I have not run SETI since before they started using GPUs to do work. Hence my question for the more experienced folks here:
I have several computers with two GPUs. But when I check them, often I find that only one, and sometimes none, of the GPUs are doing work, but the CPU is going full blast. Is this normal?
I'm (blissfully) ignorant of any Nvidia-drivers-problems, but as for either CPU or GPU running-out of work while other has work, this isn't something new when it comes to SETI@home.
Except for the VLAR-wu's that some Nvidia-cards atleast at the start crapped-out on, all wu's is device-independent, but then they're assigned to the client they're assigned to a specific device-type, either CPU, Nvidia or Ati/AMD. Meaning on client this is separate queues. Then there's problem downloading work from SETI, one of the client-queues running-out of work before the other queue is normal.
Since GPU is significantly faster than CPU, GPU will often run-out faster then problems downloading work, and since needs to download more just to keep busy it's not even certain manages to fill-up the queue even CPU is maxed-out.
For SETI@home the current limits are 400 per GPU and 50 per CPU-core.
If example GPU uses 5 minutes per task while CPU uses 2 hours, a maxed-out queue then SETI has another connection-problem so no new work can be downloaded, will lead to GPU running dry in less than 1.4 days while CPU can continue crunching for 4.1 days.