Inturpretting Benchmarks for the KyroII

[Pederv]

AT News has a link to a KyroII review and on this page is the Q3 benchmarks. Down at the bottom of the page he says the the benchmarks are cpu limited. Well I don't see it that way and here's why:

I'm gonna use the 640 x 480 results but it works the same for all the tests.

933 - 800 = 133 ____ 125.5 - 114.9 = 10.6 ____ 133 / 10.6 = 12.55MHz per frame
1000 - 933 = 67 ____ 130.1 - 125.5 = 4.6 ____ 67 / 4.6 = 14.56MHz per frame
1200 - 1000 = 200 ____ 142.9 - 130.1 = 12.8 ____ 200 / 12.8 = 15.62MHz per frame
1333 - 1200 = 133 ____ 149.7 - 142.9 = 6.8 ____ 133 / 6.8 = 19.56MHz per frame

As the cpu speed increases it takes more clock cycles to get the score to increase by a single frame. In otherwords the harder you push the card the harder it pushes back. Look at the jump from 800MHz to 933MHz, it gives you 10.6 more frames per second. While the same 133MHz between 1200MHz and 1333MHz only gives you a 6.8 frames per second increase.

When using benchmarks to compare hardware, think it through a little.

 

[Pederv]

Just to give this a bump:
Anytime you increase how fast you process the data, the increase should be linear. Anytime it's not then what ever controls what your testing is the bottleneck, in this case the video card is the bottleneck.

 

[MrGrim]

They way I see it is that higher clock speeds will increase your framerates, but up to a point. There is a point where no matter how many more MHz (or GHz for that matter) you have the card will be working at it's max, hence no increase in performance. So as you are reaching that limit the increase in FPS will be decreasing.