By Tom Sanders, 06 Jul 2007
AMD plans to remove two marketing slides from its website that provide a skewed comparison between its upcoming quad-core Barcelona processor and Intel's Xeon quad-core chips.
The slides compare a 2.66GHz Intel chip with a 2.66GHz AMD model based on benchmarks for floating point and integer calculations.
The graphic demonstrates that AMD would outperform its rival by 25 per cent on integer calculations and by 60 per cent on floating point calculations.
The benchmark data for the Intel processors is outdated, however, and AMD's benchmarks are based on estimates for a chip that is much faster than the 2GHz model that will start shipping in August.
Intel is also shipping a 3GHz quad-core Xeon, although this chip is primarily targeted at workstations.
AMD was ridiculed for the inaccurate information, especially because the company denounced Intel earlier this year for publishing skewed benchmark results.
The AMD web page does not list a publication date, but AMD spokesman Phil Hughes claimed that the information was posted early this year and was accurate at the time.
"We are working to remove that stuff from our website now. It is not an accurate reflection of the highest performance [Intel processors]," Hughes told vnunet.com.
The page was still online at the time of this story's posting.
The search for benchmark performance data on AMD's upcoming quad-core processor is fuelled in part by a lack of information from the chip vendor.
The company revealed shipping data and clock speed for the first Barcelona chips last week, but did not publish any performance data.
Combined with the relatively low clock speed, this sparked speculation that AMD might be unable to live up to its promise to take back the performance-per-watt lead from Intel.
Hughes promised that AMD would post performance benchmark data for its 2GHz chips "in the coming weeks", at which time it would also use the most recent benchmarks on Intel's chips.
The updated information should end the uncertainty surrounding AMD's performance claims.
When Nathan Brookwood, an analyst with Insight 64, scaled back the 2.66GHz data to 2GHz performance, he found that Intel would outperform AMD on the integer specification by about four per cent, while AMD's lead in floating point performance would shrink to roughly 17 per cent.
Although performance and clock speeds do not scale linearly, Brookwood's calculations cast further doubt on the merits of AMD's breakthrough processor claims.
Analyst firm Gartner questioned AMD's gains earlier this week, predicting that Opteron and Xeon quad-core performance will be close.