Toms Hardware and 6600 GT Reviews

[mehmetmunur]

I have been reading a few GeForce 6600GT reviews for some time now. Recently I got a Leadtek 6600GT and I was putting it through its paces, overclocking, 3dMk05, Doom, Half Life, Far Cry etc. Then I decided to go online and check out some scores. Thus I visited tomshardware and looked at this particular review of the newer 6600GTs that came overclocked by the manufacturer. And the more I looked the more I was puzzled by this particular benchmark. All the scores were so close together, that it seemed as if something was the bottle in the test bed, and not the GPU. And when I looked at their configuration, I realized that they were using a P4 3.2, at which point I said to myself, DUH! Half Life is an extremely CPU intensive game, and even the fx-55 has issues in some sequences, how is the P4 3.2 to fare?!!

Then I looked at Anand's review of Half Life DX9 cards and the similar sequence gives completely different results with the same GPUs but different CPU, Athlon 64 4000+ in Anand's case.

?What were they thinking!? is the first thing that comes to mind. I do not do this for a living, but I had the sense to see that there was something wrong with the picture. How can they write a comparison about video cards, when the true potential of the video card is limited by the CPU. I know that my econometrics professor would have flunked me if I omitted something like that in my regression model.

I could be wrong but let me know what you think.

 

[Munky]

That's because like half the cards reviewed in the tom's benchmark were some version of a 6600(gt), so of course you'd expect them to perform similarly. Anyway, a real test of a video card is at 1280 resolution or higher, with AA/AF, thats where modern games can push these to the limit. At 1024 res without AA/AF, the cards dont get stressed as much, and so their performance is more cpu-limited.