Tuesday, June 06, 2006

Our Coverage of Anand

On June 5, I published the SysMark 2004 comparison between Con E6300 and Athlon 64 X2 3800+. Many visitors from intel.com read that page. On June 6, Anand showed SysMark 2004 comparison between Conroe XE and FX62.

Some people keep quoting Anand's stuff as gospel truth. I have proven that Anand is a paid Intel pumper. So please stop quoting Anand as reliable source here, except for critical examination.
  • Dempsey vs Opteron benchmark: Why did AnandTech handicap the Opteron? The responses I quoted in the comments were from Anand himself. We know now, Dempsey is no match to Opteron in 64 bit performance.
  • Anand's benchmark on Yonah: X2 3800+ won by 16:6, yet Anand initially gave X2 a negative conclusion. After seeing those big Centrino ads on the same review page, it became all too clear. Anand's primary source of income is ad money.
  • IDF: Was Anand duped by INTEL? Anand pushed the Intel arranged buttons, instantly wrote that "Intel Regains the Performance Crown". After the BIOS issue, Intel called Anand back and had the machine BIOS flashed and tests redone. No one else got such VIP treatment from Intel. I suggested to AMD that it should subpoena Anand for information regarding the IDF test as part of the discovery process in the anti-trust lawsuit.
  • One month later, Intel quoted Anand's words and showed them to Wall Street analysts and investors. Intel market cap increased about $100 million that day. You notice that Hexus.net was also quoted by Intel (see page 51 of this Intel presentation to Wall Street).
  • Today, Anand published what he claimed to be an independent Conroe benchmark, while others could only push Intel arranged buttons. How could Intel trust Anand so much more than others? Anand claimed that he gathered the Intel parts. How? The Taiwanese were suddenly not afraid of Intel any more and just gave Anand the Conroe CPU?

Let's make some quick comparisons. In this test by PCStats, an Athlon FX62 got a SysMark 2004 Office Productivity Overall score of 261 . Another independent result for a slower Athlon 64 X2 5000+ (2.6GHZ,2x512KB) agreed with PCStats' results quite well. There, the X2 5000+ with a 7200RPM HD got 230 points in SysMark 2004 Office Productivity Overall. In Anand Lal Shimpi's test, FX62 (2.8GHZ, 2x1MB) only got 210, Conroe XE (2.93GHZ) got 266. Look at the FX62 sub scores from PCStats, they were 263 for communications, 297 for document creation, and 214 for data analysis. However, in AnandTech's results for FX62, the sub scores were respectively 178, 280, 185.

Let's look at PCStats.com result on Business Winstone 2004, the FX62 got a score of 36.4 there. However, at AnandTech, FX62 only got 27.9, while the Conroe XE got 32.8.

Why were AnandTech's scores on FX62 substantially lower than the scores obtained by others?

Anand is not dumb. He knew that AMD64's main advantage is low memory latency due to the integrated memory controller (IMC). AMD64 doesn't need huge cache in general because it can access memory quickly. AMD estimated that IMC's low latency gave its CPUs 20% performance edge. Intel Conroe's solution is to use large cache to compensate the lack of IMC. With this knowledge, Anand decided to use high latency 5-5-5-12 DDR2 memory for his test. As a result, FX62's low latency IMC advantage was almost eliminated.

As you can see from this newegg.com memory shopping page, most DDR2-800 memory in the market today has 4-4-4-12 or lower latency. In fact, on newegg.com, out of 59 DDR2-800 memory products, only 15 models have CAS latency of 5, the other 44 products have CAS latency of 4 or lower. 4-4-4-12 memory is 25% quicker than the 5-5-5-12 used by Anand.

AnandTech's results on FX62 should therefore be considered invalid if not fraudulent. Based on AnandTech's SysMark 2004 results on Conroe XE (2.93 GHZ) and PCStats' results on Athlon 64 FX 62:

SysMark 2004 Office Overall: Conroe XE scored 266, Athlon 64 FX 62 scored 261
Business Winstone 2004: Conroe XE scored 32.8, Athlon 64 FX62 scored 36.4

This agrees with our previous findings.

Charlie at INQ commented that "[i]t would most likely be cheaper to buy all the hardware sites out there off". I guess some company has already done that. What did Charlie know? A hardware site, a script kiddie finished reading "How to Upgrade Your PC" pushing benchmark buttons. I doubt his annual income is big. He should be cheap, as our Charlie observed.


38 Comments:

Blogger "Mad Mod" Mike said...

Wow, it's nice to know that even w/ 5-5-5-12 timed RAM, the FX-62 can hold it's own against the 2.93GHz Extreme Edition Conroe in everything but Gaming (4MB vs 1MB). Next, how about that Iranian buddy runs 3-3-3-8 timings, wait, that would actually show AMD winning, sorry, he can't do that.

7:30 PM, June 06, 2006  
Blogger "Mad Mod" Mike said...

3-3-3-8 can improve up to 18% on 5-5-5-12, add in the 193MHz Overclock and 3-3-3-8 RAM, than the FX-60 is faster clock-for-clock on non-gaming benchmarks, all that's next is for 4x4 to own Conroe up and down the block.

7:35 PM, June 06, 2006  
Blogger "Mad Mod" Mike said...

Sharikou, that website you're linking to "http://www.hwupgrade.it/articoli/cpu/1479/comparativa-processori-maggio-2006_9.html" - Is just the FX-60 at 2.8 w/ DDR1, if they used real DDR2, expect those scores to go 10-15% higher.

7:55 PM, June 06, 2006  
Blogger netrama said...

I cant believe, that Intel put
websites like hexus.net and anandtech in slides to present to the Wall Street. Is that slide for real ??

8:37 PM, June 06, 2006  
Blogger Sharikou, Ph. D. said...

I cant believe, that Intel put
websites like hexus.net and anandtech in slides to present to the Wall Street. Is that slide for real ??


Of course. It was taken from Intel's Power Point. The presentation was given in New York sometime after Intel's bad 1Q06 report. It is on page 51 of
this PDF
.

8:58 PM, June 06, 2006  
Anonymous Anonymous said...

Will this eventually tarnish Anand's reputation? I hope so!

But people have a short memory, so we'll just have to remind them often.

--

If Intel was purposely deceitful I would have to have 'some kind' of respect for that.

But the fact that Intel pumps money to web sites to make their products seem better than what they really are... and then those same Intel execs read those articles and believe what they read.

Freak'n idiots... and how much are they getting paid?

9:13 PM, June 06, 2006  
Blogger netrama said...

Thanks ..sharikou
Wow.. I checked the pdf ..it is an amusing read :

1.) Page 5 "Positioning for our best product rollout in YEARS" ...
Hmmm, Paul u mean the previous products were shitty ..right ??lol

2.) PAGE 6 .kind of weird how the Gross margin in 2006
will be 54%. if they plan to sell conroes at sub $200 range.

3.) Page 8.) "re-purpose Intel" ..ha ha ..Yes I will give
all the marketing team automatic weapons, if need arises !!

4.) Page 17 - shameless!!!

5.) Page 50 - "build emotional link" ??? what the heck is this ??

6.) Page 62 - Why is an ipod nano shown here ??

My Own Conclusion - These marketing mother-fathers are going to
take Intel DOwn !!

10:06 PM, June 06, 2006  
Anonymous Anonymous said...

The fact that Intel can resort to such desperate measures only shows how afraid they are to AMD. They are really very afraid.

10:12 PM, June 06, 2006  
Anonymous Anonymous said...

http://www.tgdaily.com/2006/06/06/intel_confirms_reorgnization_fewer_employees/

You're were correct again MR. Sharikou.


also the study about the avaliability of Conroe is RIDICOULOUSLY LOW
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2771&p=6
even the Anand guys accepted that Intel will lose money big time there. because NOONE WILL WANT OLD NETBURST tech.

10:20 PM, June 06, 2006  
Anonymous Anonymous said...

great blog, sharikou!
nothing feels better to bust such a paid bumper.

10:23 PM, June 06, 2006  
Anonymous Anonymous said...

More benchies to digest..

http://www.reghardware.co.uk/2006/06/04/preview_intel_core_2_duo/
http://www.firingsquad.com/hardware/intel_core_conroe_benchmarks/
http://www.computerbase.de/artikel/hardware/prozessoren/2006/test_core_2_duo_e6700_testdrive/
http://www.extremetech.com/article2/0,1697,1970188,00.asp
http://www.coolaler.com/modules/news/article.php?storyid=470

And don't forget the one at THG!

10:40 PM, June 06, 2006  
Anonymous Anonymous said...

There is only one conclusion after all these tests.

1. FX line should get another 2MB of L2 cache. Better latency doesn't help, better DDR2 pack also.
2. X2 should reach 3GHz ASAP!
3. AMD should lower prices.

Or just point No.3
Maybe Anand is a bumper, paid heavly, but Conroe EE beats FX.

11:18 PM, June 06, 2006  
Blogger Sharikou, Ph. D. said...

There is only one conclusion after all these tests.

You missed the whole point. All other tests were Intel controlled ones. Anand did the first reviewer controlled test. However, his comparison FX62 setup was deliberately handicapped. Take the Conroe XE SysMark 2004 scores, and find FX62 SysMark 2004 scores from other reviewers. The picture is completely different.

11:27 PM, June 06, 2006  
Anonymous Anonymous said...

Although 4-4-4- is 25% faster than 5-5-5 you do not get this sort of decrease in performance using 5-5-5- ... see here

http://xbitlabs.com/articles/cpu/display/amd-socket-am2_13.html

5% seems to be loss ..which is hardly enough for people to say using 4-4-4 would mean FX wins.

11:32 PM, June 06, 2006  
Blogger Sharikou, Ph. D. said...

The most
comparable benchmark I found is here by PCStats.com
.

1:08 AM, June 07, 2006  
Anonymous Anonymous said...

i'm shocked. always liked Anandtech, go there to get some of the latest news and it has an ok forum(not like x-bit labs).. i've been reading you're stuff Sharikou and you seem to know what you are talking about, keep it up :D it's great to see that not all is just buying in to everything that they are told.. i'm gettimg AM2 X2 4000+(65W) this summer :D always worked for me, and AMD dosn't lie about performance :).

6:29 AM, June 07, 2006  
Anonymous Anonymous said...

You missed the whole point. All other tests were Intel controlled ones

No. My point is that AMD should get 20% boost to have undisputable margin between Conroe and X2. And after that introduce 4x4....

6:42 AM, June 07, 2006  
Blogger "Mad Mod" Mike said...

If you look at the SysMark and Business benchmarks, and do what I said previously of factoring in not 4-4-4-12, but 3-3-3-8 and add in the fact Conroe has another 193MHz on the FX-62, you can see the FX-62, clock-for-clock, just like I said, is faster than Conroe.

7:32 AM, June 07, 2006  
Anonymous Anonymous said...

If you compare Anand benchmark configuration for FX62 and PCStats configuration, you will notice that PCStats used WD Raptor, in Anand it was Hitachi SATA disk. So let's wait until Review not Preview from Anand.
I think he knows that he is watched carefully and will not permit mistake in real comparision. So let's wait for better sparring parter (FX64) and the Conroe available in Q3.
I agree with memory strange config.

7:43 AM, June 07, 2006  
Blogger Sharikou, Ph. D. said...

Regarding the difference in HD.
Let's look at this test on a slower Athlon 64 X2 5000+ (2.6GHZ, 2x512KB), it used a 7200RPM Hitachi HD, the score for SysMark 2004 Office Overall was 230. This result for X2 5000+ was 10% higher than Anand's FX60 result.

8:10 AM, June 07, 2006  
Anonymous Anonymous said...

Sorry if you already have this, my Intel had a hickup while it was submitting last time... lol

With the memory issue that was pointed out during Anands test of Conroe and FX62, what would happen if they used this memory... Corsair 8500?

I am not sure how to compute the numbers, but if they both have the same 5-5-5-15 timings, and running at the memories 1066 speed, how would this end up?

And isn't Conroes FSB 1066?

Would this favor Intel or AMD?

Thanks

8:58 AM, June 07, 2006  
Anonymous Anonymous said...

It is ALL about the latency.

Check these benchmarks out:

http://staff.bit-tech.net/tim/corsair-pc10000-report.htm

9:43 AM, June 07, 2006  
Anonymous Anonymous said...

Dear anonymous who posted links:

http://www.reghardware.co.uk/2006/06/04/preview_intel_core_2_duo/page6.html

this tells you perfectly
"so bad the test was in controlled enviroments"
in short words, MORE BUTTON PRESSERS AND NO REAL TESTING!.
same technique they used with anand, and toms..

10:19 AM, June 07, 2006  
Anonymous Anonymous said...

Check out Anand's latest Woodcrest comparison. "The Intel Xeon 5160, a.k.a. Woodcrest, will simply be the most powerful server CPU this year (though it's not yet available for purchase of course). As our extrapolated calculations show, even a 2.6 GHz Woodcrest will outperform the current Opteron 285 with a 5 to 55% margin, nothing short of impressive"

http://www.anandtech.com/IT/showdoc.aspx?i=2772&p=11

11:40 AM, June 07, 2006  
Anonymous Anonymous said...

Amazing that the fallout we see on sites such as Toms, Anand, and others from readers of their sites.

I've stopped going to Tom's for awhile, but Anand... I've always found his site to be fairly valuable with technical information.

Credibility is a very fragile thing... and if you mis-lead your readers/audience, and they find out, good luck on them coming back. No visitors = no ad revenues.

Good job of info digging. I'd like to see Anand respond. :)

11:55 AM, June 07, 2006  
Blogger Sharikou, Ph. D. said...

I'd like to see Anand respond. :)

He won't. The problem with review sites is this: there is a major mismatch between their relative low income and the market value of goods they are reviewing. A review can cause massive change in a company's market cap. For Intel, a $0.15 increase in stock price is $1 billion added market cap. You can probably buy a site off with $50K, the pay off is huge.

10:28 PM, June 07, 2006  
Anonymous Anonymous said...

Sharikou, just because a website shows Intel doing better than AMD on benchmarks does not mean that the website is being bribed by Intel. You need to produce some credible evidence from a 3rd party if you expect your claims to be taken seriously.

Also, I think it's pretty funny how you censor your blog. Why do you do this? Are you afraid of criticism or something?

9:35 PM, June 09, 2006  
Blogger Sharikou, Ph. D. said...

Sharikou, just because a website shows Intel doing better than AMD on benchmarks does not mean that the website is being bribed by Intel.

Of course not. At the beginning, I thought he just overlooked something. But then with repeated acts, and the way he tried to BS his rationale. You sense sth.

I have clearly established the Anand's willfulness in rigging the FX62 benchmark. Now, why is he doing that? Once you establish a pattern, then you correlate it with the ads. You have a highly probable proof. It's not beyond reasonable doubt proof but clear and convincing.

Why do I reject some of the comments? Because they have zero value. I don't care if someone attack me or my opinion. But you have to show some good reasoning. You can't say this is wrong. You have to say why it's wrong, logically and backed up with facts. So either I or someone reader can debate it.

9:48 PM, June 09, 2006  
Anonymous Anonymous said...

"Regarding the difference in HD.
Let's look at this test on a slower Athlon 64 X2 5000+ (2.6GHZ, 2x512KB), it used a 7200RPM Hitachi HD, the score for SysMark 2004 Office Overall was 230. This result for X2 5000+ was 10% higher than Anand's FX60 result."

The thing about those benchmarks is that they were run with completely different drivers than Anandtech. Beareyes used nForce 9.16 while Anandtech used 9.34 Beta. Beareyes also used ForceWare 84.21 while Anandtech used 91.28 Beta. Also Beareyes ran everything at 1024x768 while Anandtech used 1280x1024. Personally, I don't have much confidence in Beareyes considering they were running "WindowsXP Porfessional"

In regards to the PCStats comparison, besides the Raptor issue they also have driver differences. I'm not sure whether the fact that PCStats used a SLI setup while Ananctech used a single card could of had an effect also. It still isn't easy to dismiss the use of Raptors by PCStats because both the Winstone and SYSMark Office Productivity tests are very I/O bound where HD choice is very important.

I'd be curious for you to explain why Anandtech's results seen to be perfectly inline with PCStats when Winstone and SYSMark move to the Content Creation portion of those tests. Anandtech's FX62 scored 373 on the Overall SYSMark Content Creation while PCStats score an insignificant 2 points higher. Even Beareye's X2 5000+ only scored 345, well below either system. In Winstone Content Creation, Anadtech's FX62 actually scored higher at 48.6 to PCStats 46.2. You'd think that if the lower latencies are so beneficial with 4-4-4 being "25% quicker" than 5-5-5 they'd show themselves here. If two results were enough for you two classify Anandtech's article "invalid if not fraudulent" then do two results inline with PCStats make the results valid again? Or should they just be discounted or ignored?

"As you can see from this newegg.com memory shopping page, most DDR2-800 memory in the market today has 4-4-4-12 or lower latency. In fact, on newegg.com, out of 59 DDR2-800 memory products, only 15 models have CAS latency of 5, the other 44 products have CAS latency of 4 or lower. 4-4-4-12 memory is 25% quicker than the 5-5-5-12 used by Anand."

Link link you posted to Newegg was this:
http://www.newegg.com/Product/ProductList.asp?N=2010170147+1052416064&Submit=ENE&SubCategory=147

I'm not sure if it's changed since you checked or if you are lying, but the information you talk about is not accurate. Looking at the "Drill Down" selction options under "Cas Latency" we see only 49 memory options listed. Of those 16 are Cas 5 and only 33 are Cas 4 or lower not 44. It really should be noted that JEDEC standard for DDR2 800 is for 5-5-5 timings. I'm not sure how wise it is for AMD to design a processor that really needs a non-standard latency spec to show it's full performance.

I'd also like to see you respond to the Hexus and Firing Squad independant reviews of Conroe.

http://www.hexus.net/content/item.php?item=5692

http://firingsquad.com/hardware/intel_core_conroe_benchmarks/default.asp

I know you've criticized Hexus, but Firing Squad is usually pretty reliable. Certianly statements like "Make no mistake about it, this isn’t the same AMD of six years ago that was often forced to compete on price with Intel, AMD is now in the enviable position of market leader when it comes to performance and power consumption, while Intel counters by pricing their dual-core processors at bargain-bin prices." from their FX62 recent FX62 review doesn't make it seem like they're biased toward Intel. Their Conroe review showed that a 2.4GHz Conroe could keep up with and even beat the FX62. It certainly gives support for the performance of the 2.93GHz Extreme Edition over the FX62.

8:27 PM, June 10, 2006  
Blogger Sharikou, Ph. D. said...

Regarding the differences in SysMark Office Producticity benchmarks, screen size, hard drive speed or graphics card make very little difference (less than 0.5%). This is obvious from the composition of tests (word,excel, etc). I don't like to quote AnandTech, but if you are a believer, see this test:
http://www.anandtech.com/printarticle.aspx?i=2073

As for availability of DDR2, even with your numbers, it doesn't change the overall picture. DDR2-800 with CAS latency 4 is the predominant product in the market.

As I analysed before, we should expect Conroe to do well in single threaded tests with working set comparable to 4MB. Otherwise, Conroe should be close to K8 clock/clock. Let's wait till Conroe is out and check it out with both Conroe and FX having the optimal settings.

8:46 PM, June 10, 2006  
Anonymous Anonymous said...

Well. Everyone should also take a look of this site. It’s in Chinese, but graphs should give you some idea. 3DMark comparison. They also have an English forum if you want to ask questions.

AM2 FX-62 vs Conroe E6700
http://www.coolaler.com/modules/news/article.php?storyid=470

1:04 AM, June 15, 2006  
Anonymous Anonymous said...

I always looked at Anand's site as a neutral website, not until they decided to publish 3D rendering benchmarks with lightwave 8.x showing only hypervoxels times instead of heavy FP results with radiosity or raytracing, they are totally biased.

Now with all this crap I'm replacing their bookmark with your website Mr.

Thanks for uncover this scamers.

12:40 AM, June 22, 2006  
Anonymous Anonymous said...

I've got a 3700+ San Diego and an X2 3800+ Windsor. So, to put it lightly I'm an AMD fan. However, this crap you are posting is an absolute joke. You'll post anthing to keep Intel down. Good job way to spread the BS and trash a well respected site. I'm going to buy an E6300 (~200MHz slower than my X2) and email you the results of how much it will destroy my X2. You're pathetic.

6:56 AM, July 25, 2006  
Anonymous Anonymous said...

I've got a 3700+ San Diego and an X2 3800+ Windsor. So, to put it lightly I'm an AMD fan. However, this crap you are posting is an absolute joke. You'll post anthing to keep Intel down. Good job way to spread the BS and trash a well respected site. I'm going to buy an E6300 (~200MHz slower than my X2) and email you the results of how much it will destroy my X2. You're pathetic.

6:57 AM, July 25, 2006  
Anonymous Anonymous said...

This isn't even about C2D vs X2. This is about Anand not being honest, and quite possibly being paid-which I myself discovered sometime before even seeing this blog.

Someone said that AMD shouldn't "require" or words to that effect that their processors use, well, the best memory on the market, to run at full performance. It's not a requirement, it's just that with an integrated memory controller (which Intel will be going to shortly as well, thank you very much) makes the only latency that really matters be that from the speed and timings of the RAM-whereas with Intel the cache and prefetchers take care of that, making RAM timings as well as speed almost irrelevant.

"Well AMD shouldn't design their processors to perform better with better RAM" is basically what you're saying. Does that make ANY sense AT ALL?

I will say one thing about Anand-when I wondered what the difference was between 2MB and 4MB of L2 cache on the Conroes, they managed to do that comparison without screwing it up.

3:23 AM, December 01, 2006  
Blogger JoJo said...

Wow, i guess you were right, AMD really did prove that Conroe couldn't hold up, and the Conroe line really did crash and burn in the face of AMD's superior performance. Oh wait, none of that happened. You say, 1 site shows Intel is better, and this one shows AMD is better, and your logical conclusion is that the site showing Intel ahead is clearly biased. Does anyone else see this?

5:28 PM, March 05, 2008  
Blogger AlexSuffer said...

This comment has been removed by the author.

3:13 AM, January 27, 2017  
Blogger AlexSuffer said...

http://www.theinquirer.net/?article=31994
Nice article

3:14 AM, January 27, 2017  

Post a Comment

<< Home