Pardon me if this has been asked before, but I think I read somewhere that sometimes the combination on an AMD CPU and GPU crunching can cause low GPU utilization? Basically I think I am seeing that and on one of my machines, 2 nvidia cards, 6 core AMD chip and 1 or more astropulse WU will just stop using the GPU until exit boinc and restart it. If I don't they will run for many many hours longer than they should. It all started oddly when I added the 750ti with the 560ti that has been running fine for that past 2 years. That said, it now happens with either the 750 or the 560 with no immediately obvious cause. I've tried limiting the CPU usage but that doesn't seem to have corrected the problem thus far. Just curious if anyone else has seen anything similar or if I'm missing something obvious. I should also note that this doesn't happen with MB wus...

Thanks,

Chris

---

Chris,

Are you running any work units on the AMD chip? I only ask as I had to stop using my 4 core AMD chip with my Nvidias as they weren't finishing in time and we going to high priority to make deadlines. That cause some of my GPUs to either slow way down or to stop completely. I finally just decided to stop all CPU work units till I can upgrade to an 8 core. May not have any relevance to what is happening with you but thought I'd throw that out there. Anything else you noticed when this was happening?

Zalster

You need to free at least one CPU core for each GPU to feed the GPU`s properly.

---

With each crime and every kindness we birth our future.

Nope, not running any CPU work units at all and have a full core dedicated to each GPU work unit.

Thanks,

Chris

---

How many AP instances are you running on your GPU`s ?

---

With each crime and every kindness we birth our future.

I've tried both 2 and 1. Doesn't seem to make a difference. I've also tried the -cpu_lock setting to see if it was having trouble bouncing between cores. Didn't seem to have any benefit.

Basically the WU's start out fine and then at some point in time the GPUs just stop processing some or all of them...

Thanks,

Chris

---

Try to lower your params.

-unroll 6 ffa_block 6144 ffa_block_fetch 1536

See if this helps.

---

With each crime and every kindness we birth our future.

Running it with those settings to see how things go.

Thanks,

Chris

---

Nope, still went to zero utilization again. Tried getting rid of the -cpu_sleep and updated my driver. We shall see how that goes.

Thanks,

Chris

---

So it started as you added the 750Ti?

Personally I'd never mix 2 nvidia cards together that are so different architecturally from each other.

Try taking the 560Ti out 1st and see if the problem continues.

If it doesn't and you want to run 2 cards then I'd suggest getting another 750Ti.

If the problem continues then the 750Ti maybe at fault.

Cheers.

Yeah, considered that... Thinking about putting the 750ti with the 770 and a 570 with the 560 ti. I'll pull cards after I see if any settings can fix it up. I kinda consider that my last resort since I'll need to get a different power supply for one if the machines to get all of that working...

Thanks,

Chris

---

Ok, I finally narrowed it down to the -use_sleep command line switch. Once I turned took that off the computer burned through the rest of the AP work units in no time without any problems. CPU processing times went up of course but no GPU hang ups.

Thanks,

Chris

---