This batch consists super-long workunits.

         |                       |            |Estimated| GPU | GPU
   CPU   |        GPU            |     OS     | runtime |usage|frequ
i7-4930k | GTX TITAN X (Maxwell) | Windows XP | 18h 22m | 81% | 1390
i3-4370  | GTX 980Ti   (Maxwell) | Windows XP | 22h 18m | 69% | 1316
i3-4160  | GTX 980Ti   (Maxwell) | Windows XP | 19h 55m | 75% | 1366

As the runtimes does not scale, and there's very different GPU usage I think these workunits have different number of atoms in the same batch. We'll see in 18h.

Like I said in another thread, very strange usage on my 970 on windows 10, it doesn't seem to be any different on XP. In fact, My 970 has downclocked itself from 1392 stock boost to 1164 with the same GPU load, ~75%

The long runtime was a mistake by Adria. We apologize and have stopped the WUs as well. He will resend them with shorter runtimes.

The GPU utilization is not something we can control. Some simulations (like membrane protein simulations) include CPU logic which might be the cause of that, but we cannot change that.

e1s14_2-ADRIA_FAAH_WT_3-0-1-RND6812_0
17h 39m 15s
585.750 credits
i7-4930k, GTX TITAN X (Maxwell) @ 1390 MHz, Windows XP

The long runtime was a mistake by Adria. We apologize and have stopped the WUs as well. He will resend them with shorter runtimes.

Should we abort these WUs?

The GPU utilization is not something we can control. Some simulations (like membrane protein simulations) include CPU logic which might be the cause of that, but we cannot change that.

That's ok, but you should add some extra credits in this case, as these workunits take longer thus their credit/time ratio is less.

If he could split these WUs into fourths or less that would be great.

I picked up one of these on my Win7 I7-6800K host with two GTX 980s. The task is 73% complete at 22 hours. I am only getting 67% GPU utilization but I am running 8 threads of evolution at home on the CPUs. With CPU tasks suspended GPU utilization is 79%.

e1s13_2-ADRIA_FAAH_WT_1-0-1-RND9231_0
19h 55m 8s
585.750 credits
i3-4160, GTX 980Ti (Maxwell) @ 1366 MHz, Windows XP

PERFORMANCE: 104124 Natoms 5.736 ns/day

The previous task:

PERFORMANCE: 104124 Natoms 5.058 ns/day

So these tasks are much more dependant on the CPU than the previous similar ones.

The WU still seems to use only about ~70% Usage on windows 10 having restarted the computer to fix the low clock freq bug. It only uses 7% of the 4 thread i5 2400 at 3.2ghz using 300MB of RAM. What freq are you running on that 4930k Zoltan? I'm curious just how much the CPU freq matters to these WUs

What freq are you running on that 4930k Zoltan? I'm curious just how much the CPU freq matters to these WUs

4.4GHz

I didn't manage to snag one of the new ADRIA_MI_FAAH_WTYPE_ WUs, Are they similar in performance?

I didn't manage to snag one of the new ADRIA_MI_FAAH_WTYPE_ WUs, Are they similar in performance?

Their performance is similar (74% GPU usage), but they are much shorter (~6 hours on a GTX 980Ti@1366MHz/i3-4160/Windows XP)

...but they are much shorter (~6 hours on a GTX 980Ti@1366MHz/i3-4160/Windows XP)

Ah, thanks for taking my advice Stefan

I've noticed that my (Windows XP x64) hosts which had the 368.22 driver (CUDA 8.0) were processing these workunits slower than those which have the 359.06 or 358.50 driver (CUDA 7.5), so now I've downgraded them to 359.06.