AMD is losing its patience with arch-rival Intel and venting its irritation over what it considers a lop-sided benchmark comparison.
During this week’s launch event for the 690 chipset, Henri Richard, executive vice president and chief sales and marketing officer for AMD
, vented a little spleen.
“I think we’ve been too quiet. I’m sick and tired of being pushed around by a competitor that doesn’t value competition,” he told the assembled reporters.
He was grumbling about an Intel benchmark from August, 2006, that compared a 2.8Ghz Opteron 2220 to the 3Ghz Xeon 5160 dual core Xeon, using the older SPECint_rate2000 benchmark. In that benchmark, Intel had a 57 percent performance advantage over AMD.
But with the SPECint_rate2006 benchmark, that AMD used in January, 2007, the two came out even, with a very slight edge to Opteron. In discussing the issue after, Richard said Intel used a “single core” processor, but the Opteron 2220 is dual core.
“There is not even a debate of whether the competition has a good part. We think they have a good part. But why try to masquerade things? All I want is a level playing field where we’re talking the truth,” said Richard.
The truth, he claimed, was that AMD’s upcoming Barcelona processor will run at least 42 percent faster than the Xeon 5355, Intel’s current top of the line. But those benchmarks could be as dubious as Intel’s, one analyst noted.
“You’ve got to figure that no one will put out benchmarks that don’t favor them,” noted Rob Enderle, principle analyst with The Enderle Group. He also added that in this case, Richard did have a legitimate complaint against Intel’s lopsided benchmarking.
Intel spokesman George Alfs defended the benchmarks. “The Opteron number we quoted is the highest scoring public Opteron result we could find on any SPECintCPU benchmark. We’ve seen their vague complaints, but they aren’t offering higher scores that have been listed on the SPEC Web site,” he said in an e-mailed statement to internetnews.com.
The problem, said Enderle, is that a number of well-known technology testing labs have fallen by the wayside, leaving testing to the vendors themselves. “Lo and behold, they’re biased,” he said.
Enderle thinks benchmarks are only an issue for very few people. “Outside of scientific calculations, who looks at things like SPECint?” he said.
Richard seemed to echo this, saying “I have never met a CIO who cared about the nanometers of his processors. They don’t want to know, they don’t care. They just want to get the job done,” he said.
A Preview of Future Processors
Cat fights over benchmarks aside, AMD did discuss other news aside from the 690 chipset, which was the big news of the day. It announced the Hawk family of mobile processors would ship in the second quarter of this year instead of Q3 as it initially planned.
The next generation Griffin architecture, which would replace the Turion family out now, would be available in late ’07 or early ’08. AMD is also planning on a quad core desktop, codenamed Agena, for the second half of this year.
The company also demonstrated it knows a thing or two about teraflop computing. While Intel made a teraflop chip that was purely designed as a lab experiment, albeit one that will help with future projects, AMD did its own demonstration with soon-to-be shipping products.
The demonstration involved an early version of a forthcoming Barcelona CPU and two R600 graphics cards, which are massive floating point accelerators. Using a custom calculation program, the three chips ran in parallel to produce over a teraflop of computation, or 100 billion calculations per second. However, AMD admitted practical applications for the new technology are still considered to be years away.