Hi Oliver, Is it also a problem with 1 CUDA device, or just when 2 GPUs are installed? If it's only a problem with 2 you could try setting avg_ncpus = 0.5 or 0.55 which should reserve 1 CPU core for 2 CUDA processes. Maybe not ideal but better than the current process contention.
Regards/Ed On Mon, Jan 3, 2011 at 12:22 AM, David Anderson <[email protected]> wrote: > Oliver: > I'm not sure there's anything we can do about this, > other than for you to increase avg_ncpus to 1. > > As you know, GPU jobs are run at normal process priority, > so in principle they should preempt CPU jobs, which run at idle priority. > However, apparently the Windows CPU scheduler doesn't do this. > > What's needed is for Microsoft and NVIDIA engineers to see what's going on > and fix it. This is not likely to happen since neither of these companies > has shown much interest in volunteer computing. > > -- David > > On 20-Dec-2010 3:12 AM, Oliver Bock wrote: >> Hi everyone, >> >> We just deployed a new CUDA application (called BRP3) as part of the >> einst...@home project. This app roughly up to 75% of a GPU and 3-30% of >> a CPU, depending on the GPU model/performance. Thus our scheduler >> currently issues these tasks with the following settings: >> >> hu.avg_ncpus = 0.2 >> hu.ncudas = 1 >> >> Please note that BOINC (e.g. sched/sched_customize) revision 22832 is >> used in this case. >> >> The problem is that with the settings above BOINC starts CUDA tasks in >> addition to CPU tasks that already occupy all existing CPU cores. This >> means on a system having four CPU cores and two CUDA devices, four CPU >> tasks and two CUDA tasks are launched. Although this behavior is >> intended, it doesn't really work out for us because the performance of >> the CUDA tasks is degraded significantly - GPU usage goes down to less >> than 10%, increasing the runtime by the same factor. Although the CUDA >> tasks run with slightly higher priority (below normal on Windows) than >> the CPU tasks (low on Windows) they are limited by the already >> fully-occupied CPU cores which are still required for up to 30% of the >> computation. >> >> Since we couldn't yet release a Linux or Mac OS version we don't know >> whether this is a Windows time-slicing issue or not. Are there any other >> projects running CUDA tasks in a comparable way? >> >> The only workaround in sight would be to acquire a full CPU core once >> again but that's certainly not ideal. >> >> Any ideas are welcome! >> >> >> Cheers, >> Oliver > _______________________________________________ > boinc_dev mailing list > [email protected] > http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev > To unsubscribe, visit the above URL and > (near bottom of page) enter your email address. > _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
