GT200/RV770: clock scaling

KonKort

Newcomer
Hello,

I can present you a nice test, which is about the scaling of the clock domains of GT200 and RV770.
We increase the core clock, (shader clock) and memory clock and look which clock domain increase the performance in current games in best.

The results are very interesting. First they show that GT200 and RV770 are not shader limited in current games. Second they show that increasing core clock frequeny by RV770 is as good as increasing memory clock frequency. So it looks like RV770 is a little bit memory limited. GT200 on the other hand did not perform well with increasing memory clock.

My English is not very good and the following article is in German. But many diagrams are included. Most interesting are the performance ratings, who contains all results. (Link)

GT200/RV770: clock scaling

Have fun by reading. Every comment is welcome.
 
I'm not sure how you can come to the conclusion that these chips are being "bottle-necked" or "limited" when you were simply underclocking the chips.:???:
 
Last edited by a moderator:
Why did you use such low settings in Warhead? I'd really like to know what these charts would look like at Enthusiast.
 
I'm not sure how you can come to the conclusion that these chips are being "bottle-necked" or "limited" when you were simply underclocking the chips.:???:

20% higher shader clocks results in only 7% higher performance, 20% higher core clocks results in 10% higher perfomance and 20% higher memory clocks results in 3-4% higher performance (GT200).

20% higher core clocks results in 10% higher performance and 20% higher memory clocks results in 8-9% higher performance (RV770)

Why did you use such low settings in Warhead? I'd really like to know what these charts would look like at Enthusiast.

The settings are always the same: 1280x1024, 4xAA/16xAF and 1920x1200, 4xAA/16xAF. But as you can see it is not important which settings are used (the difference between the two settings are the same)
We do not test higher settings because of VRAM size.
 
The settings are always the same: 1280x1024, 4xAA/16xAF and 1920x1200, 4xAA/16xAF. But as you can see it is not important which settings are used (the difference between the two settings are the same)
We do not test higher settings because of VRAM size.

:oops: I must have been looking at the wrong chart; it appears you did test on Enthusiast settings. Sorry.

So Warhead is mostly bandwidth limited on GT200. Interesting. I'm gonna overclock the bejeezus out of the GDDR3 on my GTX260 and see what happens. :smile:
 
Back
Top