Wednesday, March 01, 2006

Merom/Conroe performance claim is a hoax

INTEL claimed 20% performance advantage of Merom/Conroe over AMD's Athlon 64 X2 is a hoax. Upon information, clock for clock, Merom/Conroe will not be any faster than AMD64 for real applications.

This paper is a good analysis of the so called INTEL NGMA -- Bob Colwell's 1995 P6 from an architectural point of view.

Israeli amateurs will be discredited when real numbers show up.


Anonymous Anonymous said...

Sharikou's claim that Merom/Conroe performance is a hoax is not grounded on any test. We will not know the answer until next fall. The article he quotes proves nothing.

11:50 AM, March 01, 2006  
Anonymous Anonymous said...

And Intel's claim of 20% faster is a hoax because there isn't any evidence either. Yes, we'll know by next fall. Also, Intel's BSing proves nothing either.

1:10 PM, March 01, 2006  
Anonymous Zetro said...

Before netburst came out there was a bunch of talk in regards to how much faster it was going to be clock for clock. Until a product gets released to comment on how much faster it is really does not say anything.

1:34 PM, March 01, 2006  
Anonymous Anonymous said...

Once again Sharikou is abusing his eminent status as a journalist to promote his own personal agenda. He presents no real data, no hard facts, but only hearsay. Has he done these tests himself? I would like to see his methodology. He has produced exactly nothing except innuendo and hot air.

1:58 PM, March 01, 2006  
Anonymous Anonymous said...

Once again INTEL is abusing its eminent status as a x86 monopolist to promote its vaporware. INTEL presents no real data, no hard facts, but only vapor.

2:18 PM, March 01, 2006  
Anonymous Anonymous said...

Intel sucks.

2:22 PM, March 01, 2006  
Anonymous Anonymous said...

Let's talk about architecture. Intel invented it after all. If you look at the benchmarks you will find that AMD leads in many of them, but there are areas where the P4 architecture does better like video encoding/processing. With the Yonah architecture, Intel matches or beats AMD clock for clock. This is without having an integrated memory controller or the use of hypertransport. If you factor in power consumption then the metric really moves in Intel’s favor.

3:28 PM, March 01, 2006  
Anonymous Anonymous said...

This site has an article showing Yonah inferior to AMD64 in all aspects. Give it up, there is nothing P4 can do better than AMD64.

4:18 PM, March 01, 2006  
Anonymous Anonymous said...

You need to give it up. Rumour has it that the top two speed bins of Yonah will be notably lesser in real world power draw than their lower brethren, leading to 'faster' chips with real world longer battery life. This may explain why Yonah went from a planned 2.5GHz +/- a bin to 2.16GHz, the extra clock was sacrificed on the altar of power efficiency.

4:38 PM, March 01, 2006  
Anonymous Anonymous said...

Yeah, the AMD fans love comparing a desktop processor to a mobile processor. Why don't the AMD fans compare the Turion to Yonah? Oh yeah... there's no comparison in terms of speed, that's right. Given that even the fastest 2.16Ghz version of Yonah/Core Duo has a low TDP, Intel could put out a desktop version and really crank the speed up. At 2.8GHz with a reasonable TDP, nothing AMD has would be able to touch it.

OK.. Now Conroe/Merom are going to be even more efficient per MHz than Yonah, and the "Conroe Extreme Edition" is set to run at 3.33 GHz with a 1333 MHz FSB. So if a 2Ghz Yonah (with a 667mhz fsb and 2mb l2 cache) can more or less keep up with a similarly clocked Athlon 64 X2 3800, what will a Conroe Extreme Edition running at 3.33Ghz, with 4mb l2 cache and a 1333mhz FSB do to the Athlon 64 X2 - which tops out at 2.6Ghz? (Yeah, the FX-62 is set to run at 2.8 GHz) That's still a 533 MHz disadvantage to an architecture that will be more efficient per MHz. When Conroe hits the streets its game over for AMD.

(When I read this little post myself, it is so biased it's not funny... closely following the writing style of the owner of the blog here!)

P.S. "Give it up, there is nothing P4 can do better than AMD64." makes NO sense Mr. Anonymous. Considering in your comment you were talking about Yonah - a processor based on a TOTALLY DIFFERENT ARCHITECTURE TO THE P4. This is like me insulting a Ford Falcon because I think a Toyota Camry is a better car and then writing 'Give it up, there is nothing Ford Focus (a totally different car) can do better than Toyota Camry!

6:29 PM, March 01, 2006  
Anonymous Anonymous said...

Sharikou's mention of the Annadtech comparison of a low-end notebook Yonah to a high-end AMD 64 desktop chip is a stupid article to mention. It only weakens his argument and shows that he is an ignorant who doesn't understand the technical details.
If this is the superficial manner in which he gathers his data, then his study proves nothing.

7:13 PM, March 01, 2006  
Anonymous Anonymous said...

How bigotted can this guy be to call the Israeli's "amateurs". I am definitely losing faith in his judgement on these matters. Here's an article.

8:04 PM, March 01, 2006  
Blogger Sharikou, Ph. D said...

On Turion 64 against Centrino, Laptoplogic has done a comprehensive test, the Turion won 15 tests, Centrino won 7. So on single core mobile, Turion64 beats Centrino single core. Based on this, one can expect the coming dual core Turion to outperform Core Duo also.

In the AnandTech test, except motherboard and CPU, the X2 3800+ and Core Duo systems used idential components. It is a valid test, otherwise Anandtech wouldn't have done it. INTEL is going to use Yonah in server environment, called Sossaman32.

INTEL's so called NGMA mobile, server and desktop will be all based on Merom, you can see that from INTEL's own press release. INQ has long advocated that INTEL should use Pentium-M for desktop, INTEL finally followed INQ's advice.

As for Yonah's TDP, INTEL's own tech docs show it has a maximum power of 67 watts. AMD's Opteron 870HE is 55 watts. I don't see any INTEL advantage there.

Conroe running at 3.33GHZ is vaporware just like the 9.2GHZ Nehalem which was supposed to be released in 1005, show us the 2GHZ Conroe first.

Israeli amateurs got a lot to learn, first show us the beef, give us some test machines to run, and prove they can clone AMD64 instruction sets right.

Or, I will agree with some of the INTEL engineers, hyping is a brain-damaged behaviour.

8:54 PM, March 01, 2006  
Anonymous Anonymous said...

If Intel had any advantage in real chips they would be crowing from the tree tops,
they haven't and can't produce superior 64-X86 performance for multicores to scale effectivley without equivilant hypertransport and integrated memory controllers.
They can't lie straight in bed

1:15 AM, March 02, 2006  
Anonymous Anonymous said...

How could this be a hoax when the Israeli team does benchmark tests for confirmation. This is cleary the real deal, everyone that says this is a hoax is probaly jealous they wernt smart enough to think of the Yonah architecture themselves. NO it is NOT a hoax! It is very real. It should have never been called a hoax in the first place.

7:51 AM, March 02, 2006  
Anonymous Anonymous said...

deleted again. sad.

10:14 AM, March 02, 2006  
Anonymous Anonymous said...

Recall the Paxville? INTEL said it would leapfrog AMD's dual core, Michael Dell said INTEL would be supercompetitive, what happened? Only more embarrasment for INTEL. Why don't INTEL folks give out some Conroes for a test ride?

10:19 AM, March 02, 2006  
Anonymous Anonymous said...

hypertransport and integrated memory controllers. To comment on the above, until now the "hypertransport" bus has not changed since it was introduced on the Athlon t-bird?, I think. I may be wrong but it was designed with DDR266 in mind and they haven’t done a single thing to it since. If they were this wonderful company why have they not improved on the technology till now. In fact I even heard that in the DDR2 memory controller they didn’t even "improve it" they just updated it so it would support the new ram. I’m sorry but please know what your talking about when you post. At present it is NOT necessary for integrated memory controllers, I don’t believe it will be until memory bandwidth reaches a much more significant level.

10:20 AM, March 02, 2006  
Anonymous Anonymous said...

I realize that Intel has been pretty skeptical about integrating the memory controllers, but they will do it anyway. According to Eng Lim Goh, Senior Vice President and Chief Technology Officer of SGI, in 2007 Intel will standardize interface of the IA32 and IA64 processors and will integrate the memory controller into the CPU. The server microprocessors that will be available in 2007 will reportedly use common serial interconnect CSI bus instead of traditional processor system bus and will support FB DIMM memory modules. But you are right there will be no need for the integrated memory controller until system bandwidth is greater.

AMD is still losing the battle even though Intel doesn't have the IMCs. It will be lights out for AMD when Intel adds the IMCs and the bandwidth demands exist.

11:13 AM, March 02, 2006  
Blogger Sharikou, Ph. D said...

INTEL already cancelled the CSI project. All modern CPUs (Opteron, UltraSprac T1, Power5) have integrated memory controller, only INTEL is doing 1970s FSB stuff. At two CPU core level, INTEL is still marginally OK, above that, you have a situation which INQ described as many elephants breathing through a single straw. Opteron can scale gluelessly to 8 way SMP, it will soon to be able to 32P ccNUMA. INTEL is losing market share very fast. It doesn't make any sense to buy an INTEL today even for diehard intel users--they should wait and see what Conroe/Merom hoax is really about.

INTEL amateurs are no match to AMD's grand masters. Last time INTEL had some innovation was Bob Colwell's P6, today, INTEL's survival still depends on that piece of 1995 technology.

11:28 AM, March 02, 2006  
Blogger Sharikou, Ph. D said...

AMD64 is what AMD called the 8th generation architecture, or DCA (Direct Connect Architecture). INTEL's architecture is similar to Athlon T-bird or K7. In fact, INTEL's next generation Bensley/blackford is an imitation of the old obsolete Athlon MP. Athlon MP had no hypertransport or IMC but it had dual independent buses. AMD is still selling Athlon XP, but it's under the name of Geode NX.

Read this and understand why AMD is five generation ahead of INTEL.

11:36 AM, March 02, 2006  
Anonymous Anonymous said...

Your still quite wrong my friend. Intel has already developed an integrated memory controller before AMD ever did I think, but nice try. Remember RAMBUS ram, Intel had a very short lived processor with an integrated memory controller for the RAMBUS technology when that flopped, they dumped that whole idea just in case it ever happened in the future. So the fact AMD has been successful is that DDR has in fact stayed and technology has not shifted away from them. So your statement is infect very false Intel has been there and one that already and moved on.

1:04 PM, March 02, 2006  
Blogger Sharikou, Ph. D said...

Last time INTEL did an IMC was in 386 age, and it failed. Furthermore, without cache coherent inter-processor interconnect such as ccHT, processors with IMC can't go SMP easily. So those twi pieces of puzzle are related.

1:11 PM, March 02, 2006  
Anonymous Anonymous said...

Clearly, Intel is whipping AMD because it continues to use the front side bus. Intel has resisted changing its frontside bus architecture. An integrated memory controller can only work with the memory standard for which it was designed. With memory standards changing every 18 months or so, this means that companies have to redisgn the IMC's in order to enable transitions such as the switch from DDR (double data rate) memory to DDR2 memory. This is why Intel has stayed away from IMC design. Their current FSB technology will have no problem scaling to quad-core in 2007. Beyond that they may be forced to either use IMCs or add more logic to accommodate new RAM standards. They are well ahead of AMD in designing new logic that can accommodate new RAM standards. AMD will falter as their chips will not be able to take advantage of new RAMs.

1:47 PM, March 02, 2006  
Anonymous Anonymous said...

"Last time INTEL did an IMC was in 386 age, and it failed."

In 386 age INTC was the clear winner among AMD and INTC just like today. Give it up. There's nothing that AMD can do to compete with INTC. It has always been like that and always will be.

5:37 PM, March 02, 2006  
Anonymous bbbl67 said...

Intel is hoping to make its own parts perform better than AMD's by including caches that are two to four times larger than AMD's. Problem is it needs them to be that much larger, because we're already at the point of diminishing returns for caches. There was a time when you could simply increase a cache by 25% over the competition for a noticeable improvement in performance. Then later you needed a 50% increase, and later still, a 100% increase. Now we're approaching 300% increases, just to get any sort of blood out of this stone. It seems so obvious that you need an integrated memory controller these days, it's just so unbelievably dense of Intel not to even consider it. Somebody once said that Intel has a chipset market to protect and the only way it can differentiate itself from its 3rd party competitors is with a superior memory controller in the chipset. However, on an integrated memory controller environment, all chipsets are more or less equal. Intel wouldn't be able to differentiate its own chipsets from the 3rd parties, who would be undercutting it in price. AMD has no major chipset business to protect, and therefore it's got some fantastic 3rd party chipsets all competing against each other on price and features.

8:34 AM, March 03, 2006  
Anonymous Anonymous said...

This is all a moot point. It is no better speculating that Conroe will be any better than the AMD equivalents, than to speculate that they wont.
The only real comparison, as you say, is to wait for the tangible production chips. Unfortunately, all we can do is wait until release to know the truth.

5:32 AM, April 15, 2006  
Anonymous Anonymous said...

bunch of morons. Well it's now fall, and intel chips are 20% faster than AMD. Intel made the claims, whilst testing their pre-production chips. Based on how Core Duo took off, you should know by now, 2006 is the year of Intel Core. AMD needs to go back to the drawing board.

1:23 PM, October 01, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home