My recent article on Gainward's CoolFX Powerpack! Ultra/2600 graphics card sparked some discussion about the choices I made when benchmarking, regarding the filtering optimisations in the NVIDIA driver used to benchmark the card. I'd elected to turn them off, while leaving optimisations on in the ATI driver used to benchmark an X800 PRO, the card I chose to compare the Gainward to.

With NVIDIA's image quality with the optimisations on being subjectively very good on GeForce 6800, it would have made more sense to leave them on, or at least benchmark both sides of that particular coin. So I've returned to the article and added in an extra series to each benchmark graph, indicating performance at 450/1200, the card's default clock speeds with ExperTool, with default Quality optimisations enabled.

It turns out that in most cases, the optimisations allow the card to run just as fast at 450/1200 with optimisations on, as the card does at 470/1200 with optimisations off. Therefore in some tests, should I have benchmarked at 470/1200 with default optimisations on, the Gainward would have appeared even faster than the X800 PRO than it already did. However, the card was moved on by Gainward to another website before I had the chance to run additional 470/1200 numbers, with optimisations enabled.

The extra 450/1200 numbers should give you a rough idea of what to expect at 470/1200 with optimisations on, using the CoolFX. Its performance differential compared to other air-cooled cards doesn't change however, since all 6800 boards are free to use the optimisations too. Gainward were unhappy that I claimed the performance differential was too small to recommend the CoolFX, and even with optimisations on my opinion (note, it's my personal opinion and I mark it as such in the conclusion) doesn't change.

However, it's clearly the fastest consumer graphics card currently available and as such it deserves some credit, moreso than I maybe give it. I'm unswayed due to its price, but some people have very deep pockets.

My apologies to Gainward for the initial filtering choices, the card performance looks a fair bit different using the good driver optimisations.

Check out the article with updated graphs and commentary here.

Rys