AMD accuses Bapco of pro-Intel performance bias

ACCUSATIONS THAT Bapco SysMark 2002 is biased towards Intel's Pentium 4 chips and against AMD Athlon processors are flying around on the web like arrows at the Battle of Hastings.

A PDF from AMD hosted on Van's Hardware accuses Bapco of systematically favouring Intel processors over its own.

If AMD is the Norman horde in 1066, then Bapco is King Harold getting it in the eye, if the allegations in the PDF prove to be true.

The opening barb came with an article by Dean Kent on the Ace's Hardware Forum, where he said there is "something funny" in differences between SysMark 2002 versus SysMark 2001.

Kent said there: "This information tends to lend a lot of credence that the tasks are being targeted to a type of resource usage that would benefit the P4 platform, as has been alleged here by more than one person".

Dean Kent's own site is here.

Several weeks ago, AMD said it had joined Bapco, which at one time shared the same premises as Intel.

A follow up article at Van's Hardware reckons that Sysmark is no good for testing the Athlon XP processors, despite AMD joining Bapco recently. You can find that, as part of a review of an Athlon XP 2600+, here.

This article suggests that "calculated bias" has been put into the benchmarks.

It also suggests that AMD is seriously unhappy about Sysmark 2002, so unhappy that that's what it made it join the organisation recently.

And it proves that by hosting an AMD PDF on its site which details those concerns, and which you can download here.

AMD presented some results to Bapco, which prompted the organisation to produce a white paper but those results still proved unsatisfactory to the chip company, the PDF says.

It claimed that Sysmark favoured Intel tasks, inflated Intel figures, and removed elements that favoured AMD, across applications in different sets of tasks.

On page 12 of this slide, AMD claims that the 14 access tasks where the XP 2000+ had the biggest advantage over the 2.0A in Systmark 2001 were "completely removed" in Sysmark 2002.

It was such discrepancies which prompted AMD to join Bapco, the PDF said.

Another thread at Ace's expands the subject further.

The underlining theme appears to be that Sysmark 2002 is biased in favour of Intel processors.

Benchmarking, much to the astonishment of outsiders, attracts more controversy than practically any other PC related topic.

That's because vendors, hardware sites and chip manufacturers all, to some extent or other, rely on benchmarks for an assessment of performance.

We await a statement from Bapco on the accusations from AMD. µ

See Also
AMD joins Bapco