GPU client and desktop performance issues

[hiigaran]

hmm, ill give that a shot. my WU should be done in ab hour or so, but ill first try it with one unit, and see if that does anything. ill report back in a minute or two...

EDIT: nope, same problem. maybe it might work once both units are out of the program files folder, but right now it doesnt look promising.

[Amaruk]

If you are using an edited shortcut to start the client, you could try running it as Administrator. (right-click the shortcut > select Run as administrator)

[hiigaran]

nope. same problem

[Amaruk]

Last idea I have. Try creating a new directory (For example, C:\Users\[username]\FAH\GPU1), copy the entire contents of the current client (program files/FAH/GPU1?) and try to run it from there.

For reference, the contents of the GPU client's folder should be similar to this.

[hiigaran]

same problem.

[bruce]

The default location for the FAH client is the Program files directory and as long as you put the data somewhere else you won't have a problem. That's the default for the Windows installer and it works just fine.o 7/vista/XP without any special security tweaks.

[Amaruk]

The default location for the FAH client is the Program files directory and as long as you put the data somewhere else you won't have a problem.

True, unless hiigaran is using the console client.

hiigaran, what version of the client are you using?

If it is the Console client, at this point I would go ahead and do a fresh install as outlined in the guide using something like C:\Users\[username]\FAH\GPU1 & C:\Users\[username]\FAH\GPU2 for directories.

If these new clients work then simply remove the old ones. If they don't then there is some other issue besides the clients location.

[hiigaran]

its alright, i got it fixed. i dont know how, but i just did. i ended up completely uninstalling all the GPU clients then downloading them again. looks like its working now. my GPUs are chugging away nicely, and i have that annoying high pitched noise from the PSU that always comes on with high loads. thanks for the help folks!

[hiigaran]

one more thing - might as well ask here as its a quick question - regarding the core, shader, and memory clocks, which ones affect folding performance? obviously the core clock affects it, but what about the others?

[Jesse_V]

one more thing - might as well ask here as its a quick question - regarding the core, shader, and memory clocks, which ones affect folding performance? obviously the core clock affects it, but what about the others?

I haven't done things like overclocking, but did a Google search and found page: http://mytechencounters.wordpress.com/2 ... phic-card/ which seemed to explain that cores and shaders were the most important.

[bruce]

On many GPUs, the core and shader clocks are interlocked. When you overclock one of them, the other changes, too.

FAH performance is tirectly affected by core/shader clocks. Memory clocks are MUCH less important. Increasing the memory clock creates extra heat and almost no increase in FAH speed. In fact some donors have found that by reducing the memory clocks, the card runs enough cooler that they can sometimes increase the shader clocks, resulting in a net increase in FAH speed. What works for them might not work for you, though so you'll have to experiment. As with all overclocking, Stanford takes no official position

[hiigaran]

pretty much confirmed my suspicions. im already as far as i can go without instability setting in, and its too hot to up the voltage, so ill keep it as it is then. thanks.

[hiigaran]

so...ideas on why one of my GPU clients is giving me unstable machine EUEs? ive already downclocked the video card, and its only on one GPU, not the other. a few days back, i lost 80% of a work unit, half an hour ago, another 30%, and just now, after 2%.

thoughts?

[bruce]

Possiblites:

A) The GPU has a hardware problem (or is excessively overclocked, but you probably covered that one)
B) That GPU isn't being cooled adequately
C) You've been assigned a "bad WU" (Not applicable if it has been several different WUs. Check the PRCGs).

Just to rule out something strange in your configuration, modify both of your shortcuts to interchange 0 and 1. That way -gpu 0 will be started on the WUs that were previously running on -gpu 1 and vice versa.

Have you tried -gpu 0 plus -gpu 2 instead of -gpu 1? Some folks have reported the numbers as applying to each monitor connection, not each discrete GPU.

[hiigaran]

lets see if i can address each point...

A: not sure how i would be able to confirm that, other than through benchmarks and artifact scanning...both of which have come up clean.
B: average GPU temperatures are 90 on GPU 1 and 85 on GPU 2. GPU 2 is the problematic client, and in any case, the 590 is specced with a max temp. of 97 degrees (and manufacturers generally assign the lowest safe value anyway). in any case, the temperatures have gone up to 94 degrees before.
C: ill check them next time round. are there specific 'bad' units, or...how do i know?

ill also try those flags when i get the problem again. so far its been behaving itself (and i certainly dont want it to crap out on me when its 91%!).

what i have noticed, which could just be coincidence, is that the usually happen about 10-20 minutes after turning on my computer and opening the clients. not sure if that means anything though.