A few observations about the Gamma-ray pulsar search #2 v1.07 (FGRPopencl-nvidia) app performance,

System - W7, Boinc 7.0.64, i7-3770K@4.2GHz, 8GB DDR3 @ 2133MHz, SATA6-SSD, GTX660Ti, GTX660 (PCIE3 X8).

GPU utilization factor of BRP apps 1:
GPU usage fluctuated up to ~33% on both cards with GPU power moving between 28 and 36%
GDDR usage was just under 290 and 300MB.
CPU usage was one thread per GPU app and kernel time was low.
GPU temps 38°C and 42°C.

Performance didn't noticeably improve/deteriorate when suspended/resuming CPU work.

GPU usage drops to zero for 1min 10seconds mid-run (@40% and @95% progress).

LATeah0069U_16.0_860_-9.14e-10_2 382268 8 Jun 2013 | 18:13:58 UTC 9 Jun 2013 | 7:59:06 UTC Completed and validated 769.03 765.22 70.00 Gamma-ray pulsar search #2 v1.07 (FGRPopencl-nvidia)
LATeah0069U_16.0_860_-9.05e-10_2 382259 8 Jun 2013 | 18:12:55 UTC 9 Jun 2013 | 7:12:37 UTC Completed and validated 770.06 766.15 70.00 Gamma-ray pulsar search #2 v1.07 (FGRPopencl-nvidia)

A RAC of 7 or 8K from a GTX660Ti is misery.

See message 112492 (earlier in this thread) for a visual demonstration of that, and Bikeman's reply for an explanation. I wouldn't worry about RAC at a beta project like this - it's a work-in-progress.

Recently all the tasks for this app on this computer
https://albert.phys.uwm.edu/show_host_detail.php?hostid=6867
have started error-ing after @ 2 sec with an Exit status 11 (0xb).
Previously had no problems running. Has anything changed on your side or is the error code reflecting something that has changed on my side ?
Have tried resetting the project but again it just throws errors. Everything was fine up until 13/6/13 by which time I had completed a few 100 units, but haven’t managed a single successful unit since.
Any ideas I can try ?

Did you changed anything on your machine around the time the problems started? Updated the graphics driver? Overheating?

---

No, nothing changed my side as far as I'm aware.
I believe that signal 11 errors are usually memory related ( unlikely as the card in question runs primegrid and gpugrid no problem ), problem with 64bit system running 32 bit apps ( have checked 32 bit libs are still installed ) or bad work units.
Am I missing anything obvious ?

I'm not sure if it is a bug but I can see:
http://albert.phys.uwm.edu/workunit.php?wuid=405509
The WUs on opencl-nvidia are validated together and my WU (CPU) is marked as invalid. It might be that my computer sent wrong answer but it is the only invalid WU in albert (on eintein I did't have invalid WUs too as I remember).

---

I think I received my first task of this type:

Gamma-ray pulsar search #2
hsgamma_FGRP2
version: 1.07

I have 3 GPUs: GTX 660 Ti, GTX 460, GTS 240.
I have GPU Exclusions such that the Albert tasks can only run on the GTS 240.

That being said, I'm noticing VERY LOW GPU utilization on my GTS 240, for this app.
It fluctuates between 4% and 10%, and that's all. Even if that is the only task being processed, it never gets more than 10% GPU utilization. It does use a full CPU core while processing.

Is this low GPU utilization on this GTS 240... expected behavior for this app?

Thanks,
Jacob