AMD ready to crush Intel in a price war
Chinese independently discovered that Con XE 6800 2.93GHZ leads AMD64 FX62 2.8GHZ by less than 9% with 4-4-4-12 RAM.
In other news, IBM is going Opteron in a bigger way.
Intel and AMD's pricing after July 24 will look like the following:
Since an Athlon 64 X2 3800+ frags a Pentium D XE 965, the pricing table above simply makes all Netburst CPUs 100% undesirable from all aspects. You should note that 90% of Intel's units in 3Q06 are Pentium D, Pentium 4 and Core32.
The fact that only AMD's FX CPU has 2x1MB cache indicates that AMD's yield is near 100%. Basically, AMD puts in a wafer and says let them be FX62s, and they all become FX62s. With APM3.0, AMD is super efficient. I expect the X2 4200 to become the top selling CPU.
posted by Sharikou, Ph. D. @ 7/20/2006 06:21:00 PM
40 Comments:
Tnx sharikou, here inblog I finde all teh infrmation i need, it changes my upgrade plans. 9% witn new 2993 mghz and 3 Y O arqitecture with 2800mhhc is othing. I guess ew chips from Amd will b estronger,
This is pretty weak. The reviews came out on the 14th, which means you have spent six days reserching reviews in order to find 1 that has AMD within 9%.
I like your blog, I like the fact that there are opinions from both sides here, but this is one of the weakest post from you that I have ever seen.
Keep blogging.
Torrenza!
IBM doesn't want to be left behind.
News Brief:
The Intel fanboy talked off the ledge today is resting fine and is receiving psychiatric care. He keeps mumbling “only 9%”.
Is it just me or if you look at the rest of the benchmarks (page 5,7) some are not "just 9%".
I know this is just Winrar, Lame, and Windows office benchmarking which are just SW benchmarkiing programs that noone uses in real life but I thought it was worth mentioning...
Can you say wow..
After all that searching and all they find is to be behind by 9%..
I'd think they would have found one benchmark AMD could win.
Things are pretty bleak there
The Doctor
Entrophos the only problem with that is the fact that FX series are going to stay 90nm until, little help here, I think middle of next year, I cant remember exactly.
Actually, I found the article... here.
When you read it youll also notice FX66 isn't until Q3 2007.
Given the pre-existing ties between AMD and IBM I'm surprised that IBM hasn't jumped on the bandwagon sooner.
How'd you calculate your scores. I get a 19% advantage, and this includes the GPU limited benchmarks with 3DMark, Hl2 and Doom 3. It doesn't include the memory bandwidth tests or any of the Sisoft benchmarks.
Given how easily Conroe wins the CPU-intensive benchmarks, it'll probably take a 4GHz FX to match it.
Since you probably don’t understand Chinese, you probably got the wrong impression. I will summarize that article for you. Conroe WINS! The reviewers LOVED Conroe! The best that FX-62 can do is losing 9% or more.
My opinion on any processor is that how easy to overclock a cpu to a certain speed with just stock cooler is how ready that manufacturer be selling that speed on the market. 3Ghz on AMD, even FX-62, with stock fan is not really easy. However, from all the reviews, 3Ghz+ on Conroe is quite common.
I don’t care who makes the processor. To me, a good processor is a good processor. You can say Intel is bad as a company. But denying a good product because of that is just stupid.
The fact that only AMD's FX CPU has 2x1MB cache indicates that AMD's yield is near 100%. Basically, AMD puts in a wafer and says let them be FX62s, and they all become FX62s. With APM3.0, AMD is super efficient. I expect the X2 4200 to become the top selling CPU.
Actually, it means that AMD can't afford to manufacture large numbers of 230mm^2 CPUs at the new pricing levels dicated by Conroes dominance.
The 1MB FXs that don't meet the clock requirements now become Opterons.
"I know it's a marketing tactic, but it would be a smart one at this point."
It would smart. They need to have something out there that can compete with the high end Conroe. I feel they will loose fanbase if they don't show that they are still competitive (not talking about 6 months from now). I also believe alot of people have held back on upgrading especially after AM2 only brought a 5-10% improvement. Those are the people they might lose.
Overclockers has their own opinion on AMD's numbers.
http://www.overclockers.com/tips00996/
"You should note that 90% of Intel's units in 3Q06 are Pentium D and Pentium 4s."
You mean deshtop right? Mobile continues to be faster growing then desktop these days and mobile units are not P4's.
FX62 being the only product at 1MB does not mean perfect yield; what it means is that any die that have cache hits get re-badged as a 512MB X2 5000....but you already must have known that based on your vast Si processing background, right?
Or it could mean that the majority of the yield hits are on the logic portion of the die as opposed to cache so it doesn't matter what the cache size is from a yield perspective.
You mean deshtop right? Mobile continues to be faster growing then desktop these days and mobile units are not P4's.
No. Actually, 90% of Intel's units are legacy stuff, server, desktop and mobile (Pentium-M, Core32)
Let me repeat your ORIGINAL quote since you can't apparently read your own words or modify them after the fact so that noone thinks you made a mistake:
"You should note that 90% of Intel's units in 3Q06 are Pentium D and Pentium 4s"
What part of this statement mentions P-M and/or Core? Perhaps you meant to say P4 and "others", but that is not what you stated. I don't believe Intel is changing pricing on mobile chips yet; so price war represents <90% of Intel's units in Q3.
It is good however that you update your blog after the fact and then state I'm wrong...you're new statement/change is actually noticeable because your table makes no mention of Core pricing and your previous sentence refers only to "netburst" which the Core is not based on. Good to see integrity!
Lol. What you don't get is AMD isn't making any money on those chips, Intel is.
If AMD wasn't stuck on primitive 90nm they might be in a slightly better position.
Intel is going to be punished in the next quarter or two, but AMD is going to take it even worse.
I'm not an idiot like some people so I won't claim AMD is going "BK", but they'll suffer.
Expect AMD stock to go below $15, INTC stock to hover around $17-19.
I'll come back and feast on your tears when AMD is in the crapper.
After attending a high level meeting of 14 high end gaming computer companies - you would recognize them, but I won't mention their names, but they are the leaders of the industry. Every single one of them is going Conroe big time. AMD was a speaker and said that their first quad cores would be Opterons. AMD is ahead in servers and Intel won't catch up anytime soon while they're now behind on desktop and seems to have sold out the gamers who brought them where they are today. Being the only primarily server vendor in this meeting, I pointed this out to them and told them that we didn't need quad core Opteron to be competitive with Woodcrest, which is a nice part - we have had access to them and they run well, but not as well as Opteron for the overall price. AMD would do much better by skipping this 4 x 4 nonsense and coming out with a true quadcore part because these high end gamer types, although they are a small market on their own, are heavy influencers. Opteron is fine and will be able to hold its own for a year...but AMD has sold out their core spokesmen and Intel will certainly try to hold them forever.
9% performance lead? -In games that is, least we forget Sharikou.
I'm concerned about 3D modelling, video dekoding/encoding and decompressing/compressing performance. That's where the Core architecture shines. I don't play games, so your 9% comment is worthless concerning my use.
It doesn't surprise me if you still remove comments like this, but I've got a point if you don't.
FX62 being the only product at 1MB does not mean perfect yield; what it means is that any die that have cache hits get re-badged as a 512MB X2 5000....
AMD as different cores for 1MB and 512KB parts, what you say makes no sense.
While Intel from the same part Conroe 4MB for example:
-Dual core 4MB part
-Dual core 2MB part
-Single core 2MB part
-...*
*maybe more
AMD can not do that because design limitation (did you ever saw one X2 being re branded as 3000?), or if it can they don’t do it because the yields are at 90% or more.
How do you explain that AMD with 1 factory does 20% of the market and Intel requires 5 or 6?
Overclockers has their own opinion on AMD's numbers.
http://www.overclockers.com/tips00996/
Don't you have any new material, I've only seen this post like 100 times. All it is saying is that AMD's average chip price is going to go up because they're switching to an all dual core lineup. What you - and the artice - fail to mention is AMD makes each core cheaper than Intel makes a core. Intel is switching to dual core also. Intel loses bigger in this scenario. In fact their setting themselves up for a lot of trouble due to the fact that they decided to start a price war that favors a company who can make a cpu more effeciently. Guess what? Intel can't make a cpu more effeciently than AMD. INTEL LOSES!
I'm not a big fan of these price wars (as I invest much more money into the tech sector than I make up for in "consumer savings"). As of late I've seen Intel's stock drop %35 and AMD's get nearly cut in half. This is not a good thing, for either camp.
But my one hope in all of this is that the price war wont last long. It can't. AMD simply enough cannot sustain a long drawn out price war - especially if they have any desire to increase capacity by building new fabs. Intel also does not want to continue this price war, as they recognize the unrest in their investor base. It seems the only reason Intel initiated this to begin with is a "this will hurt you more than it hurts me" mentality.
I'm not going to argue about performance. Every tech oriented group on the planet has claimed Intel Core superiority (even Sharikou, although he tends to down play the performance gap compared to mainstream sites). What I will say though is that this price war is bad for both companies. Both companies will come out with lower margins and will take a major hit in terms of profitability.
These are the conclusions that people arrive at when they lack bias.
quote from http://www.informationweek.com/hardware/showArticle.jhtml?articleID=190900525&subSection=Processors
In some cases, executives said, AMD walked away from business when price points became so low the deal was deemed unprofitable.
Henri Richard, AMD's executive vice president of worldwide sales and marketing, said AMD would only take business that makes sense for the company. "We are not going to chase what I call lighting a cigarette in front of a gas leak," he said.
CPU prices from below
http://www.hkepc.com/bbs/itnews.php?tid=633569
http://www.hkepc.com/bbs/itnews.php?tid=632181&starttime=0&endtime=0
Minimum AMD CPU price is USD51
Minimum Intel CPU price is USD39
please allow me to use my limited logic analysis here:
1) AMD is a GOOD company, and will do good thing for humanity
2) Selling CPUs is to gain money, no matter how little, is true for both Intel and AMD
3) Hector is good man and won't lie
assume everyone want to make at least 10% for profit but AMD can't push it below USD 51. so, i'll assume its low end CPU cost is USD 46, and intel low end CPU cost is USD 35.
should AMD make more than 10% on it and so that can prove Shariko's point AMD's APM is far more superior than Intel's Copy exactly? it can't without violating point number 1 and 3. Since AMD is so good and supportive to the humanity, of course AMD would support USD100 PC initiative. The key is low CPU cost. A GOOD AMD would definitely sell cheaper CPU when it can and a good Hector would not lie.
So, can intel actually selling at lost? it can't either since Sharikou think that intel is so evil and thus it would simply make its CPU to support the PC USD 100 initiative to sell it at lost.
So, the conclusion is what the industry has recognised except for Sharikou, his fren mike and the AMD marketing VP, Intel has far more superior manufacturing :)
btw, AMD's 'matured yield' 7 years ago is 60-70% :) i'm not sure what is the number now.
Sharikou, I wonder what you think of the Inq's latest story that AMD and ATI are rumoured to be going to their shareholders to get approval for a merger?
Remember when I said I'd be back to feast on your tears?
AMD down 12%.
Cry for me. Cry.
This is only the beginning - like I said AMD will bottom out around 15 or so in coming months.
AMD is making desperate move after desperate move.
AMD is the company with the inferior chips that it has to hock at firesale prices to get anyone to buy them.
AMD is the company that just obsoleted all of its current customers and motherboards with the recent introduction of "4x4".
AMD is the company that announces vaporware product after vaporware product (4x4, Torrenza, etc) with nothing to back it up.
AMD is the company that is going to see 95% of its market share blown to bits by the "perfect storm" of Conroe, Woodcrest, and Merom.
Conroe 3.2Ghz is already being tested by the reviewers and will further slam AMD into the performance doldrums.
Meanwhile AMD is desperately trying to buy a graphics company so AMD can alienate their #1 motherboard partner, Nvidia.
AMD is just a bunch of morons throwing spaghetti at the wall and hoping something sticks. Only a fool would buy AMD today.
"How do you explain that AMD with 1 factory does 20% of the market and Intel requires 5 or 6?"
Intel produces a lot of chipsets, embedded solutions, wireless products, etc.
I explain your statement above by stating that most of Intel's CPU production comes from 3-4 factories which actually agrees very well with 80%/20% market share split. (Nice try though)
btw, AMD's 'matured yield' 7 years ago is 60-70% :) i'm not sure what is the number now.
You heard of Sematech?
Their annual rating of fabs list AMD as the best yielding fab in the world... Intel is nowhere near close or near the top.
If AMD is obsolete Intel fanboy above, then what do you call all the Netburst products that are still over 90% of their mix? Who's in trouble now? ;p
"then what do you call all the Netburst products that are still over 90% of their mix?"
I'll try again...ever hear of mobile?
You know this new-fangled technology where computer chips go into this device called a laptop that people can carry around. Or by your logic is that less than 10% of Intel's chips?
"Their annual rating of fabs list AMD as the best yielding fab in the world... Intel is nowhere near close or near the top."
I assume you are refering to AMD's analyst slides?
A) Other than AMD no other fabs were labeled because Sematech does this anonymously for its member comapnies (and provides a graph to each member company which only has their company identified)
B) I think the actual slide was "efficiency" not YIELD.
C) There were no units on the graph so it is impossible to say whether AMD was 5%,50%,500%...
Or perhaps you were referring to AMD normalized yield graph which was trying to show they hit mature yields faster. That of course had a line marked "mature yield"; which obviously has to be interperted as 100%? And this had no comparison of other competitors (because yield information is highly condifential and is not shared between companies!)
Keep on spouting the misinformation under the guise of facts.
You heard of Sematech?
Their annual rating of fabs list AMD as the best yielding fab in the world... Intel is nowhere near close or near the top.
I happen to come from a place where AMD assemble it's chip. And as i said, the data i gave is 7 years old and refering to the test yield, not sort yield,. And if AMD was the best 7 years ago , then it must be a joke. It just means 30+% out of the 'supposingly good' die were scrapped after assembled.
Even More Obsileet P6 procsesors for molble wich are going to be maid into CORE2 procesors with some hacks above Pentium m / Core duo/solo.
"I'll try again...ever hear of mobile? "
Then keep trying buddy... you mean those 32-bit laptops? Awesome... have fun when Vista comes out, and enjoy.
"I assume you are refering to AMD's analyst slides?
A) Other than AMD no other fabs were labeled because Sematech does this anonymously for its member comapnies (and provides a graph to each member company which only has their company identified)"
Yes, and this shows AMD at the top... If Intel were that great, wouldn't you think they'd be touting this at THEIR analyst meeting? APM is truly a great piece of software to increase yield.
"B) I think the actual slide was "efficiency" not YIELD."
Then I guess you can't think and have a pea sized brain. It's not efficiency... it's a normalized yield. Regardless, it's still tops.
"C) There were no units on the graph so it is impossible to say whether AMD was 5%,50%,500%..."
Hence... normalized yields.
"I happen to come from a place where AMD assemble it's chip. And as i said, the data i gave is 7 years old and refering to the test yield, not sort yield,. And if AMD was the best 7 years ago , then it must be a joke. It just means 30+% out of the 'supposingly good' die were scrapped after assembled. "
7 years? You mean the stuff that came out of Austin Fab25? Aweomse!! Glad you can come out of a hole... and welcome to Y2K... wait, it's 2007.
Fab30 in Germany has been tops in Sematech fabs list several years running now. Go back to your hole and live with the cavemen please.
Your logic is equivalent to saying ALL American cars have poor reliability because it did 7 years ago. GM actually has some cars that are tops in JD Power & Associates Reliability now.
Oh by the way... I heard Model-T's had some issue from my dad's dad's uncle back in the early 1900's... so the reliability, etc must still be poor cause no one can improve.
Lemme guess... you said you used to work for a place that assembled AMD chips? You're at Intel now...? Assembling their awesome 4MB Conroe parts? How are those yield coming along? Random Defect density of those LARGE Cache's must be fun to sort huh?
====
I like the blog and oftentimes disagree with Shak... but for me it's entertaining.
The Intel fanboism is pretty extreme. You do a great job of ignoring all of the ignorant comments...
"Then I guess you can't think and have a pea sized brain. It's not efficiency... it's a normalized yield. Regardless, it's still tops."
OK, now I have to call you out and let folks know that you are a complete idiot. Here is the link (page 17)
http://www.amd.com/us-en/assets/content_type/DownloadableAssets/DarylOstranderAMDAnalystDay.pdf
Please, oh wise on explain to us pea brains what units "performance index" is measured in?
Also if you could point me to the place where Intel is on this graph to confirm your obviously true statement that "Intel is nowhere near close or near the top"
If you knew anything about a fab, performance index could mean anything: cycle time, WIP turns, fab inventory (don't confuse this with finished product inventory), WS per square foot of clean room...
Please point me to any place where it says performance index in that graph means normalized yield.
Your statement (and Sharikou's): "then what do you call all the Netburst products that are still over 90% of their mix?"
My response: This is clearly wrong as Netburst architecture is not used in mobile and mobile is significantly more than 10% of Intel's mix.
Sharikou's response - in stealth mode update his blog after I point this out (by adding Core 32) hoping no one would notice and then respond to my comment "I said all Intel's legacy products"
Your ignorant response: "Then keep trying buddy... you mean those 32-bit laptops? Awesome... have fun when Vista comes out, and enjoy."
Please explain how I'm wrong saying that Netburst is not 90% of Intel's product mix? Or do you just want to change the subject again so people don't realize your ignorance?
"Their annual rating of fabs list AMD as the best yielding fab in the world... Intel is nowhere near close or near the top."
http://www.amd.com/us-en/assets/content_type/DownloadableAssets/DarylOstranderAMDAnalystDay.pdf
The previous poster was indeed correct, this is a plot of "performance index" which could mean anything and is not the same as normalized yield. It is also not clear if Intel is even on this graph.
You should own up to your mistake.
you said you used to work for a place that assembled AMD chips
Nope, i didn't say that. You know there are some area that is full of industry area or r&d hub, like san jose, mountain view, etc? You can know something about the other guy by word of mouth.
And anyway, I have no intention to bash AMD. What i try to do is tell the blog owner about the truth on who has better manufacturing prowess, using his own statement (AMD is good, intel is evil, and things like that).
The 7 years old data has nothing to do with the logic, just an addition piece of information for him to realize the APM, at least, was not that great at all. I do acknowledge I have no idea about their current yield, and not trying to create mis-information as the blog owner often like to do.
And often, the AMD fanboy here like to call others as fanboy, whenever they state out points that is against AMD, and the funny is that it is not about the AMD CPU, but its manufacturing capability.
refer to my blog on this, as i do acknowledge AMD will continue to be Intel's strong competitor, despite the fact Intel is taking the lead now. That's something an AMD fanboy would agree on the first part, but not the later part of the statement.
Computing-Intensive
Doctor Doctor, I can't claim credit but I found this great one at another fanboy site. Post if you are secure enough in your fanboyism..
All AMD has is price. Got to drop them as they have no benchmarks, have no power leadership. Got to keep the factory full as you lose less money selling at a loss versus selling nothig. Poor AMD, more like banckrupt AMD..
OLMY is the man! (8:12pm EST Fri Jul 21 2006)
Intel: "We're about to release a processor that uses much less power than the best AMD processor, and will beat it by 40% in benchmarks"
AMD: "Intel lies. Everybody knows that AMD has the best benchmark performance. That's why we're winning market share"
Intel: "We'll prove it. Here are some machines running a pre-production Conroe, and some machines running a top of the line AMD CPU. Watch the benchmarks. We win."
AMD: "Intel lies. The benchmarks were rigged by Intel's PR department. If those benchmarks were run in a fair configuration AMD would win hands down."
Intel: "Ok, we'll let a few of you run your own benchmarks. We win."
AMD:
Intel: "Ok, the chip is released. Now you can all try it. See? It's just like we said. We win by 40%. Every independant reviewer agrees. We told you we weren't lying!"
AMD: "Ah yes, maybe your benchmark numbers are better. But you clearly lied about your power consumption! If you measure power consumption the same way we do (but not the same way that the rest of the world has been doing it for years) we win hands down! You environmental terrorists! Your power consumption is robbing humanity of its future!"
Intel: "Uh, what the hell are you talking about? Independent reviewers already confirmed that our power consumption beats AMD when measured the way that it is usually measured. But now they went back and did it your way and we STILL kick your butt! So now everyone agrees that we win on both power AND benchmarks!"
AMD: "Benchmarks? Nobody pays attention to benchmarks anymore. That is SO 48 hours ago! Benchmarks haven't been relevant since we last had the best benchmark numbers. We phased out the use of benchmarks last week when it became clear that we were now losing. Can't you keep up?"
What's next? I'm guessing we will go to:
AMD: "Power consumption? Nobody cares about power consumption anymore! The REAL measure of the success of a processor is the fanboys/sale ratio"
The Doctor
All the hoopla about the Conroe is good PR for Intel, but it won't magically make its outdated inventory (Celerons/Pentiums) sell. I'm buying budget desktop systems and I wouldn't touch low-priced Intel desktop CPUs with a 20 foot pole. They're just no good compared with even the Semprons.
Intel has to play it very, very carefully in the coming quarters. The Core 2 lineup is basically its last chance. The amount of money it makes in the short term is fairly irrelevant - it's the PR that matters most here (as the armies of zealots from either side can attest to). With all of Conroe's strengths, all it takes is one bad chipset issue, supply problems or god forbid another Pentium-style CPU bug to nail Intel.
"Intel's aggressive launch schedule for its Core 2 Duo chips has sent ripple effects throughout the industry as the world's largest chipmaker dumps old processors on the market to make way for the new chips. The resulting price pressure has caused problems for Intel's partners and competitors, as well as Intel itself." - Tom Krazit
"The market was caught by surprise that Intel was bringing forward their product launches; they thought they had a longer lead time to get rid of their inventory. Intel's actions have sort of shaken up the industry and put additional pressure on the pricing environment in the second quarter." - Charles Smulders, Gartner analyst.
~~~
"According to our sources, the current C1 stepping of the 965p chipset might have some compatibility issues with some ram. From what we understand, it has some issues with Kingston and ADATA DDR2-800 modules. There is also a upcoming C2 revision which is said to improve async memory performance."
Maybe we should wait for a C2D revision C2, a 965p revision C2, and a motherboard revision C4...
Bechya intel will have pentiumI stile processor bug, did you hear about the raid debacle with woodcrest wich is a modified CONrow. so peple that could aford it have raid and it will run slower.
Post a Comment
<< Home