Effect of GPU Overclocking on PPD?

First of all hello everyone
I used to be on the old CPC forums a few years back but started university 2 years ago and the world of PCs became very distant to me as all I had to tinker with was my laptop Now however I have managed to get my rig I built back in 2008 through to uni and have started folding again I know my PPD is nowhere near everyone in the top ranks' but I'd like to do everything I can to maximise the points I can squeeze out of my current hardware!

Now I know people are saying "GPU folding is almost dead" but since I don't have and can't afford a more powerful CPU based setup at the moment my GPU is my biggest producer.

The GPU I'm running is a BFG 8800GT OC2

Core Clock: 675MHz
Memory Clock: 975MHz
Shader Clock: 1674MHz

Currently using the v7 client for the GPU (used to use the GPU2 client - have had no difference in tpf between the two clients)
and have no flags set up on this client.

I've got two main issues:

1) Most people seem to be averaging 4500-5000ppd with an 8800GT and I'm currently showing a 4200ppd with a TPF of 1m 38s. The only WUs I've been getting lately have been Project 10122 based with 494 points per unit.

2) I've been browsing the Folding forum for a while and couldn't find much about GPU overclocking and it's affect on PPD. Does one clock speed affect the GPU's folding capability more than the others?
Also I've tried overclocking the GPU further (checking the OC for stability first using 3DMark06) but any further overclock only has a detrimental effect on the TPF moving it to 1m 44s.

My main concern is that my PPD seems to be decidedly lower than other 8800GT users and was wondering if anyone had any idea how I can resolve this? (without buying new hardware before anyone tries to be smart )

If anyone wants/needs more information just let me know

Thanks in advance!

 

I believe shader speeds were important. Also make sure you turn the power save feature off in the Nvidia control panel. And watch them temps.

 

What I've read/been told is that the core clock and memory clock will have a relatively small effect on folding. As MrGumby rightly says it is definitely the shader clock that makes the most difference.

Unforunately, I don't own an 8800GT so I can't offer any kind of comparison. The only thing I can think that might effect your PPD is if you have the -advmethods flag turned on through the client or not?

 

As MrGumby says shader speed is the thing to go for.

I don't have any knowledge of 8 series GPUs but with the 2 series you can change the shader independently of the core (in step changes of 54).

If that is the case with the 8 series, I would leave the memory and core at stock and boost the shader as far as possible using MSI Afterburner or something similar.

 

Thanks guys (and girls...never know)

The Shader Clock can be raised independently of the Core Clock and by the 54MHz increments you stated One_Box so I've raised the shader clock from 1674MHz ---> 1890MHz after finding this to be the maximum clock without artifacts appearing while running a FurMark stability test. The max temp was 82C which is the same as at stock settings.

As a result my TPF is now a more palatable 1m 32s average and holds steady at 64C when folding which is only one degree more than it was previously. So I'm now sitting on around 4700ppd, but then this happened....

So it managed around 8% of the unit and then shutdown....then it started a new unit....

The new unit is currently 36% done and having no problems, the temp is still sitting at 64 degrees.

I was thinking of seeing how this unit progresses and then if the same thing happens dialing down the shader clock to see if that helps at all. Any thoughts?

 

Back off the shader OC a touch, not worth leaving it at a value where UNSTABLE_MACHINE errors occur.

 

Yeah this unit got the same error after 86%.

Have backed the shaders down to 1836MHz so will see how that holds up then if need be lower it some more.

Thanks for the help again
Though by the looks of things it's going to be a day without points for me today while I experiment...

 

I expect 1836 will be OK, - you will generate more points in the long run by not pushing it too far - good luck.

 

Thanks, it's producing a TPF of 1m 29s now on big units and 1m 15s on smaller units so if I was to run it 24/7 I'd be looking at a lowly 5000ppd now 1836MHz would appear to be the magic number for this little old GPU. In the end of the day it's all about the science though

Thanks so much for your advice again One_Box, MrGrumby and BonaSwirl. Much appreciated