SPCR

A few days ago I noticed that my SMP client only hit a checkpoint every couple of hours rather than the 17-20 minutes it used to. That made me look at how much CPU it was using, which turned out only to be one core. This seems to be a known issue with some WUs, that's being worked on. A workaround (since if it only uses one core, it probably won't meet the deadline anyway) is to remove the core, there's a new one released that'll discard these WUs when it gets them.

Also, SMP is really nice compared to the old uniprocessor client (I got a dual core back in May, which brought me from ~100ppd on an Athlon XP to ~1500ppd on an Athlon 4850e.

So when you say 'remove the core', what do you mean, like kill the job, wipe out the work dir and the queue.dat file and start it up again? Checking my throughput, it looks like I may have a couple of these bad WUs.

Hey, I got my jobs fixed, now things are back to normal, so thanks for the tip!

cordis wrote:So when you say 'remove the core', what do you mean, like kill the job, wipe out the work dir and the queue.dat file and start it up again? Checking my throughput, it looks like I may have a couple of these bad WUs.

Plus delete FAHcore_A1 or A2 to get the updated one.