GTX285 Bottleneck investigation

[thilanliyan]

http://alienbabeltech.com/main/?p=9368

 

[vj8usa]

That's weird, I keep getting a "406: not acceptable" message every time I try to go to the ABT site.

 

[thilanliyan]

Works fine for me in IE8.

 

[ArchAngel777]

Nice write-up BFG. I am surprised by the results. I expected the shader performance to drop more and what surprised me even more was that the memory bandwidth took second place (with AA) and coming pretty close to the core performance drop-offs.

 

[McRhea]

Isn't it safe to assume that the Core clock will have the biggest impact to frame rates when overclocking? I would think it's common sense that you would want to get the core clock as high as possible, and then push the shader clock followed by the memory clock to get maximum frames.

That was my experience when OC'ing my 4890... pushing the Core as high as it would go (925/1100) gave me better frames (very slight, but still better by 3 frames avg) than pushing the memory clock higher than my core (900/1200). I didn't touch the voltages on my 4890, and I'm using the stock fan on the card.

Here's two articles that concur with BFG's findings:
AT's 4890 OC article

AT's GTX 275 OC article

 

[BFG10K]

Originally posted by: McRhea

Isn't it safe to assume that the Core clock will have the biggest impact to frame rates when overclocking?

The funny thing is, when I did these tests on a GTX260+, the three clocks were quite even in terms of impact:

http://forums.anandtech.com/me...id=31&threadid=2287281

Also the memory on the GTX285 had an even bigger impact than it did on the 8800 Ultra.

I think the selection of games plays a large part in the outcome of the tests.

 

[dguy6789]

Extremely interesting results. I love it when benchmarks are done to answer some of the more elusive questions about hardware.

 

[AzN]

Originally posted by: BFG10K

Originally posted by: McRhea

Isn't it safe to assume that the Core clock will have the biggest impact to frame rates when overclocking?

The funny thing is, when I did these tests on a GTX260+, the three clocks were quite even in terms of impact:

http://forums.anandtech.com/me...id=31&threadid=2287281

Also the memory on the GTX285 had an even bigger impact than it did on the 8800 Ultra.

I think the selection of games plays a large part in the outcome of the tests.

So what you are saying is that on different cards it behaves differently after testing diff. cards.