Sunday, December 10, 2006

Misinformation Rolls Onward

I had hoped that hardware review sites like Toms Hardware Guide and Anandtech would somehow rediscover the integrity they used to have. I've hoped that financial analysts would stop measuring Intel with a shorter yardstick than they use for AMD. However, even a brief look at recent articles shows that bias towards Intel is very much alive and well.

In this THG Clovertown article we see several problems right away. For example the only two compilers used are: Intel C++ Compiler Version 9.1.030 and Intel Fortran Compiler Version 9.1.029. This guarantees that Intel will have an advantage because the Intel compiler purposefully skips optimizations on non-Intel processors. Secondly, we see that the bulk of the "tests" are either single or dual thread. I'm still baffled why anyone with common sense would test a quad core processor with single threaded code. This is a bit like testing a 500mph test vehicle in rush hour traffic in L.A. rather than someplace more suitable like the Bonneville Salt Flats. I suppose you could make the same lame argument that L.A. traffic is more like real driving conditions but a test is useless if it is artificially limited as so many of the recent tests have been. We also see that in the supposed "load" test that the testers fail to actually achieve any load on the processor. Intel gives the TDP of the 5355 as 120W while the 5160 is only 80W. Yet in the THG chart on page 11 the 5355 actually draws less power. I had strong hopes that THG might get back on track during 2007 but this level of incompetence or outright bias is hard to break. THG remains useless for real testing and benchmarking.

Another interesting article was this Economist piece. The theme of the author is clear as the article is entitled The Empire Strikes Back. Backing this up is a supposed graph showing that Intel's market share is on the rise and growing faster than AMD's. The only problem is that this graph is bogus. To get this imaginary increase they include numbers from VIA whose cpu's sell at a price below both Sempron and Celeron and therefore have no effect on Intel and AMD processor sales. The numbers for VIA were temporary inflated because VIA was dumping large numbers of processors that were End Of Life. By adding these anomalous VIA numbers in they make Intel's Q2 share look worse and give the impression of a changing trend. In reality, Intel's share has been slowly dropping while AMD's has been slowly rising. The change is roughly 0.66% per quarter. This trend should continue and if it does AMD will indeed arrive at 30% volume share by the end of 2008. I guess the problem is that if Intel has been slowly losing share for the last six months it would be impossible to use those snappy catch phrases in the article like, "right-hand turn", "turning the wheel", "successfully navigated the transition", and "Intel cemented its comeback". I guess the true phrases like, "Intel continues slow loss of share" or "Intel's reorganization was not deep enough" just wouldn't generate the same kind of excitement. In reality there has been no big turnaround in share for Intel in spite of the introduction of Core 2 Duo. Today, Intel continues to slowly lose share to AMD. Perhaps it is time to give up on the financial analysts as well.

Over at Overclockers Ed is claiming that AMD will only hit 2.9 GHz with dual core on 65nm. I'm not really sure where Ed got this information from. Common sense would suggest that if AMD can hit 3.0Ghz with FX-70 on 90nm then they should be able to do at least a little better with 65nm. If Ed had speculated 3.2Ghz I think that would be reasonable. Even 3.1Ghz would be a bump over 90nm speeds. However, I think speculating a speed that is lower than the speed for 90nm is just absurd. Yet in the same article he speculates that Intel will hit 4.0 Ghz with dual core on 45nm. Considering that Intel has not yet shown or announced anything faster than 3.0Ghz I would have to say that this is optimistic. The question is why the optimism always seems to be in Intel's favor. Overlockers is another website that is unreliable for a balanced look at Intel and AMD.

A website that is usually much better, Ars Technica, however showed some serious errors in their Barcelona article. This is the faulty paragraph:

It does not seem that K8L will catch up to Conroe in terms of the theoretical peak number of 128-bit SSE operations per cycle, however. K8L's two floating-point/SSE pipes give it two 128-bit SSE ops/cycle, and its FSTORE pipe can do another 128-bit SSE move per cycle, for a total of three per cycle peak. This is half of Conroe's peak theoretical throughput of six 128-bit SSE ops/cycle.

Conroe cannot do 6 SSE ops per cycle. It is true that sometimes Conroe can do a simultaneous SSE multiply/add in a single cycle however 6 ops is pure fantasy. Where the 6 ops number comes from would be two simultaneous multiply/adds plus one load plus one store. However, the actual datapath to L1 cache can only handle two 128 bit words at a time. So, a more realistic number is two 128 bit SSE operations per cycle which interestingly is what K8L will be able to do. It is true that sometimes Conroe could go faster by executing combined multiply/add instructions but the article fails to count the L1 datapath advantage that AMD has over Conroe. My best guess is that AMD will actually come out slightly ahead of Conroe on SSE.

I think that there are many who are still hoping for Intel to pull decisively into the lead again as they did in 2002. However, this is not happening and won't happen. The true situation is that AMD and Intel are becoming more and more alike both in terms of corporate structure and in terms of processor design. I'll talk more about this in my next article, Walking Into A Mirror

21 comments:

Anonymous said...

This guarantees that Intel will have an advantage because the Intel compiler purposefully skips optimizations on non-Intel processors.
Is this another conspiracy?:)
http://techreport.com/reviews/2006q3/e6300-vs-sff/index.x?pg=7
There's a recent review with Intel+MS compiler where both sides gain.

Maybe Ed was talking about the K8L dual cores..which we've seen from HKEPC to top out at 2.9GHz, and even that's questionable, assuming that AMD wants their quads to be the fastest at everything, especially for Octo FX.

So is there any chance that one day you will stop talking about the conspiracy against AMD and talk about some interesting stuff like Quad FX and 65 nm, both paper launches as of now, the same thing you ragged about Intel not too long ago?

gdp77 said...

THG remains useless for real testing and benchmarking.


please tell us which exactly benchmark(s) would show the true power of quadFX and would not favor Intel

Yet in the same article he speculates that Intel will hit 4.0 Ghz with dual core on 45nm.

Maybe this is because today's conroes can hit out-of-the-box >3.5 GHz and up to 4.0GHz stable?

Anonymous said...

Scientia says Intel compilers skip non Intel CPUs. I shown a bench where it benefits AMD. ? Are you saying that AMD would benefit more if Intel supposedly didn't disable more? I looked on spec.org, searched AMD as company, compiler by Intel, and out of 67 results in Q306, 32 were compiled with Intel. If Intel's compiler is so evil, why don't they do their own?

http://www.hwupgrade.com/articles/cpu/10/quad-fx-the-first-quad-core-amd-platform_index.html
Closes the gap a lot with Vista, but QX6700 is still ahead.

The initial 90 nm CPUs weren't high clocked/high TDP to begin with though. We're seeing high TDP to begin with with Barcelona specs.
http://www.hkepc.com/bbs/itnews.php?tid=709944
New English version:)

BTW, I think calling people's 'argument's silly is silly enough, but it's easy to see what you're actually trying to say by calling people's arguments 'stupid', 'to the point of lunacy', 'lacking common sense', etc, not directed toward anyone;)

Ed cites the 3.5/4.0 from VR-Zone and Inq but even acknowledges that they may be a bit hopeful. And I don't see anyone saying over 4GHz.

And if HKEPC isn't a trusted rumor mill, then I don't know what is. When did I imply that they were trying to prove anything? I said that initial 2.9 goals can't seem to be met and that 2.5 rated at 120W does not seem like they'll be able to push more. What info in the past have they gotten wrong? Lowering quad speed to 2.5 could also show that yields aren't very good, which can be supported by lack of Brisbane info and 'simulated' benches with just 40%/70% gains. Not that those are small numbers, but I would expect more from doubling cores as well as an enhanced core. Rhetoric?.. You're the one that seems distraught with the stream of not so good AMD news lately..:)

http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543~114483,00.html
AMD Athlon 64 FX-70 series dual-core processors are expected to be available from NewEgg throughout North America on November 30. Worldwide product availability is expected by early 2007.
http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543~114609,00.html
..immediate availability..
If those aren't paper launches, I don't know what is. They are claiming immediate availability, not shipping.

gdp77 said...

gdp77, thanks for actually sticking to the topic buddy, as scientia hasn't even mentioned 4x4.

I sincerely apologize for mentioning 4x4. I was under the impression that the link provided from scientia was the 4x4 vs q2c review. My bad...

However my argument still remains unanswered. I would like scientia to point out the benchmark(s) that thinks they are objective and unbiased.

Anonymous said...

Sciantia said...

"In this THG Clovertown article we see several problems right away..."

I just looked at the review, and well I am not seeing how the compiler is relivant when they are not comparing Intel to AMD.

Could you explain?

"Intel gives the TDP of the 5355 as 120W while the 5160 is only 80W. Yet in the THG chart on page 11 the 5355 actually draws less power."

Well since you stated that most test were single or dual threaded, was the test in question (during load) a dual threaded test, thus allowing the other two cores to stay idle and allow the higher clocked dual core to use more power?

---------------------------------

"This trend should continue and if it does AMD will indeed arrive at 30% volume share by the end of 2008."

Why should it continue?

Intel is no longer designing processors based on Netburst, the move is to the Core architecture and as of now it is better than K8 in 2P, Mobile and Desktop.

If your talking about 4P and above, where AMD's processor design shines, that is a relatively small percentage of the overall market.

So where will these continual gains take place?

----------------------------------

"However, the actual datapath to L1 cache can only handle two 128 bit words at a time. So, a more realistic number is two 128 bit SSE operations per cycle which interestingly is what K8L will be able to do. It is true that sometimes Conroe could go faster by executing combined multiply/add instructions but the article fails to count the L1 datapath advantage that AMD has over Conroe. My best guess is that AMD will actually come out slightly ahead of Conroe on SSE."

I am sorry but this is either incorrect or it is showing a very strong bias on your part towards AMD...

Please look at what I have bolded.

1. You say they are the same 128bit.

2. You say Intel maybe faster due to the ability to combine.

3. Now your last statement says AMD will be slightly ahead while having the same two 128bit, and while not being able to combine?

Please explain.

"However, this is not happening and won't happen."

This is pretty bold statement considering that (supposed) future clock speeds will easily allow intel to pull away from AMD.

Projected performance numbers of K8L are now showing they will be extremely close and maybe a little ahead, but that is only on the 65nm level, if they end up being equal clock for clock AMD is in deep trouble.

45nm will easily allow Intel to get a Xeon 5355 at 3.5GHz inside the 130W envelope, AMD at 2.5GHz for a quadcore will be strugleing...

Just my 2 cents.

Scientia from AMDZone said...

The Intel Compiler advantage has been known for years and even Intel doesn't deny it. Tryng to refute it by linking to an article is pointless.

The problem is that THG uses the Intel Compiler at all. The only way to obtain comparative results is by compiling code on a genuine Intel system and then moving that code over to AMD. At least this used to work; Intel may have put in an initialization check.

I suppose what AMD really should do is create its own optimized compiler and provide it to places like THG and Anandtech. If AMD doesn't do this then I would agree that they are overlooking an obvious problem. The Intel Compiler problem will only get worse with SSE 4.

I've looked over the market history for AMD and Intel back seven years. There is simply no way to make the case that Intel is gaining or making a comeback; it just isn't there. I know this may seem in some way unfair or unjust after Intel released the clearly faster C2D but it isn't happening.

This could be related to the big price cuts. It could be related to Intel's overstock. It could be related to Intel's habit of abusing partners. I don't know the exact reason but there it is. Unless the numbers change Intel will keep slowly losing share and 30% will be a reality for AMD by end of 2008.

AMD's bus advantage is due to being split rather than due to width. This should be enough to offset Intel's combined ops unless the code is very carefully tuned for Intel.

I'm a little puzzled about the benchmarks question. You want me to mention the few benchmarks in the THG review that are actually relevant? If you were reviewing a 15 passenger van would you put 4 people in it? Thread limited testing is pointless.

Scientia from AMDZone said...

A paper launch is when a company releases the specifications of a product with their claims of performance without engineering samples to back it up. A hard launch is when the actual product or large numbers of engineering samples are available to back those performance claims. Most hardware is launched via paper launch.

No, this isn't accurate. A paper launch can be accompanied by engineering samples. Samples sent out to reviewers don't count.

Anandtech had a very good definition back in 2001 before they got so biased in Intel's favor. They said that when a sample was sent to a site for review that the chip should be available to the public at the end of that month. If it takes 2 months for a chip to actually become available after review then that is a paper launch no matter how many samples were sent out.

Vaporware is hardware that is talked about but never released. Examples of this include Whitefield, Tejas, and Nehalem.

Conroe was the perfect definition of a paper launch because reviews were encouraged even though the chip was not available. In contrast, K8L was a preview and therefore not a paper launch.

Scientia from AMDZone said...

There is also a soft launch where product is available but only in very limited quantities. This often happens with Intel because D1D is usually the first to produce a given chip but it is not a volume FAB.

A hard launch is when a product is announced or reviewed and product is available in significant quantities.

Scientia from AMDZone said...

Ashenman, I have no idea where you got the idea that engineering samples have anything to do with paper launches. Engineering samples are always available long before a product is actually in production. For example, there were engineering samples of Itanium out a year before release. Itanium was not a paper launch however because the review was not sanctioned by Intel. Bootleg reviews by people who come into possession of engineering samples also do not count.

Scientia from AMDZone said...

I just looked at the review, and well I am not seeing how the compiler is relivant when they are not comparing Intel to AMD.

Could you explain?


Sure, look at this review where they are comparing Intel with AMD and notice that they only use the same two compilers.

Woodcrest and Opteron

Yes, it would be nice if this were some type of exception but it isn't. THG routinely tilts the tests in Intel's favor.

Anonymous said...

There's and interesting phenomenon I've observed a couple of years ago when I had a summer job in a PC retail store : Intel owners simply refused the fact that similar AMD solutions were superior from a Perf/$ and Perf/Watt point of view. Simply because the lack of IT knowledge they simply wanted "Pentiums" and wouldn't want anything else. Also more educated friends of mine simply refused any reasonable arguments about AMD being better by arguing Intel was more stable than AMD and, Intel being faster for optimized applications!

This is a phenomenon I couldn't understand to this very day.

Now something like this is happening with AMD fans denying things they shouldn't. I guess It takes a long time to switch sides especially when you are a supporting fan!

This brigs me to the point I was trying to say:
TH (and others) realized that Intel has the biggest crowd of supporters that for three years had to put up with their supporting company taking beatings. If TH would have continued to be supportive of AMD they would have lost bigtime, they would have lost all they Intel supporting readers. Because all those readers ever wanted all of this time was for Intel to dominate once more.

It's hard to understand why someone would sustain a monopolistic company but I guess it's something like admitting your judgment is wrong. Owning and Intel and admitting AMD is better is like admitting you made the wrong judgment, which is one of the most difficult things to do ...
This is somewhat happening with AMD fans also ...

Scientia from AMDZone said...

Now something like this is happening with AMD fans denying things they shouldn't. I guess It takes a long time to switch sides especially when you are a supporting fan!

Who besides Sharikou is doing this? C2D is clearly ahead by a good 15% in integer IPC and a lot more than that in SSE. I suppose K8 does do okay with FP instruction but these should be becoming less and less relevant. I find it funny when I say things like this and people (who have to post anonomously btw) call me an AMD fan. That is sad.

Anonymous said...

I wasn't calling you a AMD fan, and I didn't mean Sharikou either.

But AMD does have it share of die hard fans ... [:)] I' was one of them a while ago. I still am an AMD fan but not die hard and not biased anymore.

Christian H. said...

As the most vocal AMD supporter on Tom's Forums, I can say that it their site is filled withIntel fanboys whomay not be as interested if "semantics" make Intel look bad compared to AMD.

it's the same most places. it is interesting how no other market has such fierce anti-competitor sentiment.

It's like Intel represnents some last vestige or this or that. (another story)

You are right though about the media and it's Intel spin. There is plenty of money for two CPU companies to operate in the world market, but Intel has to be the pinnacle of X86 achievement (RISC archs make that a contradiction in terms) for some reason.

Even now, people will applaud Opteron dualies, but say that QFX is a flop. (?)

I actually expect no balanced reprting from Ed Stroglio. He is like the king of the editorial "one-liner."

I ahvebeen saying for a year that AMD will contnue to grow with Core 2 out because AMD has a good portfolio that only requires one distinct chipset and now only TWO sockets(939/754 are as good a dead) AM2 and SocketF.

They also have the advantage of being the underdog and everyone loves to root for the underdog in these situations.

And notwithstanding Intel's predatory tactics that kept OEM margins razor thin while calling AMD the value alternative with less stability, there are companies who wouldlike nothing better than to stab Intel in the heart.

These ridiculously low Core 2 prices have kicked the bottom out of thebudget space and I see lots of small builders going out of business because of it.


But anyway, if MAD raises their prices they are responding. if they lower them they are responding. if they leave them the same they are responding.

Basically according to the Intel fanboys, AMD can only respond after Core 2 and everything they do is that.

They can no longer innovate their own products, they have to be "playing catch up."


No wonder I won't buy Intel. I'm not supporting such a flawed system of bias.

Anonymous said...

You claim that it does, I provide evidence that the compiler doesn't. I think the burden of proof is on you again, a simple "I'm right, nothing you can say matters" doesn't cut it. And good point Enumae, they aren't comparing your precious AMD with the compiler that helps all anyways.. I think you should also tell Tech Report and AMD and all the others that use the Intel compiler. I would rather AMD be faster though more behind with the Intel compiler than slower and closer without.

http://www.sudhian.com/index.php?/articles/show/691/1
AMD doesn't just grow because it's smaller and they haven't had nonstop growth for 7 years.

http://www.tgdaily.com/2006/11/03/amd_65nm_roadmap_update/
There is supposed to be a 90 nm 3.2GHz in Q207, now 2.5GHz K8L? quad, and 65 nm K8 2.8GHz in Q207, with higher clocked 90 nm CPUs to be introduced in Q307. Kind of hard to correlate one process/arch with another with clocks all over the place.

..You cannot justify twice the wattage of Quad FX for at best similar performance compared to something that is readily available;)

I don't have a problem with people that have preferences, just people that don't admit them;) I wasn't sure what rhetoric meant at the time, but it seems like about from the 2nd half of Scientia's last entry, you started arguing with emotion and name calling:( And sorry for my unnecessary use of fancy language, I'm trying to stop:)

Thanks for telling us that K8L is not a paper launch..? Conroe was PR released as was QX6700 with a date for availability. Quad FX was released to look not so behind QX6700 on the charts and claimed immediate availability. Call it what you want.

http://techreport.com/ja.zz?id=133143
http://forums.anandtech.com/messageview.aspx?catid=28&threadid=1972469&frmKeyword=&STARTPAGE=1&FTVAR_FORUMVIEWTMP=Linear
AMDroids are so plentiful and funny:D Seriously, you don't want me to spam this anymore necessary, do you?

I think it's silly to say that one is not biased because one acknowledges superiority. Even Sharikou can't spin 65 nm or Quad FX. This 99% pessimistic against Intel blowing up tiny cracks and silly conspiracies..

Khalif, people are praising dual Woodcrests now;) You can buy that Quad FX and melt/deafen your self just because it's AMD, that's your prerogative.
No wonder I won't buy Intel. I'm not supporting such a flawed system of bias.
Pot, meet kettle.

ashenman, I agree. I hope AMD does all it can to destroy Intel for the good of all of us. Analyzing it though, like you said, they're going to have a hard time;)

Anonymous said...

TheKhalif said...

"...if AMD raises their prices they are responding. if they lower them they are responding..."

If you do not think AMD lowering prices as a direct response to Intel you are naive, look at the current market...

1. AMD changed prices on Quad FX due to benhmarking that was showing Quad FX was over priced on a price/performance stand point, original (rumor) price for the two FX74's was $1500, on launch day they were at $1000, and a few days later $900.

2. Athlon FX62 was around $1000 ($1200 street price) at the release of Core 2 Duo, its at about $700 now.

These are just two examples of Intel forcing AMD to lower prices.

Here is a thought, have you seen AMD raise prices on any desktop processors since the release of Core 2 Duo?

The answer is no, they are milking Mobile and Server processor prices, as are Intel.

AMD is indeed playing catchup, and to doubt this is just ridiculous, and I would like a few examples of this from you in regards to desktop processors.

You may want to label me as a fanboi because I disagree with your opinion, but I am hoping that you can at see these two examples make sense, and AMD is indeed on defense.

Anonymous said...

Ashenman said...

"You'll probably notice very very soon that a lot of battery manufacturers are going to start to push lower power density batteries (due to the exploding batteries problem, and due to the high cost of lithium ion and the low cost of the newly popularized lithium polymer), and that mobile processors are going to be pushed to consume even less power."

Well your right...

1. "Instead of storing the lithium ions in organic solvents, the ions are held in a non-flammable polymer matrix. But there are more advantages than just the lack of explosions. Li-poly batteries do not require a metal casing to squeeze the battery's electrodes together so they can be up to 20% lighter than Li-ion batteries. Also, the form-factor of Li-poly batteries is much more flexible than the necessarily boxy or cylindrical Li-ion cells, and they can be as thin as a credit card."

And your wrong...

1. "Of course, they come with their disadvantages too. Most important to the consumer market, they are more expensive and they lose capacity faster than Li-ion batteries... Most Li-poly batteries on the market today require some fluoropolymers in the matrix. Fluoropolymers are expensive and difficult to create, requiring a lot of energy and chemicals. And, most troubling, there are currently no good programs in place to recycle lithium-polymer batteries..."

Also

"They need to be disposed of more rapidly than Li-ion batteries, they can't be recycled, and they contain some truly dangerous and energy intensive chemicals."

Scientia from AMDZone said...

Red, you are making a fool of yourself. The Intel Compiler problem has been around since Intel released SSE and has been documented over and over.

Secondly, even Intel itself admits that the problem exists and that they purposefully turn off advanced features for non-Intel processors. The fig leaf that they give for this is that they can't be certain that advanced features will work properly for non-Intel processors.

Red, I am truly sorry that you don't understand programming or compiler design. I'm sorry that you think that the links you provided give some strong evidence to support your claims which they don't. It is a waste of time to debate something with you as well established as this. Nor am I going to debate review bias when that too has been documented for the last three years.

Anonymous said...

Ashenman said...

"Once the materials are more available as production shifts to them, it will obviously become cheaper."

Yes, the price will come down, but it depends on the adoption of the technology like you stated, but, this does not mean it will be cheaper to produce than Lithium Ion even when production shifts.

There is nothing conclusive to say that Li-Po will be cheaper than Lithium Ion.

From what I read and posted it is more expensive and has a few drawbacks, mainly "they lose capacity faster".

PS: My post was not to prove you wrong, just to simply show more information.

Christian H. said...

Khalif, people are praising dual Woodcrests now;) You can buy that Quad FX and melt/deafen your self just because it's AMD, that's your prerogative.

I guess then that you were the most ardent opponent of Prescott and demanded that people get their money back if they chose it over FX60.

abinstein said...

"I think it's silly to say that one is not biased because one acknowledges superiority. Even Sharikou can't spin 65 nm or Quad FX."

The problem of your comments is not only many of your links don't work, but you are mixing unbiased facts with biased arguments.

Quad FX isn't good for K8 X2 to compete with Core 2 Duad. I don't think anyone with a clear head is going to spin that. However, Quad FX is not just for K8 X2 to compete with Core 2 Duad. Your belief of it is only so shows your true bias.

You first talk about QFX, then say people are praising dual Woodcrest. This is very wrong because one is enthusiast product whereas the other is server. A better comparison is dual Opteron and dual Woodcrest, where the former loses some on integer but wins some on floating-point. In short, their performances on dual-socket are comparable. Now for servers you not only want performance but also scalability. Woodcrest is losing here.

I also don't get your argument on AMD 65nm and K8L. Are you saying that they are late because some website claimed they would be early? Why don't you roll yourself back to AMD's Q2 2006 analyst day presentation and see how well the roadmap was kept up?

K8 is not competitive to Core2 on personal/desktop market; we all know that. This is unfortunately AMD's bread and butter over the years and I believe largely responsible of AMD's bad sales lately. However, K8 (not even K10) is still very competitive on high-end server market, and not yet have competition on 8-core setup (yet). What is the true bias is that people tend to related single-socket enthusiast performance to those of the higih-end servers. Apparently AMD's server sales also suffer from such common yet silly misconception.

In the end, it is wrong to say Core 2 is superior, becasue nothing in computer microarchitecture works simple like that. Core 2 is superior for PC enthusiasts or entry-level servers. K8 is still superior for mid-to-high end servers, especially when you consider future scalability.