AMD pulls out of BAPCo in benchmark spat

Chip maker claims that latest benchmark suite favours Intel chips over its own

AMD has resigned from the BAPCo benchmark organisation and refused to endorse its latest SYSmark 2012 suite, claiming that the new benchmarks are unrepresentative of the workloads used in everyday computing and are biased in favour of Intel's chips.

BAPCo is a non-profit consortium set up to develop a set of objective performance benchmarks, which are widely used in the industry and technology press to evaluate PC performance.

However, AMD claims that SYSmark 2012, which was released earlier this month, does not base its scores on the kind of workloads that are likely to be encountered by users running modern web-based applications and other software that makes extensive use of graphics.

In a blog posting, AMD chief marketing officer Nigel Dessau said that only a small number of the benchmark operations in SYSmark 2012 actually contribute to the final score, and that a large proportion of the score is actually based on system performance during optical character recognition (OCR) and file compression tasks.

"The SYSmark benchmark is not only comprised of unrepresentative workloads (workloads that ignore the importance of heterogeneous computing and, frankly, favour our competitor's designs), but it actually generates misleading results that can lead to very poor purchasing decisions," Dessau wrote.

This is not the first time that AMD has claimed that benchmark tools unfairly base their performance metrics on operations that play to the strengths of Intel's chips rather than its own, but the company now feels so strongly about the matter that it has withdrawn its support from BAPCo entirely.

The move follows AMD's launch of it's A-Series chips last week, combining multiple x86 CPU cores with an on-chip GPU. This design not only gives a boost to graphics performance, according to AMD, but also allows applications to make use of the massively parallel processing capabilities of the GPU.

AMD's position, therefore, seems to be that its chips offer better performance for applications with graphical content or that can be accelerated with GPUs, including IE9, Adobe Flash, JavaScript and HTML5, and that SYSmark 2012 simply does not reflect that in its tests.

However, BAPCo told V3.co.uk that AMD had been heavily involved in the development of SYSmark 2012, and had voted in favour of the majority of the development milestones.

"BAPCo believes the performance measured in each of the six scenarios in SYSmark 2012 fairly reflects the performance that users will see," the organisation said in a statement.

Unconfirmed reports also claim that another chipmaker, VIA Technologies, has pulled out of BAPCo for similar reasons along with graphics firm Nvidia.

At the time of writing, VIA had not responded to requests for confirmation from V3.co.uk, while Nvidia declined to comment on the rumours.