- Saturday, 11 July 2009
- INQ Mobile
- RSS
Look wise, say nothing, and grunt. Speech was given to conceal thought - Sir William Osler
First Shanghai benchmark numbers tip up
IT SEEMS THE first numbers from AMD’s soon to be launched Shanghai have tipped up online, although AMD is keeping them very quiet indeed.
Results published by SPEC on a benchmark for power efficiency include what appear to be the first Shanghai scores, ahead of AMD’s launch on Wednesday.
Shanghai at 2.7GHz manages to score 860, while the 2.5GHz manages a score of 731.
All well and good, except that with those kinds of scores, Shanghai doesn’t even come close to Intel's already existing Harpertown chips, let alone Nehalem.
Intel chips certainly seem to be the big winners as far as this particular benchmark is concerned, with Power Leader running Chipzilla’s L5430 Xeons and scoring 1135. IBM also pulls two very high scores from its hat with Intel’s Xeon X3360 2.83GHz (1064, 1054), while NEC also posts a high score of 1010 on a Xeon L5420 2.5GHz.
But there’s also something a bit suspect going on in these benchmarks, something which makes us think AMD might be up to some shenanigans.
AMD submitted its own results for Intel Xeon L5420s at 2.5GHz and somehow managed to come up with a score of only 561. How come? Well, by very badly configuring the system to ensure the Intel server got the lowest possible score, it would appear.
The system sports eight FBDimms at 2GBs each when, with only four FBDIMMS at 4GB each, a much better score can be achieved. The system would run faster and would also cost less.
But AMD seems to be stuffing its box to crank the power up via C, thus artificially hammering the power efficiency. What this amounts to is picking the worst case scenario and possibly using that as a base from which to post a particularly hostile submission.
AMD also appears to be using different memory configurations on its own machine, specifically 4x4 instead of 2x8. This, we assume is because 2x8 would slow AMD down a fair bit.
So, what could AMD have up its sleeve? Could it be that these curiously low Intel numbers might make an appearance in performance comparison graphs at next week’s Shanghai launch to 'prove' how good the new product is compared to rivals? µ
L’Inq
SPEC
Benchmark
Share this:
lol they want to bore us to death, a power benchmark? I mean this is the product that will save AMD, rival Intel, and just kick its stupid little Intel butt! But no, we will release a benchmark only on power consumption. seriously who is running this ship over at AMD, I think its stuck in a whirlpool, its going around in circles, and bits are flying off the ship, and its gaining speed almost into the center to be completely lost and forgotten about.

lesson1 AMD
To keep people interested launch benchmarks that SHOW THE PERFORMANCE, not the amd shanghai used 12.33211123453 watts V the Intel core 2 used 13.98330443005632

AMD WINS YEY (rolls eyes) ohh brother give me stregth
The answer is very simple and has been published in HPCWIRE http://www.hpcwire.com/features/17886984.html and Sandia ( Sandia Micro-Benchmark Suite ) National Labs research over 2 years ago. http://www.cs.sandia.gov/smb/overhead.html and further demonstrated by Lawrence Berkeley National Labs http://www.cs.berkeley.edu/~samw/research/papers/ipdps08.pdf in an IEEE award paper this year. There is a world of difference in performance between a <4k page file and 64k and 128k page files. Intel's problems with large page files has figured in the decision to use AMD and IBM in the Petascale R&D http://www.nitrd.gov/pubs/2006supplement/hec_rd.pdf and also why Intel has yet in 6 years to win any of 72 possible HPC awards from DARPA. The author doesn't realize that benchmark optimization cuts both ways. The only way to really tell is to run the same benchmarks under the same operating conditions. The author should spend next week in Austin Tx at the SC'08 meeting. Discussions on these issues by noted experts are on the agenda. 
"AMD also appears to be using different memory configurations on its own machine, specifically 4x4 instead of 2x8. This, we assume is because 2x8 would slow AMD down a fair bit."

I hope you just made an honest mistake here and not deliberately misleading the readers. I checked the links you provided and AMD is sporting the same mem config as the intel at 2.5GHz submitted by AMD. Both systems have 8 x 2GB mem config.

Of course you do have a point. While configurations may be the same, consumers should not believe product benchmarks of a rival company from its rival. Consumers should do their own evaluation before buying.
AMD is acting so nvidia-like.
Are you trying to pull a fast one on your readers. Yoou start of by saying that you suspect that AMD has some shanghai scores in spec. In other words, this is speculative, you are not sure, it is not actually factual, since you cannot guarantee that it is indeed the score for Shanghai. Yet, that doesn't stop you from ranting about how AMD appears to be massaging the results, and you state it like it is fact. Is this a pre-emptive strike? Oh well, sounds like an article from a loser, but I guess you already know that. :-)
Shanghai 2.5GHz used 2x8 DIMM just like L5430 Xeon in AMD's benchmarks. The stacked deck claim by not using 4x4 doesn't add up.
AMD is comparing it's own 2.5 GHz Opteron 2380 to Intel's 2.5 GHz Xeon L5420 using the same memory configuration on both systems (8 x 2GB) as well as the same hard drive and NIC. Where is this conspiracy you are alleging?
Check out how this moron tries to spin these scores as a good thing for AMD.

http://sharikou.blogspot.com/
"All well and good, except that with those kinds of scores, Shanghai doesn’t even come close to Intel's already existing Harpertown chips,..."

http://www.spec.org/power_ssj2008/results/res2007q4/power_ssj2008-20071205-00023.html

http://www.spec.org/power_ssj2008/results/res2008q4/power_ssj2008-20081017-00092.html

... E5450 vs 2384 

i think shanghai looks ok ... (80W tdp intel vs 75W acp amd)

The first Opteron versions are still only HT1 and DDR2....
Well obviously the Shanghai platforms were badly configured as well as those AMD tested Intels.
AMD put a rather generic config for the Xeons. I don't have an idea where all that performance hit came from though, besides FBDIMM and the SATA-2 drive. Maybe SUSE doen'st have a good power management by default?
I hope AMD don't go down this route, and hope for their sake that you are wrong. People just havnt got time for this crap anymore. I have amd earmarked for my next upgrade, when they produce a decent overclockable processor, however, should they pull such a short sighted stunt in the meantime, they will be placed even with Intel as far as i am concerned.

It was nice to see ati loosing with the hd3k series, and not messing with benchmarks etc. Hopefully this will remain at least.
Funny, I've always heard code names, known they were places...but never until my city was one of them did I realize how cool it was. :P

Long live Intel...and AMD, step it up. This is pathetic.
i'm not server person, so I have a potentially dumb questions.

The AMD setup had 2 JVM instances (761 score)
The Intel setup (by AMD) had 4 JVM instances (531 score)

Is the # of JVM instances significant and if yes would a higher # help or hurt the score?

Based on the article after this one, it does seem AMD is quite defensive about this and is probably playing a little fast and loose with the setups.
Please dont ever leave the Inquirer. Youre my favorite
When every other tech site just recycles, inquirer innovates. Dont know how you found these first shanghai numbers, but well done! Thats why Im sticking with the Inq, its tech journalism, not just tech "news".
One is single socket and the others are all the expensive Low power versions of the Xeon.
I think power will be very close between intel and AMD now that they are both on 45nm.

What you should concentrate on is that the 2.7Ghz Shanghai has a higher ssj_ops score compared to the 3Ghz duals socket version of intel.
Which, kids, is why you distrust AMD's numbers on Intel as much as Intel's numbers on AMD. This is possibly even more blatant than the outright fraud Apple tried when they launched G5 systems, and produced some "benchmarks" vs that terrible child-murdering x86 (which were considerably lower than any you could find by Intel on spec.org). Apple appeared to run the Intel chips on milk and harsh language, and configure them by hiding them behind the fridge, and it seems that dear old AMD are doing the same.

Of course, Intel would do the same- if they needed to lie about their performance currently they'd do so.. probably more competently, too. However, they can appear the paragon of truthiness(tm) right now, as they currently have the best part.

The moral here, best beloved, is that they're all a bunch of lying abstrads. Sadly, we have to see through all the subterfuge to select parts which fit our needs for price/performance and efficiency- and in some cases, software support.

Intel, Nvidia (calm down Charlie, you might soil yourself again) and even darling AMD are not to be trusted- none of these sorts of outfits are. If you shake hands with a representative of one, count your fingers.
..AMD/ATI/DAMMIT using cfg's that make 'em look good, OMG, what is the world coming too...? The marketing boys will be happy, it's all a big conspiracy, ask Drashek....
Sigh.

AMD once again tries to cheat against Intel by submitting an under performing Intel configuration and calling it apples to apples with AMD's.

Please.

AMD fanboys should quit crying. Demand that AMD compete fairly and stop cheating like this.

Great expose article by George Ou on this latest deception by AMD

http://www.formortals.com/Default.aspx?tabid=36&EntryID=130

So Intel hands you your ass (again) and all you can say is "look how little power we draw to be p0wned like that !" ?
Maybe you should draw some more power and get competitive on the performance level, hmm ?
Ahh, the beauty of competition.
C'mon, that's 8 FBDIMMs versus 4 DDR2s .. of course the power consumption goes up! A single FBDIMM already higher power consumption than a single DDR2 stick.

AMD is a big cheat. Anyone can remember AMD's simulated Barcelona benchmarks? History repeats again.