Yesterday I installed the newest version of BOINC in to my PC that is running Windows Vista 64 bit edition (service pack 2). I have an NVIDIA CUDA capable GPU, and when first installed, BOINC recognized the GPU inmediatly (im using the latest drivers from NVIDIA 186.81), downloaded a workunit for the GPU (as well for the CPUs) and started computation normally.

There is a problem when I restart or shut down the PC: when the client restarts at boot, there are sometimes when the client doesn't recognize a CUDA device in the system, therefore loosing all data computed before. I don't know exactly why this happens as there are other tmes when the PC is restarted or rebooted and there are no problems at all.

This has happen at least thrice the last day, and now the server won't give me any other GPU workunits. The GPU never had a chance to compute an entire workunit because of this problem. Here are the latest messages from the client: (this time, after a reboot when the client didn't recognize a CUDA device, now recognized it)

04/09/2009 08:53:14 a.m. Starting BOINC client version 6.6.36 for windows_x86_64
04/09/2009 08:53:14 a.m. log flags: task, file_xfer, sched_ops
04/09/2009 08:53:14 a.m. Libraries: libcurl/7.19.4 OpenSSL/0.9.8j zlib/1.2.3
04/09/2009 08:53:14 a.m. Data directory: C:\ProgramData\BOINC
04/09/2009 08:53:14 a.m. Running under account Jorge Ramos
04/09/2009 08:53:14 a.m. Processor: 2 GenuineIntel Intel(R) Core(TM)2 Duo CPU P7350 @ 2.00GHz [Intel64 Family 6 Model 23 Stepping 10]
04/09/2009 08:53:14 a.m. Processor features: fpu tsc pae nx sse sse2 pni
04/09/2009 08:53:14 a.m. OS: Microsoft Windows Vista: Home Premium x64 Edition, Service Pack 2, (06.00.6002.00)
04/09/2009 08:53:14 a.m. Memory: 3.75 GB physical, 3.63 GB virtual
04/09/2009 08:53:14 a.m. Disk: 282.99 GB total, 227.62 GB free
04/09/2009 08:53:14 a.m. Local time is UTC -5 hours
04/09/2009 08:53:14 a.m. CUDA device: GeForce 9500M (GF 9200MGS + GF 9400MG) (driver version 18681, compute capability 1.1, 256MB, est. 4GFLOPS)
04/09/2009 08:53:14 a.m. Not using a proxy
04/09/2009 08:53:14 a.m. SETI@home URL: http://setiathome.berkeley.edu/; Computer ID: 5093139; location: (none); project prefs: default
04/09/2009 08:53:14 a.m. No general preferences found - using BOINC defaults
04/09/2009 08:53:14 a.m. Reading preferences override file
04/09/2009 08:53:14 a.m. Preferences limit memory usage when active to 1918.71MB
04/09/2009 08:53:14 a.m. Preferences limit memory usage when idle to 3453.69MB
04/09/2009 08:53:14 a.m. Preferences limit disk usage to 10.00GB
04/09/2009 08:53:14 a.m. SETI@home Restarting task ap_23ap08ab_B3_P1_00107_20090903_28463.wu_0 using astropulse_v505 version 505
04/09/2009 08:53:14 a.m. SETI@home Restarting task ap_23au08ae_B5_P0_00353_20090901_11107.wu_2 using astropulse_v505 version 505
04/09/2009 08:53:14 a.m. SETI@home Sending scheduler request: To fetch work.
04/09/2009 08:53:14 a.m. SETI@home Requesting new tasks for GPU
04/09/2009 08:53:28 a.m. SETI@home Scheduler request completed: got 0 new tasks
04/09/2009 08:53:28 a.m. SETI@home Message from server: (Project has no jobs available)
04/09/2009 08:54:44 a.m. SETI@home Sending scheduler request: To fetch work.
04/09/2009 08:54:44 a.m. SETI@home Requesting new tasks for GPU
04/09/2009 08:54:50 a.m. SETI@home Scheduler request completed: got 0 new tasks
04/09/2009 08:54:50 a.m. SETI@home Message from server: (Project has no jobs available)
04/09/2009 08:56:06 a.m. SETI@home Sending scheduler request: To fetch work.
04/09/2009 08:56:06 a.m. SETI@home Requesting new tasks for GPU
04/09/2009 08:56:11 a.m. SETI@home Scheduler request completed: got 0 new tasks
04/09/2009 08:56:11 a.m. SETI@home Message from server: (Project has no jobs available)
04/09/2009 08:57:27 a.m. SETI@home Sending scheduler request: To fetch work.
04/09/2009 08:57:27 a.m. SETI@home Requesting new tasks for GPU
04/09/2009 08:57:32 a.m. SETI@home Scheduler request completed: got 0 new tasks
04/09/2009 08:57:32 a.m. SETI@home Message from server: (Project has no jobs available)
04/09/2009 08:58:48 a.m. SETI@home Sending scheduler request: To fetch work.
04/09/2009 08:58:48 a.m. SETI@home Requesting new tasks for GPU
04/09/2009 08:58:53 a.m. SETI@home Scheduler request completed: got 0 new tasks
04/09/2009 08:58:53 a.m. SETI@home Message from server: (Project has no jobs available)
04/09/2009 09:05:14 a.m. SETI@home Sending scheduler request: To fetch work.
04/09/2009 09:05:14 a.m. SETI@home Requesting new tasks for GPU
04/09/2009 09:05:19 a.m. SETI@home Scheduler request completed: got 0 new tasks
04/09/2009 09:05:19 a.m. SETI@home Message from server: (Project has no jobs available)
04/09/2009 09:32:38 a.m. SETI@home Sending scheduler request: To fetch work.
04/09/2009 09:32:38 a.m. SETI@home Requesting new tasks for GPU
04/09/2009 09:32:43 a.m. SETI@home Scheduler request completed: got 0 new tasks
04/09/2009 09:32:43 a.m. SETI@home Message from server: (Project has no jobs available)
04/09/2009 09:34:58 a.m. SETI@home Sending scheduler request: To fetch work.
04/09/2009 09:34:58 a.m. SETI@home Requesting new tasks for GPU
04/09/2009 09:35:03 a.m. SETI@home Scheduler request completed: got 0 new tasks
04/09/2009 09:35:03 a.m. SETI@home Message from server: (Project has no jobs available)


This problem only happens with the GPU, the CPUs are working just fine. any help is appreciated.

I don't know about any other problems, but Seti is running very low on work and is likely to run completely out for the weekend if new data doesn't come in from the telescope before the end of the day.

---

Boinc V 7.4.36
Win7 i5 3.33G 4GB NVidia 470

could i have more info on this:

"CUDA device: GeForce 9500M (GF 9200MGS + GF 9400MG) (driver version 18681, compute capability 1.1, 256MB, est. 4GFLOPS)"

is this computer a laptop?

Hi:

Yes... it's a Dell Studio XPS 1340

Jorge.

i wonder why boinc is reporting 2 on-board cards...

Hi:

The Dell XPS 1340 has the option of the "9500M GPU", not the original one, but a combination of the mGPU 9400M + a discrete 9200M GS GPU working in Hybrid-SLI mode, thus creating a boost in performance when the two are active (which is this case). This is much like the case of the MAcbook Pro which are now available with the 9400M mGPU + a discrete 9600M GPU.


Jorge.

Maybe this problem could be solved if, BOINC don't erase (CUDA) computing data if suddenly, it can´t recognize a CUDA device in the system, when it was working fine prior a reboot or a shut-down.

The work done so far can be recalled after another reboot or shutdown, when BOINC would recognize the CUDA device.

Could you please post the start up messages for when your BOINC does not find a CUDA capable card in your system?

---

Hi:

I think I know what the problem is... Because my PC is a Laptop and it has a hybrid-SLI system, there exists the posibility that when I rebooted the PC, the energy savings protocol was active, and the discrete GPU (which is the one responsible for CUDA in my system) was inactive a the time when BOINC started automatically and tried to detect a cuda device (which I think it does scan the system after every reboot).

I havn't been able to test this, mainly because I am full of workload now (CPU and GPU) and havn't restarted my PC since.