+1

Every 7.13 WU I receive immediately fails with computation error, as well.

Thanks SekeRob,

Done.

TKH

First also check up if OpenCL 1.1 is loaded too [see event/message log at start-up], a *must*.

There is no authorized GPU list. There's a list of explicitly blocked cards. Some cards do work, but not for everyone, not in every hardware/BIOS/OS/software/driver configuration, so it becomes a "are you lucky". Research latest WHQL and or Beta drivers, and see if you can get that Kernel time down.... trial and error.

(Yesterday saw an odd log where it was suggested that an OpenCL 1.0 enabled card was getting work done... I'm waiting on the owner to come back in his own time to tell if the results validate... could not as there is the special OpenCL extension that only got included from 1.1 [so is my reading]... see https://secure.worldcommunitygrid.org/help/viewSearch.do?searchString=opencl )

I'm having similar issues using a Mac. Interestingly, I was able to start running a GPU based work unit for HCC and it was making rapid progress - as you might expect. Incredibly rapid progress, indeed. However, as I was trying to watch a video at the time, I changed the settings to 'Use GPU according to preferences' instead of all the time because it made the computer unusable. I did this with the intention of switching back to enable the GPU computation afterwards.

When I did so, the unit resumed and instantaneously reached the 'Computation error' stage, as did the next few work units which my laptop attempted to run. There were no changes in the use of the computer - indeed, it was under less stress because the video had been stopped. There's less running now than there was earlier. I'm using 4 of the 8 virtual cores to keep temperature down and these are running at max 25%load for BOINC. Would that affect anything? I wouldn't have thought so, but you guys will know better.

I had to pause the work unit because the video became more of a sequence of juddery stills whilst running a GPU work unit too, but I had expected it to resume without issue. My laptop appears to have OpenCL properly installed; below is a paste of the log entry as BOINC starts up:

Mon Apr 1 00:32:28 2013 | | OpenCL: NVIDIA GPU 0: GeForce GT 650M (driver version CLH 1.0, device version OpenCL 1.1, 1024MB, 1024MB available, 29 GFLOPS peak)

I attempted to reboot BOINC to see if that would make a difference new GPU tasks still failed. I'm using BOINC 7.0.47 so that my laptop will not automatically switch to use dedicated graphics whenever BOINC is running apart from when doing GPU crunching.

As I'm typing this now, one unit failed instantly and the next one is on 12.183% after 1 min 23 seconds. I'm perhaps a little confused right now because they ran one after the other!

I hope this slight ramble information helps you guys to figure out what's going on - whether it is a BOINC issue or whether the laptop isn't capable of doing the work, or anything else that might be causing things to fail. If you need any more information, I'll try to do my best to help further but the seemingly random nature of work units running and failing makes it hard for me to give more information other than 'nothing has meaningfully changed' between when units work and don't work!

Cheers,

Matt

After rebooting, the first unit hit a computation error immediately, the second is running without issue. The issue doesn't appear to be fixed by rebooting unfortunately.