AMD answers to controversial use of retired benchmarks in China
Summary: Earlier today, I wrote up the news of AMD's newest Opteron: the 3.0 Ghz 2222SE.
Earlier today, I wrote up the news of AMD's newest Opteron: the 3.0 Ghz 2222SE. In addition to news of the new 2P server-targeted dual-core processor's release, AMD also released some benchmark charts that show the 2222SE leading Intel's dual core 2P server targeted offering: the Xeon 5160. Competitive benchmark charts have been a touchy subject for the two companies as of late. During a recent press conference, AMD's senior vice president and Chief Marketing Office Henri Richard came out swinging and swinging hard at Intel for what he considered to be unscrupulously deceptive benchmarking practices. This, in addition to AMD's charges that Intel has violated U.S. antitrust law.
After digging into the issue a bit more, I concluded that Intel was indeed using deceptive benchmark charts in public forums -- charts that painted AMD in a far more negative light than could justifiably be done. Richard, and subequently, other AMD officials, were adament about the sorts of benchmarking practices that had to stop and committed to serving as a role model when it comes to the do's and don'ts of fairly presenting benchmark data. No sooner did AMD put that benchmarking integrity stake in the ground than it engaged in the very same benchmarking malpractice (usage of retired benchmarks) that it had accused Intel of. Those retired benchmarks appeared during a slide presentation that was given in Beijing, China last week by AMD's manager of worldwide Opteron market development John Fruehe (pronounced "Free"). Additionally, as I pointed out in my earlier blog post, the benchmark charts that AMD used today in briefings with the press as well as in an advertising campaign raise additional questions about how to most fairly present benchmark data.
Today, after discussing the Opteron 2222SE news with Fruehe, I asked him whether or not that was the right thing to do. To hear or listen to how that interview went, you can use the audio center above to download or stream an audio version of it to your PC or you can play the video below. Although he's not a lawyer and admits as much, Fruehe argues that there are certain challenges when it comes to offering previously unpublished benchmark data when a public company like AMD is in its SEC-enforced quiet period (a period just before a public company officially announces its quarterly earnings results.. a period during which a company is legally muzzled from releasing certain material information). As you can hear (or see), Fruehe eventually agrees that AMD can't be seen as a pot calling the kettle black. As soon as the video version is ready, I will drop it in here.
Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.
Talkback
I listened - Berlind was wrong
Yes it does. If Intel used old benchmarks when there was no reason they could have used more current ones, they were lying. If AMD used old benchmarks because they had no choice, they were between a rock and a hard spot. John Fruehe said AMD had updated benchmarks that would be released as soon as they were legally allowed and were much better.
It may be a convenient rock for AMD, granted. And I don't know what kinds of restrictions Intel might have been under when they made their presentation. If they were under the same "quiet period" restrictions, nobody mentioned it in the interview. (That's something I wish Berlind would have investigated, but I think he was more interested in an argumentative confrontation.)
But if that difference existed, Berlind dismissed it when it really did matter a lot.
No...it doesn't matter.
Actually, there should be...
Maybe Tom's Hardware or AnandTech could do the public benchmarking, but it would be far better if a totally independant, add free, respected organization did it.
Until then we will all be dealing with this crap unless the SEC steps in and threatens fines to Intel and AMD for misrepresenting their products abilities vs each other.
The Quiet Period Isn't A Joke
They did. Did you listen to the interview.
And then he wrote, "Tell the public that the information you DO have at the moment isn't enough to convey the true power of your product. Then when the "quiet period" ends (yawn), release the new (and convenient) figures."
They did that too. Did you even read my post?
Your post kind of defines what is sometimes called a "knee jerk reaction."
you're taking me out of context
db
More at... If a quiet period, why release anything?
This is like saying that you found the cure for bone cancer, but forgetting to mention that the cure only works for invertabrates.
Both companies are guilty of deceptful marketing. If Via could produce some results with their chip, I would jump ship to them.
AMD hypocrisy
Both Intel and AMD should...
Poor excuse, I know. I would like to respect AMD more than I do now.
AMD didn't say, "Intel lied first"
They said, "Intel lied, but we were unable to present all the information we had."
Companies, especially since the Sarbanes-Oxley Act, can get in serious trouble violating the rules.
It may seem like I'm defending AMD. I'm actually not. (I think they are just as capable of misrepresenting the truth as Intel is.) I'm saying that Berlind should have been a little more careful in his reporting.
Want more hypocritical lies
http://news.zdnet.com/2100-9593-6173644.html
Read this post
http://talkback.zdnet.com/5208-9593-0.html?forumID=1&threadID=32312&messageID=594715&start=-1
Which benchmarking software?
Every article I read on this story seems to omit this information.
Are we talking about Intel and AMD using say, a version 1.0 to test the competitor's CPU and using versoin 3.5 to test their own?
On Compute Clusters
Per job/thread/core bandwidth is also critically important.
IBM offers a single-core version of its dual-core POWER4 and POWER5 processors targeting the HPC market.
Clovertown has little value in HPC. Each core has half the bandwidth access as Woodcrest, and there is greater contention for resources. The clock speed is lower, and therefore the per thread performance is less.
In x86, dual-core rules in HPC. Quad core is a ways out. In RISC/EPIC, the competition was single-core POWER4/POWER5 to single-core Madison Itaniums. That could change now that Itanium is dual core. Regardless, expect for both IBM POWER5 and Montecito for HPC customers to shut off the multithreading.
You Dont't Understand Quiet Period
Quiet Period is more misinformation!
LOW BLOW BY AMD!!