Gtx 570= 2:16

Core 795 Shader= 1590 Memory= 1905 CPU=i5 2500k @4.3 Latest Nvidia drivers installed. W7 64-bit Current WCG boinc

With one core for GPU only.

Besides crashing on bf3 on occasion I haven't noticed any. And I don't let the monitor sleep or.rig for that matter. I've heard its.bad.but.until.I.see serious problems I'm just going to let it be up to date.

Thanks for the post though.

I wonder, since the 7970 was apparently overclocked (most stock 7970's run at 1000mhz, correct me if I'm wrong), if this could have resulted in an invalid result?

Perhaps overclocking to 1125Mhz is a possible reason for invalids, although that's quite a mild overclock for a HD 7970, wouldn't even require overvolting, probably just need to push the power slider to 20%. Others have run these at 1240 Mhz without producing errors at MilkyWay which is a very demanding GPU application. This application is unlikely to cause overheating problems on a HD 7970 as it runs for less than 14 seconds every minute.

The earlier task with the shorter runtime was valid at the same frequency. With such a large proportion of the runtime of HD 7970 processed tasks on the CPU, if considering overclocking as a cause of invalids then CPU overclocking is also a possible cause.

The throttled GPU causing stalling and erratic progress of GPU processing was just something I considered as a potential cause of invalid results. Stalled/throttled GPU processing was experienced by some with POEM OpenCL GPU application as that also has a default ncpus value of less than 1 which causes CPU overcommitment when all cores are loaded with CPU tasks.

The first 3 tasks I ran here on my HD 5870 were all marked Invalid. I was running with an underclocked memory speed of 900MHz instead of the default 1250MHz. When I increased the memory to the default and overclocked the core from default 870Mhz to 925MHz the last batch of 24 tasks were all valid. So my experience with this card on this application tells me to beware of any memory underclocking rather than mild core overclocking. Hard to deduce anything though other than suppositions from such a small number of completed tasks and the validator was acting ornery on the first batches too with a large number of replications being marked Invalid or Too Late.

The application on fast GPUs is partly CPU bound and also runs only on the CPU for about 50% or more of the runtime, so the GPU core speed makes little difference to total runtime. On a CPU bound GPU application as the GPU speed increases so does the throttling and GPU kernel time remains about the same. With the current application on fast GPUs, most important for efficiency is allocation of sufficient CPU resources, second is CPU speed and GPU core/memory clock is last. On slower GPUs that are not as CPU bound, GPU core/memory speed would have more effect on runtime.

Awesome time on that 7950. Comparing your time with my 6950, your total time is longer but faster in the other times. Did you leave a thread free or are all threads crunching? You can see the first post here to look at mine. Would have thought your total time would be allot less time. If you did not have a thread free, can assume it was sharing CPU time with another active wu.