For the driver restart part of the problem, you're misconfigured the Nvidia OpenCL AP app:
Running on device number: 0 DATA_CHUNK_UNROLL at default:2 Number of app instances per device set to:2 DATA_CHUNK_UNROLL set to:12 FFA thread block override value:16384 FFA thread fetchblock override value:2048 Priority of worker thread raised successfully Priority of process adjusted successfully, high priority class used How powerful do you think a GTS 250/9800GTX+ is? Even running Stock app settings the GUI will be laggy, increasing the number of instances, the Unroll value, and the FFA thread block value, will all make the app cause driver restarts, Reduce them to the default settings and try again. Claggy >----Original Message---- >From: elliott.ch@verizon. net >Date: 29/09/2013 14:25 >To: <[email protected]> >Subj: [boinc_dev] (no subject) > >In cc_config.xml: > > <exclude_gpu> > > <url>http://setiweb.ssl.berkeley.edu/beta//url> > > <device_num>0</device_num> > > <type>NVIDIA</type> > > <app>astropulse_v6</app> > > </exclude_gpu> > > > >Stdoutdae.txt: 29-Sep-2013 08:21:01 [---] Unrecognized tag in cc_config.xml: ><exclude_gpu> > > > >Thanks a lot. > > > >If anyone is interested, all the recent astropulse_v6 version 604 >(opencl_nvidia_100) apps > >on the beta site are erroring out on NVidia GeForce GTS 250 cards. I have >tried four > >different GTS/GTX GeForce 250 cards, and two different display drivers, >version > >320. 49 and 327.23, the latter being the most recent. The error pattern is: > > > >Start astropulse_v6 app(1) on GPU > >Runs for ~1:33 minutes: seconds > >Desktop posts a message: Display driver stopped responding and has recovered >(version, e.g., 327.23) > >There is a watchdog timeout error in the Event log, but it contains no >useful information that I can find. > >The affected WU(1) goes into a scheduler wait > >A different WU(2) is started with the same app > >WU(2) is processed for ~1:33, the display driver stops, WU(2) is put on >scheduler wait, and the AP app goes back to WU(1) > >WU(1) starts from the beginning (0:00 time) and runs for ~1:33 > >This cycle repeats indefinitely until eventually the WU is aborted. > > > >When the WU is being processed on the GTS 250, the GPU runs very hot, ~180 >F. The only way > >I could cool the card was by placing a large Honeywell fan (picture: > >http://www.kaz.com/kaz/fans/products/honeywell-turbo-force-room-air-circulat >or-ht-908/) > >a few inches from the GPU with the air stream playing directly on it. Then >it runs at about 120 F. > >I have been using these 250 GPUs, on a different motherboard, for several >months and this is > >the first time this problem (including the overheating in the summer months) >has > >appeared. The 250 replaces a 460, which is in the shop. The > >other GPU in the system is a 550 Ti, which is not having any trouble with >the AP WUs. > > > >The "Tasks" page for Computer ID 58103 indicates that most of the AP WUs >that errored > >on ID 58103 also errored out or were aborted on other NVidia GPUs, although >some > >were successful on ATI GPUs. > > > >Charles Elliott > >_______________________________________________ >boinc_dev mailing list >[email protected] >http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev>To unsubscribe, visit the above URL and >(near bottom of page) enter your email address. > _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
