Soldato
... Crysis warhead is 87% for nvidia, not 76. I haven't checked any of the others.
Single-instance minimums are not exactly the best indication of overall performance anyway. Too many other factors can influence instantaneous minimums (CPU limitation, memory bandwidth limitaton, DRAM to VRAM paging etc). It's unrealistic to expect them to scale linearly with the average framerate.
If you want to play at statistics, you could look at the relative increase in minimum framerate going from single to multi-GPU. That's what the review is designed to investigate after all, and shows the nvidia cards in a much better light (they come out on top in every test). But still, instantaneous minimum framerates are too unpredictable to be a very useful measure.
The entire variation over the benchmark is given for each game though, so the true performance is easy to see from those plots.
I agree that it would have been good to include 1920*1200 as well, though clearly that.
Single-instance minimums are not exactly the best indication of overall performance anyway. Too many other factors can influence instantaneous minimums (CPU limitation, memory bandwidth limitaton, DRAM to VRAM paging etc). It's unrealistic to expect them to scale linearly with the average framerate.
If you want to play at statistics, you could look at the relative increase in minimum framerate going from single to multi-GPU. That's what the review is designed to investigate after all, and shows the nvidia cards in a much better light (they come out on top in every test). But still, instantaneous minimum framerates are too unpredictable to be a very useful measure.
The entire variation over the benchmark is given for each game though, so the true performance is easy to see from those plots.
I agree that it would have been good to include 1920*1200 as well, though clearly that.
Last edited: