Yep, except if it gets a wu with an odd number of packets then one GPU goes idle right at the end. Supposedly he is now issuing wu with an even number of packets for dual cards.Originally Posted by Maxwell
Yep, except if it gets a wu with an odd number of packets then one GPU goes idle right at the end. Supposedly he is now issuing wu with an even number of packets for dual cards.
Huh - I can't get it to keep a constant load on the 5970 without dedicating a core to it.
Same here. I've stopped processing MOO packets right now, as I need all 4 cores.
Ahhhh, then you should have asked the right question at the beginningI have a core dedicated to the GPU ... and I do mean CORE not THREAD! Each machine is running 2 PG wu on Cuda and 1 Moo wu on ATI. The 5970/570 computer uses 26% of the processor without any CPU crunching going on.
Heh... I've the same setup now, Bryan. Moo (5970) has one core on my AMDx6, DRTG (570) has another core. The remaining four are being crunched on.
This makes sense. Thanks, folks.
Dual 5830's......gave up due to looping/stuck work......had one on 89% for 8 hours.....grrr.
Kinda messes with your throughput, don't it! That sucks!
What method do you use to dedicate a full core to your moo task?Ahhhh, then you should have asked the right question at the beginning
There is a program called ProLasso (Project Lasso) that allows you to set default priority and affinity for any task. Maxwell uses this also and it is FREEBIE.
I set my Moo and PG wu to run on and only run on threads 6 and 7 (core 3 of my I7). I also raise their priorities to HIGH. Then I assign whatever I'm crunching CPU to threads 0-5. That way the GPUs have their own unobstructed core to use.
I HIGHLY recommend this program!!!!!!!!!!!!!!
Process Lasso. Don't let the thing fool you - you can use it for free. I'm loving this thing...
Posting Permissions
Reply With Quote
