ZDNet Technical Director calls shens on latest Barcelona performance figures

Wednesday 04th July 2007, 12:12:00 PM, written by Rys

ZDNet's Technical Director, George Ou, has called the latest round of pre-release Barcelona performance data by AMD, blatantly deceptive, cherry-picked and simply wrong.

After the most recent round of performance data found its way out onto the web, where AMD compare a simulated 2.6GHz Barcelona quad-core processor to parts from Intel's quad-core Xeon product line, George has taken serious exception to the data.

"I’ve seen benchmarks get cherry picked and twisted before but this is just outrageous. AMD is deliberately leaving out Intel’s best scores, leaving out Intel’s best products that shipped months ago, and putting in theoretical Barcelona scores for products that don’t even have a ship date. After Henri Richard (AMD executive) came in front of our ZDNet cameras to slam Intel for “un-ethical behavior” and promising not to do the same, we have caught them on four separate occasions behaving unethically. After this latest incident, it’s clear that AMD has no intention of behaving honestly or ethically."

Grabbing the latest figures from SPEC.org, something AMD didn't do, George shows that the theoretical performance figures offered by AMD don't look so great in that different context.

He also has issues with the fact that the 2.6GHz Barcelona Opteron is simulated and won't ship this year, whereas Intel have been shipping the comparison Xeon since April.

For the ZDNet Technical Director to use such strong language and go after AMD like that is quite something, just as it is for AMD to publish a set of figures like they did in the first place.

Read George's 'blog post in full over at zdnet.com.

Discuss on the forums

Tagging

amd ± zdnet, barcelona, shens, spec, opteron


Latest Thread Comments (12 total)
Posted by Farid on Wednesday, 04-Jul-07 11:53:20 UTC
Quote
After claiming to have the high-road on ethics, AMD showed hypocrisy on three separate occasions (one, two, and three). But this latest round of deceptive benchmarks is so outrageous that it’s criminal.
This will be an interesting thread, I think.
Quoting Simon F
"Shens"?
Shenanigans. If that's what you're inquiring about.

Posted by Geo on Wednesday, 04-Jul-07 13:02:44 UTC
". . .so outrageous that it's criminal" and ". . no intention of behaving honestly or ethically" are the kind of statements that can get you introduced to gentlemen in $1,000 suits with shark smiles. I wonder if AMD will try, like Nvidia did with FM long ago, to pressure a softening of that language, or decide to leave well enough alone.

Posted by Farid on Wednesday, 04-Jul-07 14:00:14 UTC
Quoting Geo
". . .so outrageous that it's criminal" and ". . no intention of behaving honestly or ethically" are the kind of statements that can get you introduced to gentlemen in $1,000 suits with shark smiles. I wonder if AMD will try, like Nvidia did with FM long ago, to pressure a softening of that language, or decide to leave well enough alone.
FutureMark is a software company that is in business partnership with Nvidia. This is a ZDnet editor expressing his point of view in an editorial piece.

Start suing the press is one of the most major mistake a company can make, PR-wise.

Posted by dess There aren't the "most recent" figures! on Wednesday, 04-Jul-07 15:30:56 UTC
Quote
After the most recent round of performance data found its way out onto the web, where AMD compare a simulated 2.6GHz Barcelona quad-core processor to parts from Intel's quad-core Xeon product line, George has taken serious exception to the data.
This is stupid. AMD publicized these figures back in november last year!

What AMD publicized now are comarisons to an Opteron 2222:

Image: http://www.xbitlabs.com/images/news/2007-07/2pserver.jpg

Image: http://www.xbitlabs.com/images/news/2007-07/4pserver.jpg

(Note that the numbers should be interpreted as x% faster.)

Posted by Geo on Wednesday, 04-Jul-07 15:55:16 UTC
Quoting Farid
FutureMark is a software company that is in business partnership with Nvidia. This is a ZDnet editor expressing his point of view in an editorial piece.

Start suing the press is one of the most major mistake a company can make, PR-wise.
If he'd said, as he should have IMO, that it OUGHT to be criminal, he'd be expressing his point of view. As it is he's said something that very well may be actionable, regardless of whether it's actually acted upon.

And let's not rehash 2003, but that's an awfully simplistic view of what happened.

Posted by trinibwoy on Wednesday, 04-Jul-07 16:23:49 UTC
Quoting Geo
If he'd said, as he should have IMO, that it OUGHT to be criminal, he'd be expressing his point of view. As it is he's said something that very well may be actionable, regardless of whether it's actually acted upon.
I didn't know the use of conditionals was a requirement for a statement to be considered an opinion. Isn't it enough that the words came out of his mouth (or pen) ?

Posted by Geo on Wednesday, 04-Jul-07 17:50:19 UTC
What conditional would that be? "ought" isn't conditional in the least. It's advocacy. As in "There ought to be a law making this kind of thing illegal so it'd be criminal to pull this crap".

Posted by trinibwoy on Wednesday, 04-Jul-07 18:52:02 UTC
You're right, it's not a conditional but my question still stands. Why do you need a qualifier in there at all? I know jack squat about libel law but I just find it amusing that someone could be sued because they didn't explicitly state that what they are saying is an opinion. Considering the definition of an opinion and all....

Posted by Geo on Wednesday, 04-Jul-07 19:36:19 UTC
Because there are actual objective standards to reference re "criminal", so by using that word he's arguably acting with a reckless disregard for the truth, which is one of the standards for libel. Unless, of course, he can show it to be criminal by pointing at the relevant law that's been broken.

FM by using "cheating" (which it seems to me while a nasty word is less nasty than "criminal"), actually was in a somewhat better position legally as it would have been very hard for NV to have shown that FM acted with reckless disregard to an objective standard as to what constitutes "cheating".

Tho this is all academic --I don't actually expect a suit from AMD over this. Just moderately surprised that ZDnet didn't have a paid weasal add a few out clauses to the language that really wouldn't have weakened the point they were making at all.

Posted by dess on Thursday, 05-Jul-07 04:06:50 UTC
Well, George Ou at first suggested that those results were just made up by the marketing at AMD after they had to admit the low launch clocks. Not true. These numbers were advertised in April, when they were correct ones, as 1. Intel introduced the 3GHz Xeon 5365 later, 2. the better benchmark results for Intel that he took for his comarison also were submitted later. Anyway, also "leaving these numbers posted on AMD's website is blatantly wrong". Oh well.

The other objection, so that there will be no 2.6 GHz Barcelona short term is perhaps a more good point.

But let's look how he resolves the situation: by taking the most recent (and best to date) results on Intel's part. Fine. Let's jut ignore the fact those are produced using a newer complier with significant performance improvements... (And say the difference is because of - what else - AMD's deliberate misinformation action.) I would think making a comarison using results with the older compiler for both is still more correct, as a compiler probably affects both sides similarly, so the ratios remain relatively intact...


Add your comment in the forums

Related amd News

RWT Analyzes Bulldozer Benchmarks
AMD Bulldozer microarchitecture analysis
Say hello to GLOBALFOUNDRIES
AMD completes deal with ATIC to create The Foundry Company
AMD Propus to be released in Q2 & Q3
AMD launch 45nm Phenom II processor
AMD goes Asset Smart; splits into two
Beyond Programmable Shading course notes available
AMD launches FireStream 9250 with 200Gflops DP via RV770
AMD GPGPU solutions get extra support from industry partners