Can overclocking the GPU cause decreased performance?

[Twizted_3kgt]

I ask because it seems the more my 660 Ti downclocks, the higher benchmark score I get.

Here's a run at the maximum stable overclock I could get 1280/3206:

P4937 3DMarks

Graphics score was 5794.


And here's the run that downclocked to 705/3206:

P5123 3DMarks

Graphics score was 6036.



I have 2 other "inbetween" clock speed runs that follow the same trend. I don't get it, what gives?

 

[PP Mguire]

Could be not enough voltage, overheating causing throttling, or CPU bottleneck.

 

[Twizted_3kgt]

Could be not enough voltage, overheating causing throttling, or CPU bottleneck.

I think the voltage was excessive, 1.175v which is maxed and not even necessary to keep it stable. The temps didn't get above 65C. So that leaves the CPU, which is old and probably the problem. Hadn't thought of that.

Ill have to find a comparison 660ti run on a better CPU.

 

[PP Mguire]

Another thing is it could not be getting enough power. When I was cranking the clocks on my 580s and the lights started flickering I noticed I was pulling over 1000w from the wall. Meaning I was exceeding overload on this 850w and the higher I clocked the more it crashed and the lower my score was.

Just as a comparison, an i5 750 and GTX 580 both overclocked get 7k. A 660ti is faster than a 580 so your hold back is for sure the CPU in points. In the reducing points it's still one of the things I mentioned.

 

[Twizted_3kgt]

Another thing is it could not be getting enough power. When I was cranking the clocks on my 580s and the lights started flickering I noticed I was pulling over 1000w from the wall. Meaning I was exceeding overload on this 850w and the higher I clocked the more it crashed and the lower my score was.

Just as a comparison, an i5 750 and GTX 580 both overclocked get 7k. A 660ti is faster than a 580 so your hold back is for sure the CPU in points. In the reducing points it's still one of the things I mentioned.

Looks like I would need some hardware to measure wattage, what do you use? Kill-a-watt?

I'm sure the CPU is bottlenecking my score. It was when I first got the 560 Ti a year ago, so probably moreso with the 660 Ti. But so far I haven't seen any difference between benchmarked FPS in games versus what I'm getting, which is really my main concern and the reason I haven't upgraded CPU/mobo.

I've just never seen an overclock actually lower a score/fps.

 

[PP Mguire]

Yea I use a Kill-A-Watt.

Yea you're right there, upgrading the CPU will only make a small difference in games and a huge difference in benchmarks. Not exactly worth it.

That's why I think there's something up. Instead of going down it should only make it slightly go up. Are you sure your load temp is only 65? Have you tested this with something like Furmark?

 

[Twizted_3kgt]

Yea I use a Kill-A-Watt.

Yea you're right there, upgrading the CPU will only make a small difference in games and a huge difference in benchmarks. Not exactly worth it.

That's why I think there's something up. Instead of going down it should only make it slightly go up. Are you sure your load temp is only 65? Have you tested this with something like Furmark?

Well I've been using the stress test and GPU burn-in from Kombustor 2.3 to see where the max GPU temp gets to. 65C is the highest I've seen it go since rerouting my wires and eliminating fan controllers (hard wired to 100%). But it takes forever for it to get that high, and that was overclocked as high as it would go with the voltage as high as it would go.

Without the overclock/overvolt it levels off at 61-62C. In BF3 and modded Skyrim it stays around 50C with a max of 60C. Idles at 30C.

I did have a 2-3 max fps increase when overclocked in the stress tests though. So maybe its just 3DMark? Wouldn't really make sense though.

Edit: So I tried Furmark just to be sure on the temps since you mentioned it, same temp 61-62C is where it levels off.

I also did no overclock vs overclock in Furmark, and I gain 1fps overclocked (1254/3206) vs not overclocked. Maybe it is just 3DMark?

 

[PP Mguire]

Well like I said stability could be an issue. 3dmark will drop score for various reasons so there is really no way to tell. If the overclock isn't stable enough with the supplied voltages it could drop score.

Also, are you raising boost clock or standard clock?

 

[Twizted_3kgt]

Well like I said stability could be an issue. 3dmark will drop score for various reasons so there is really no way to tell. If the overclock isn't stable enough with the supplied voltages it could drop score.

Also, are you raising boost clock or standard clock?

Boost clock and then forcing it to run boost clock constant. I don't think it lets me raise standard clock.

 

[PP Mguire]

That could be it. Boost clock too high could be unstable forcing regular clock speed or lower boost.