hello there folks

as i ve been seeing, the top computers all crunch at least 2 wus at a time.

i m with you guys so far, 2 work units lower the time of the gpu idling in between work units, when the cpu is processing the last bit. but here comes the thing. even with 2 work units simultaneously, i have a lot of gpu idling, because the modfit separation units need some 4-5 seconds cpu time in the end.
even with 3, i still get gpu usage of only 50% sometimes, like every other 10 minutes.
the times from the top computers tell me, however, that most of you crunch 2 at the same time.
have i missed sth in my app config?
i cannot see a difference in changing cpu usage from 0.05, 0.25, 0.5 or 1.0 either.

my question is: while knowing that for chips, there is no such thing as a wearout (only due to too high temperatures, afaik), i wonder if it is of any significance for the GPU itself whether it calculates a single work unit or several (up to 8 or whatever) at the same time.
because if it is not, then why not run 8 at the same time, which basically guarantees that the gpu always is at full load. (temperatures wouldnt be a problem, stable ~64° here with milkyway running @1180mhz [r9 280x])

regards

I am currently also using a 7970 to run units, but only 1 at a time and am seeing this:
15.35 5.77 26.74 Milkyway@Home Separation (Modified Fit) v1.36 (opencl_ati_101)

and this:
33.37 3.29 106.88 MilkyWay@Home v1.02 (opencl_amd_ati)

You said you are running 2 units at a time and are seeing this:
21.07 4.78 26.74 Milkyway@Home Separation (Modified Fit) v1.36 (opencl_ati_101)

and this:
32.20 2.45 106.88 MilkyWay@Home v1.02 (opencl_amd_ati)

I don't see where you are doing the units any faster or slower overall than I am, both examples are just examples and not a result of adding and dividing etc.

My pc is a:
GenuineIntel
Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz [Family 6 Model 23 Stepping 10]
(4 processors) and I leave one core totally free just for the gpu to use.

Your pc is a:
AuthenticAMD
AMD Phenom(tm) II X4 940 Processor [Family 16 Model 4 Stepping 2]
(4 processors) do you also leave a core totally free for the gpu, or do you just use the app config file to control your cpu usage?

To be honest with you, i dont quite know how to "leave a cpu free" for the gpu.
for the time being, all i have set cpu-wise is the cpu_usage in the app_config.xml.Do you mean the option in the boinc manager where i can say use maximum ...% of cpu? in this case, 75% then with 4 cores?

they definitely are faster when running 2 at once. because the gpu doesnt go idle or low load % after each work unit.

what clock are you going with? mine is currently 1180MHz, i think the standard 7970 clock is lower?

and this:
32.20 2.45 106.88 MilkyWay@Home v1.02 (opencl_amd_ati)

this one must have been a workunit that came after a modfit separation wu, they usually take 37 seconds when running 2 wus at a time. but because the modfit takes about 5seconds of cpu time, the ones following go faster than usual, thereby avoiding or lowering gpu idling time

See this thread. http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2301

Edit: Oh wait, sorry. You're talking about GPU.... Link is for CPU above.

To be honest with you, i dont quite know how to "leave a cpu free" for the gpu.
for the time being, all i have set cpu-wise is the cpu_usage in the app_config.xml.Do you mean the option in the boinc manager where i can say use maximum ...% of cpu? in this case, 75% then with 4 cores?

No the line above that that says "on multiprocessor systems use at most [] % of cpu, zero mean no restrictions". Change that to 99% and your pc will leave one cpu core free for whatever else the pc wants to do with it, open webpages, process documents, feed the gpu etc.

they definitely are faster when running 2 at once. because the gpu doesnt go idle or low load % after each work unit.

what clock are you going with? mine is currently 1180MHz, i think the standard 7970 clock is lower?

Yes mine is slower at 955 core and 1375 memory speeds as reported by gpu/z.

and this:
32.20 2.45 106.88 MilkyWay@Home v1.02 (opencl_amd_ati)

this one must have been a workunit that came after a modfit separation wu, they usually take 37 seconds when running 2 wus at a time. but because the modfit takes about 5seconds of cpu time, the ones following go faster than usual, thereby avoiding or lowering gpu idling time

If you are happy running it the way it is, then stay with it, Boinc is all about sticking with it for the long term, being happy with the way your pc runs is key to that. I am NOT saying what I do is better, I am just saying that what I do works for me.