Bottlenecks: A 5-Series Trifire CPU Scaling Investigation

First! J/k Anyhow I had to read that twice to take it all in. This leaves me with the "Now what?" feeling.
 
Last edited:
Two things really come out of these runs.

A) Somethings that we consider 'gpu benchmarks' aren't really benching the GPU at all.

B) Never take these review sites seriously that show, for example, three GTX 285s hanging in performance with three 5870s. There is no comparison, the CPU bottlenecks in many/most situations are making them seem closer in performance. For example, there's a article with three 285's making a similar claim as mine except they were getting ~75FPS in Crysis at 4.2GHz (almost exactly what I get)... except they could only get that with 0xAA. That's a massive difference in performance.
 
Ihatethedukes said:
Two things really come out of these runs.

A) Somethings that we consider 'gpu benchmarks' aren't really benching the GPU at all.

B) Never take these review sites seriously that show, for example, three GTX 285s hanging in performance with three 5870s. There is no comparison, the CPU bottlenecks in many/most situations are making them seem closer in performance. For example, there's a article with three 285's making a similar claim as mine except they were getting ~75FPS in Crysis at 4.2GHz (almost exactly what I get)... except they could only get that with 0xAA. That's a massive difference in performance.
Click to expand...
It's true. What your experiments show, though, is that when comparing 3dmark bechmarks of various video cards, even the slightest score difference still complies to a high difference in GPU performance.
 
You must log in or register to reply here.
Back
Top