The market for workstation class cards differ from the gaming market in several ways.
In our current context, the most relevant is that the set of benchmarks used change much less over time. It follows that the kind of benchmark manipulation that has now been exposed for 3DMark is much more likely to remain profitable for a longer time. Furthermore, the group of benchmarks used is narrower, making it easier to avoid suspicion if the effort is spread out over the quite small set of standard benchmarks.
I've never been comfortable with some of the improvements that have been seen between driver revisions in the workstation space, nor the huge performance differential shown in workstation benchmarks between "gamer" and "pro" drivers for the same hardware.
I have a feeling that a close scrutiny of how the workstation drivers behave for the standard benchmarks could yield very interesting results. The scientific/engineering community has absolutely no interest in being decieved by IHVs regarding performance expectations, and the removal of foul play(ers) would greatly benefit the community.
I'm not really part of the graphics end of the scientific community any more other than as an occasional user, so if there are people reading this who are better positioned to take action in this matter, I urge you to start doing so.
The end result can only benefit us all.
Entropy
In our current context, the most relevant is that the set of benchmarks used change much less over time. It follows that the kind of benchmark manipulation that has now been exposed for 3DMark is much more likely to remain profitable for a longer time. Furthermore, the group of benchmarks used is narrower, making it easier to avoid suspicion if the effort is spread out over the quite small set of standard benchmarks.
I've never been comfortable with some of the improvements that have been seen between driver revisions in the workstation space, nor the huge performance differential shown in workstation benchmarks between "gamer" and "pro" drivers for the same hardware.
I have a feeling that a close scrutiny of how the workstation drivers behave for the standard benchmarks could yield very interesting results. The scientific/engineering community has absolutely no interest in being decieved by IHVs regarding performance expectations, and the removal of foul play(ers) would greatly benefit the community.
I'm not really part of the graphics end of the scientific community any more other than as an occasional user, so if there are people reading this who are better positioned to take action in this matter, I urge you to start doing so.
The end result can only benefit us all.
Entropy
