weeks! Really, if you don’t see this as a perfect example of model engineering and chip fabrication, then I really don’t know what the color of the sun is on your planet.

GOT THAT?

AMD’s bullshit spin drew so much attention during the last 9 months; people were fixated on the chip that eventually fell flat on its ass. Meanwhile, Intel was just banging out 45nM parts like mad. Rest assured that those parts are boxed and ready for release and shipment, bank on it. I did.

Maybe you see real confidence in Intel, and it is scaring the piss out of the completely subjective, bias, fanboy hopefuls. Don’t mistake this for arrogance, this dear boys,is an absolutely perfect business model of regaining market share, increasing performance, lowering prices, raising share prices, and executing cutting edge technology.

Let me define arrogance. Arrogance is when you take half your cash assets and put the company in debt for four times that amount with absolutely NO regard for the competition. You take shareholders assets, when you have NO successful WORKING future product releases anywhere in sight, and BLOW IT on a frivolous pipe dream.

Arrogance is when you take a good company that had a TOP product and squander peoples jobs, and all your resources, in one year, AND NOT EVEN FACTOR THAT YOU HAD ANOTHER POWERFUL COMPETITOR, waiting to eat you alive.

You dump your business partners, trash your loyal enthusiasts, and dump a successful viable platform for one that no one wanted; just so you proudly say you have a big time OEM as a partner. Then in all your arrogance, you find the deal you made isn’t making the money you hope for. In fact, the deal is strangling you AND your business partners.

What you are seeing here is NOT Intel’s arrogance; it is AMD’s arrogance coming to fruition.

This is gospel, and history.

SPARKS

Giant said...

... trash your loyal enthusiasts, and dump a successful viable platform for one that no one wanted;

I'll agree with that. AMD was hailing the Socket 939 platform as the platform of the future, with a lifespan of many years. The performance was good, much better than the Pentium D's that Intel had. So I built a 939 system with a 4200+ X2 CPU. Less than a year later AMD dumped the whole thing for AM2.

Anonymous said...

Sparks you completely misread my comment - if it turns out that tri-cores start selling (I'm not sure why that would be, but I don't think this is an impossible event), it's a safe bet that Intel would jump in as well. This has nothing to do with how well Intel is executing. While I think Otellini was trying to be funny and boast a bit about manufacturing capabilities, comments like these sometimes come back to haunt you.

Much like when AMD introduces the Shanghai MCM (or whatever the 8 core MCM is called) - the first question they should get asked is I though MCM was a bad solution, why aren't you doing this in a "superior" native design? Why no octa-core for dummies book?

Back to Otellini - when you clearly have the upper hand there is no need to rub it in... he should have just taken the high road and focused on Intel's excellent execution and product roadmap. Intel should try to get back to the days where they didn't even acknowledge AMD publicly and let their superior products speak for themselves.

Anonymous said...

>Back to Otellini - when you >clearly have the upper hand there >is no need to rub it in...

I think the 'AMD' question just got a bit tiresome for him ..

Anonymous said...

While I think Otellini was trying to be funny and boast a bit about manufacturing capabilities, comments like these sometimes come back to haunt you.
But I doubt they'd really have any impact. "Tri-cores" would need to become wildly popular for Intel to bother, IMO. Or they could always push lower-end quad-cores into that niche and make it seem as if they're a better deal.

Otellini may be hoping for a sort of "dual effect" from his comment: tweaking AMD's nose, and creating the suspicion that buying a three-core CPU means buying a "crippled" or "broken" CPU. Nothing like tossing a little bit of FUD out there to get the fanbois frothing at the mouth.

Axel said...

Anonymous

Or they could always push lower-end quad-cores into that niche and make it seem as if they're a better deal.

This is almost certainly what Intel will do in Q1 08 as 45-nm ramps. They will simply price Kentsfield farther into the mainstream and the faster Yorkfields at a premium. Yields on Kentsfield are mature, so Intel would have to generally disable a healthy core on each Kentsfield die to sell it as a cheaper tricore. Why do that when they can simply price that Kentsfield just a bit higher than AMD's tricores and deliver quadcore performance? The cost of the die is the same.

In Q1 08 with the launch of the desktop Penryns, it is likely that Intel will push Kentsfield down into the sub-$200 range. I see a price war looming between Phenom X3 and Kentsfield, with AMD being eventually forced to price these X3s into dual core territory. This will cannibalize their own dual core sales and drive AMD's fab output to a mix biased increasingly towards the huge 283 mm2 quadcore dies, while Intel blissfully keeps the price pressure on with fast MCM Yorkfields at 107 mm2 x 2.

AMD's restructuring is on its way sooner than most people realize...

pointer said...


The Xeons, E6600, E6700 and X6800 account to 5% of all Intel production leaving the other 95% as broken selling products? Or as you call them defective products!!!


There is a fine difference between the approaches. Intel will disable cache even if it is working for E6300/E6400 while waiting for the half cache design to be validated. This would help sell chip at a particular performance envelop, while protecting margin on higher performed CPUs.

For AMD case, it is more of salvaging than CPU performance positioning. Anyway, I don't think there is anything wrong with that. I wonder if AMD is doing this as after-thought, in such they need an extra stepping (since they only able to release this produc in Q1'08) to help isolate the bad core from the rest of component and reconfigure the share resource for the available cores.

pointer said...

Anonymous said...
First of all, what’s stopping Intel (besides good yields and common sense) from disabling one core from the 2nd chip to create a triple-core?

The cache moron! The stupid FSB! And Intel stupidity also helps here!


you seems to be pretty clueless on the Intel Core as well as the FSB. Intel is able to configure cache to be used fully by a single core, and there is no issue for FSB protocol to support 3 cores (instead of 4).

The only possible issue however, is whether Intel already has such design/fuse to diable a core completely from the rest (less complex than barcelona case here)

Giant said...

They could also package together a dual core die and a single core Celeron D 4xx die and create a tri-core CPU that way. But what's the point in this? Awesome quad core performance starts right now at $279 from Newegg.

Earlier I was encoding two videos, working with Photoshop and running Houdini all at the same time, the system was still totally responsive. This is all on a $266 product today. Coupled with that, 4GB of DDR2 memory, fast hard disks (Dual 500GB SATA2 16MB Cache, 7200RPM) and a Geforce 8800 GTS. All mainstream gear. Three years ago to achieve this level of multitasking you'd have needed a high end workstation costing around $5000 or more.

Giant said...

I'm post this here, since Scientia is on a censoring rampage once more.

Indeed. Scientia, why are you censoring posts here? First Axel's post with the Xeon 5150 vs. Opteron 2357 in LINPACK. There was nothing wrong with his post, it was pure fact.

Indeed, the only post I could see that you might have a valid reason for removing is mo's post that is full of profanity. But he wouldn't have posted that had you noted removed his prior post!

You claim to be unbiased, and not an AMD fanboy like Sharikou. But removing posts in such a manner leads people to seriously doubt these claims Scientia. I'll post his at Roborat's blog as well, since you'll in all likelihood delete this.

For people who do want to see the Nehalem demo in video, here is the link:

http://www.intel.com/pressroom/kits/events/idffall_2007/webcasts.htm

It's in the Pat Gelsinger webcast, towards the end.

Mo said...

damn scientia, I just went off on him in a post that will likey get deleted. I'd post it here but it's just full of profanity from me.

I'm just pissed off that this Greg character can make a FALSE claim and i provide evidence that he is wrong.
Greg said that They only showed a die shot of Nahelem. This is false, Intel loaded windows and OSX and ran 3D programs.
I post that and Scientia deletes my post and keeps Gregs.

WTF???????

Giant said...


I post that and Scientia deletes my post and keeps Gregs.


I know. Earlier he compared data of a Xeon 5150 and an Opteron 2347 in a LINPACK test. He said that the site that did the test provided no details on the hardware configuration.

But it was easy to work this out The Xeon 5150 is a 2.66Ghz dual core part, for DP servers. So a total of four 2.66Ghz cores. The Barcelona 2347 is a 1.9Ghz quad core for DP servers. So eight cores at 1.9Ghz vs. four cores at 2.66Ghz.

The Xeon system was at a huge disadvantage in the test, but Scientia does not admit that, and when Axel (and later me as well) pointed that out he just deleted the posts.

GutterRat said...

I just posted this on scientia's blog. He again deleted one or two of my posts, so cross posting here just in case.


Censoring my posts AGAIN, scientia?

AMD Principal MTS and "fellows" are leaving for NVDA, INTC, even IBM.

Despite their employee agreement that sequesters employees from poaching from AMD for 1 year, some are finding it too tempting not to leave.

No financial reward, uncertainty about future prospects...the perfect storm.

SPARKS said...

“Back to Otellini - when you clearly have the upper hand there is no need to rub it in... he should have just taken the high road and focused on Intel's excellent execution and product roadmap.”

He has been, and will continue to do so.

Look, I don’t mean to take you task for this, or anyone else, for that mater. However, let’s look at this from a 40 BILLION dollar a YEAR perspective. That’s $40,000,000,000. Further, let’s take AMD’s FUD, Power Point campaign over the last year, which in the final analysis, came to mere fizzle. Remember all those AMD exec’s pounding the pavement with Barcelona performance figures? “40% over Clovertown, they cried.” Intel did bite on a few spins and responded, true, but they sure as hell didn’t tap into 40 billion in resources to go “tit for tat” with AMD’s marketing spinsters! (I believe one server exec opened up and cried foul sometime in June or July.) Basically, that was it.

Here’s my point. When AMD had their chips on top last year, they had huge banners and ads on Park Ave, and they were spending TONS of money, bragging about the fastest chip in the world. It was perceived as cute and funny. “See the Scrappy Little Company is showing Intel a thing or two, ha, ha.” And that was fine and ACCEPTED, because it was true. AMD hired a sky writer to crash an Intel product launch, no criticism, all in fun. That was the perception. Call it the underdog license.

Now after the scrappy little company is getting its ass kicked downtown on Wall Street, at every turn for the last year, Ottelini passes one comment and gets called out on the carpet for it. “Arrogance, the shear gall,” they cried.

I’ll tell you what, I’d be willing to wager that Ottelini has had his hands full, keeping his aggressive (and well paid) marketing people on a short leash. If you think, by any stretch, these people have forgotten 2 years of AMD’s thumbing their noses in Intel face, think again. In fact, if I worked for Intel, in that position, with a product like Core 2 (and Quad), for a year, I WOULD WANT TO GET INDUSTRIAL SIZED MEAT GRINDERS.

Why do I mention the 40 billion? If he gave them one quarter of one percent of the year’s income, $100,000,000, and if he really wanted to be arrogant, all he’d need to do is write, on one miserable ‘Post-it’, “Go get ‘em boys”, and the gates of hell would open for the “Scrappy Little Company”

Wait, why spend the money? Read the ‘Wall Street Journal’, the gates of hell have already have opened. He KNOWS this, too. If you were him, do you think you’d be just a bit proud of your company, and your accomplishments? With that, on the whole, I think they’ve be extremely reserved, and so has he. And, there will be more to follow.

Obviously, modest, selfless, and low keyed, the man is a walking success story. Cut the guy some slack, will ya?

He is a crowning example of what every corporate CEO should aspire to, not the example, the model. A true leader, I for one would give anything to shake his hand.

SPARKS

Anonymous said...

"Obviously, modest, selfless, and low keyed, the man is a walking success story. Cut the guy some slack, will ya?"

How much of Intel's recent sucess do you think is directly attributable to Otellini? I'm not saying he's not a good CEO (heck in comparison to Ruiz he's a genius), but what has really turned Intel around? Products and process technology.

Otellini has ZERO influence over process technology - this is driven by folks like Mark Bohr and the extremely well run CR and TD divisions. And by the way the one area that was relatively untouched by the restructuring was the process TD - the organization and structure was kept exactly the same and there were very few layoffs in these organizations. On the product side he clearly has impact on marketing/pricing and strategy, but has little real influence over actual architectural decisions. Core2 was well under development before he took over.

O'm not saying he's not a good CEO, but like a baseball manager they tend to get too much blame when things go poorly and too much credit when things are going well. Otellini has done a good job of staying out of the way of things out of hie expertise and focusing on the business side of things.

My point is - what value does it add to even acknowledge AMD's existence at this point? You have a company that is executing, process and product technologies that are further ahead - publicly he should just ignore AMD (privately you obviously need to be aware). You have the press now singing the praises of the execution and roadmap and now you want them to even get the slightest hint that arrogance is setting in and start writing the what if Intel becomes a true monopoly articles or the inevitable will innovation slow down if AMD is not competitive article?

When things are going well, you should just shut up and continue on executing and modestly enjoy the benefits of your success. Taking a shot at your competitor serves no purpose. You don't seem to understand PR at all - when a small competitor takes shots and pulls stunts, it is usually considered "feisty" or "scrappy", when a large dominant company does the same thing (or even less) it will naturally be perceived as arrogant - that's just human nature and Otellini, with the background in sales and marketing that he has should understand this most of all.

I'm a huge Intel fan but I don't like to see things like this - while funny (and I laughed hard when I fist read that line), I realize there is always the potential for backlash.

Andy said...

With regard to the tricore argument....

Why have 3 when you can have 4 :P

Intel doesn't need to make tri-core cpus.

The chances of a defect hitting a single core only are so low it's probably easier for them to throw it in the bin. The cache Intels cache is more likely to get hit.

I think it is pretty easy to tell Intels yields are through the roof and probably have been since day one. Intels worry is balancing the cpu's they ship to supply the different types of consumer.
I doubt there are many Core2 CPU's that wouldn't hit the clock speed for an Xtreme Edition.

GutterRat said...

Another post censored by Scientia the tool.

Posting here :)

Scientia,

I would offer that the one painting himself into a corner by proclaiming G3MX and SSE5 as The two most significant developments of 2007 is you.

AMD only has slideware and paper docs to fling around whereas as evidenced by this week's IDF Intel's busy executing on its tick-tock product design strategy.

With respect to the various memory technologies in design and already out there: there is no one size fits all answer. It may very well be that FBDIMM2 is more appropriate for high-end SKUs than DDR2/3 with the associated power, density, bandwidth tradeoffs.

There's nothing remarkable about AMD's 'SSE5' and if we are to believe AMD the target is Bulldozer. AMD got caught flat footed with SSE4. Who will use these SSE5 instructions?

I am looking forward to a future disclosure of Larrabee and its ISA extensions. If it's an x86 core suitable for graphics, then I would imagine it would have some extensions to match of beat AMD's alleged 'SSE5'.

Please continue with your humorous posts.

SPARKS said...

“You don't seem to understand PR at all –“

Man, you’ve got that right. Nasty business as far I’m concerned. In my line of work I don’t factor perceptions at all. I could care less. Are we going to bring the job in on time, is the boss making money, and are the customers happy.

Performance is king. In my line of work there is no spinning and embellishment. Either the damned thing works, or someone’s unhappy. I’ll deal with the problems, I’ll make it happen, and just stay the hell out of my way. Yes, this is true I’m the company prima Donna, and pain in the ass. After the smell of B.S. clears the room if they what an honest assessment, I’ll give it to them with the gloves off.

Karl Marx said, religion is the opiate if the masses. Not quite. What he said was,

"Religion in any shape or form is regarded as pernicious and deliberate falsehood, spread and encouraged by rulers and clerics in their own interests, since it is easier to control over the ignorant”

Well, just substitute the word ‘religion’ with 'Public Relations', and you have AMD during the last year. Now, it is three core, tri-core, hard core, who give a ****, its all nonsense.

However, I will concede the argument; Otellini could have been more tactful, especially now with Intel’s, size, superior resources, stellar performance and execution. The point is yours.

It just felt so good to hear him say it.

SPARKS

Intel Fanboi said...

Yeah, I posted this over at Sharikou's, but what the heck:

AMD marketing is currently testing potential brand naming for the tri-core. Here is a leaked list of what they are considering:

Oopseron

Subprimeron

Tritanic

Quasinodo

Quadconomo

Brokealona

StrikeThree

Oddcore

Refurbium

Unobtainium

Faileron

Bankruptelona

Tri-plegic

TriedandFailed

QuadCoNoMo

Subopteron

Rejecteron

Phuckitron

Defecteron

Tri-Hard

Dual Core Plus

Quad Core Minus

3/4 Quad

Tri-De-Fecta

Tri-to-sell

Lemonade

Triple Cripple

Shortbus

Giant said...

Scientia is WORSE than Sharikou. He's enabled full comment moderation. Now the only comments that will show are one's that agree with his articles. Certainly, no one correcting his false statements would be permitted to post.

All this while he claims to be unbiased! That is just so low and pathetic that it's stupid.

pointer said...

Giant said...

Scientia is WORSE than Sharikou. He's enabled full comment moderation. Now the only comments that will show are one's that agree with his articles. Certainly, no one correcting his false statements would be permitted to post.

All this while he claims to be unbiased! That is just so low and pathetic that it's stupid.


I actually wondering when will you all learn. I stop posting there once he start name labelling me out of no reason (may be there is a reason, I disagree with his statement:)). At the same time, i observed he deletes post that disagreeing with him as well.

Anonymous said...

"It just felt so good to hear him say it."

I agree...as for PR, I'm an engineer by trade but it is unsettling how decisions are not always made upon the best technical solution. Especially as companies get bigger, the BS politics and perception become difficult.

One of the better bosses I've had took me aside one time and said - "sometimes perception is reality, while you may be technically right that doesn't mean people will agree with you" It kind of pissed me off until I realized he unfortunately was right.

Anonymous said...

Intel Fanboi:

Unobtainium is unfortunately taken. In many an engineering meeting when various materials are being tested unsucessfully, the question is often asked if anyone has tested unobtanium?

I do have some alternative suggestions:

- limpalon
- bankruptium
- lowyieldalon
- scrapium

and my personal fave:
- defectron

InTheKnow said...

Gutterrat, it seems you have singlehandedly shut down Scientia's blog. I'm interested to see if my contrarian view will make it through moderation.

He pulled my post that gave him raw production numbers and closed the AMD vs Intel output topic for discussion. Seeing my current submission get posted will go a long ways towards determining if I bother to post there again.

Frankly, I value his viewpoint since it is different from my own. But if an open discussion of all aspects of his post isn't permitted, then it is just a soapbox and I might as well just read AMD press releases.

Anonymous said...

Funny in Scientia's Top developments of 2007, K10 is not listed! Who knows maybe he is coming around!

Also not sure why he listed SSE5 as a top development as there is not a single product that exists today (or in 2007) that uses it. Apparently anything announced ON PAPER can be a top development.

I also love how his blogs just continue to stray off point in just a few paragraphs which is why his blogs are 50 pages long. He moves from top developments of 2007 to Intel "hitting the wall" (AGAIN, gee, hadn't heard this before!) Funny no mention of AMD's K10 wall at 2.5GHz (maybe) in 2007!

The blogs seem to have gotten more and more one sided as things have gotten worse for AMD.

Apparently he missed a few (off the top of my head):
1) Intel demo-ing a chip which has IMC and CSI (Nehalem). All this does is eliminate the one last spot where AMD had a competitive edge - but hey that's not important.

2) Releasing 45nm products in less than 2 years after 65nm (Jan'06 --> Nov07), also not important. This includes replacement of the gate oxide material which has been used for 20-30 years with a new material, also not important.

3)Shipment of well over 1,000,000 quad core chips prior to a paper quad core launch by their competitor.

I still find it amazing that Scientia doesn't even mention K10 as a top development in 2007! One would think the "most important launch of 2007" (as stated by AMD) would make Scientia's well thought out, carefully researched and completelty unbiased blog!

Anonymous said...

"Frankly, I value his viewpoint since it is different from my own."

Unfortunately, this value is not reciprocated on Scientia's blog as he thinks anything that questions or challenges the validity of his point is often taken as a personal attack or just dismissed without consideration.

I was following the whole how many cores were used in the benchmarks he posted on the K10 vs Core 2 and found it amazing that the person who challenged it was shutdown because he couldn't definitely prove it. What was amazing was Scientia couldn't prove his own view either so in the absence of data he just assumed his point of view regarding the core count was correct.

Part of me hopes things turn around so Scientia doesn't completely lose it and go (further) off the deep end.

Giant said...

bankruptium - That's AMD. Intel is still raking in over $5bn in profit/year.

lowyieldalon - That's AMD too. AMD's yields are at 30% as we've seen. That's why they're selling three core CPUs next year!

scrapium - This describes AMD too. All AMD dies are scrap, so they'll be used for making unreliable three core CPUs.

defectron - AMD is the king of this too. They love defective CPUs - why else would they produce a three core CPU out of dies that were intended for FOUR core CPUs?

limpalon - AMD enjoys limping along too! Losing a cool $7M/day.

Gutterrat, it seems you have singlehandedly shut down Scientia's blog. I'm interested to see if my contrarian view will make it through moderation.

That's doubtful. I sent a post through refuting his claims that Intel is stuck at 3Ghz even with 45nm. (They're releasing a 3.16 / 1333 and 3.2Ghz / 1600 version come November... what is Scientia smoking?!)

It's obvious that Intel is using all the good high performance dies for the 3GHz+ high end server processors since the 45nm production is taking place solely at D1D for now.

Intel has zero competition in the high end desktop segment now, so they don't need more than 3Ghz. Best to use the high-speed dies for server CPUs to keep Barcelona at bay.

Giant said...

Oh, $10 says that neither my post nor your post go through. ;-0

Anonymous said...

err....giant....those were all suggested names for Intel fanboi to add to his tri-core naming list (did you not realize this as you went thru each name and realized they all described AMD?).

For a REAL FUNNY TAKE ON TRI-CORE:
http://www.overclockers.com/tips01218/

'What if Dell and HP put out tri-core product lines? What are you going to tell them if they want more, "Sorry, we didn't screw up enough, we promise to do worse next week?"'

'Disable a healthy core Incur all the expenses of making a four-core, then deliberately sabotage it so you can sell it for less. This is just what a company staring bankruptcy in the face needs to do.'

The overall analysis is actually pretty dead-on and pretty damn funnny/sarcastic... the site may be a little pro-Intel, but if the points are valid they're valid...

Anonymous said...

Now that I think about it, tricore is sounding more and more like one of those niche products which AMD touts the features and/or capabilities of, and yet is mysteriously nowhere to be found when you actually want to buy one. (remember those energy efficient Athlons that were benchmarked, but were more than a bit difficult to find?)

So my prediction - they'll be some tri-core kicking around early on for the desktop market as AMD works on its yield of quad core. AMD will claim victory and the ability to "uniquely" service market demand. If/when AMD improves yield, these chips will magically vanish in terms of availability as AMD realizes, surprisingly enough, they can charge more for a quad then disabling one of the cores and selling it as a tri-core.

The AMD spin will be: "as the market has shifted toward quad core, we no longer see market demand for tri-core therefore we will stop 'making' them." Hey, it's AMD and they're all about satisfying market demand - look no further than K10 launch which "focused" on low speed chips because that's "what customers were demanding"!

And if you think I'm crazy look back a few blogs ago and look for my prediction of AMD to roll out a new benchmark at the K10 launch! (average power consumption anyone?)

The only question is will Intel to be stupid enough to react and 'make' their own tri-core or put one on the roadmap, I really don't hope they take the bait on this one - it seems like they have layed (laid?) off this for now.

GutterRat said...

intheknow and giant

I have now gotten into the habit of cross-posting onto roborat's (thanks, REALLY) blog because Scientia has demonstrated he'll censor posts that don't meet his "editorial bar" which is subjective.

Scientia does not like to be challenged with hard data.

Scientia is an AMDZone regular, as well as abistein.

They say you can't take the stripes away from a tiger, right?

GutterRat said...

Now that I think about it, tricore is sounding more and more like one of those niche products which AMD touts the features and/or capabilities of, and yet is mysteriously nowhere to be found when you actually want to buy one.

I'm just waiting for Scientia's spin on this whole "tri-core" development.

How this is "customer centric innovation" is beyond me. Who are the EXACT customers demanding this feature?

Wake me up when you find some.

Giant said...



err....giant....those were all suggested names for Intel fanboi to add to his tri-core naming list (did you not realize this as you went thru each name and realized they all described AMD?)



LOL! Sorry! I should have read your post again. I thought you were the pro-AMD guy that posts as Anonymous.

Ho Ho said...

Interesting that Scienta completely ignored my latest post about SIMD where I explained his mistakes. I reminded it to him but unfortunately I wrote it after he turned on full moderation and I doubt it would get through.

Anonymous said...

Non native quad core beats naive quad core hells yeah

Giant said...

Skulltrail platform early details:

http://blogs.zdnet.com/Ou/images/skulltrail-benchmarks-leaked.jpg

Dual 3.4Ghz Quad Core CPUs with dual 1600mhz FSB and 12MB L2 cache per CPU anyone?

(Where's Scientia with his Intel can't get past 3ghz claim now?!)

I imagine that would be one hell of a rig. It would certainly put my computer to shame!

Anonymous said...

A good interview with Mooly Eden:
http://www.tgdaily.com/content/view/33958/136/

Many will think he's arrogant (I think it is more a style thing than arrogance), but what separates him from most others is that he really understands BOTH the technology and the market. I think the last paragraph on the last comment says it all.

Anonymous said...

Scientia, needs to lay off the drugs:

"I don't see a monolithic 8 core Nehalem as practical. Such a chip could easily be 520 sq mm."

err... how big is a quad Nehalem? Apparently ~260mm2? Nope! Might want to do a little research on Nehalem quad core die size before pulling a number out of the air!

So we understand a quad core chip on 65nm for AMD is practical, a octa core chip on 45nm is not? Hey, if Intel has problems with yield they could just launch "native" septa-cores or penta-cores...and then say it's because the market is demanding them!

"Yes, Newegg (and others) have Barcelona so that should remove the notion that it was a paper launch.

I'm not sure what to say about the $790 price tag though. I doubt it will stay that high for long."

Ahhh... product 8 days after launch - why not launch when the product is actually available?

As for the price tag, it will stay that way for as long as there are enough suckers, ummm I mean AMD fans, willling to buy an 2.0GHz chip when they'll be far faster ones available shortly.

"Given the above I can easily see Shanghai slipping and then coming out once again in the midstream and low voltage sectors rather than the high end, as per the 65nm process for K8."

Welcome to CTI - the initial rev of AMD's new process technology is usually about same performance as the previous tech (except with worse yield). This is the great shell game that feeds (and fools) the AMD fans into thinking hey, we're only a year behind. The problem is that it takes AMD another year to get the tech to actual performance while Intel's process is there at launch (thus a 2 year gap minimum!)

"but I can see Intels release of Nehalem slipping as well, possibly into Q4."

Funny Scientia's original estimate was 2009! His revised estimate, less than 1 month ago (aug 23 blog) was Q4'08. Now it might "slip" to Q4'08! Intel has said H2'08 so much like AMD's mid-year crap (which apparently means end of Sept!) this would not be a "slip", even if it did launch in Q4'08.

Ahhh... Scientia getting the spin machine into overdrive - I'm surprising he's not saying Nehalem should be here in Q2'08 so if it comes in Q3 or Q4 he can claim it to be 1/2 year behind! I guess if you make enough predictions, one of them is bound to come true or validate whatever future spin you might want to make. (The other key ingredient is to make your comments so long that people can't find the original prediction)

Anonymous said...

I'm not sure what to say about the $790 price tag though.

The high price tag usually means that there is a shortage of supply and vendors are pricing the product accordingly.

Anonymous said...

"I'm not sure what to say about the $790 price tag though."

It says Newegg realizes the only idiots (umm... I mean customers) who will be buying the first rev of these chips are loyalists who will pay any price - might as well gouge as much out of them as you can when you get the chance (it could be the last one). Newegg (not AMD) is setting this price.