I am not getting that error, but I am also not sure if I am getting tasks for the GPU(s).
You wont be getting any GPU tasks because your GPUs don't meet the documented requirements. You need 512 MB cards to have a chance of meeting the min memory requirements.
Quote:
- How can one tell from the tasks? Unlike PrimeGrid, there is no parenthesis indicating the type of processing approach, i.e. "(cuda32)", and the "(SSE2)" or "(BRP4SSE)" does not tell me..?
SSE and SSE2 are CPU capabilites required by CPU only tasks. If you see either of these in the app name you know you are dealing with a CPU task. With NVIDIA GPUs, if you see 'cuda' as part of the app name, you know you are dealing with a GPU task. All the current app names are listed here. When a GPU task is running you will see something like (0.20CPUs + 1.00 NVIDIA GPUs) if you are running the GPU tasks one at a time. With modern GPUs and sufficient GPU memory, you can run multiple tasks concurrently.
Quote:
Sun Dec 2 06:40:31 2012 | | NVIDIA GPU 0: GeForce 9600M GT (driver version 5.0.37, CUDA version 5.0, compute capability 1.1, 256MB, 53MB available, 120 GFLOPS peak)
Sun Dec 2 06:40:31 2012 | | NVIDIA GPU 1: GeForce 9400M (driver version 5.0.37, CUDA version 5.0, compute capability 1.1, 254MB, 231MB available, 53 GFLOPS peak)
....
These lines tell you that BOINC has recognised your GPUs and that they may be of use to other projects but unfortunately not for Einstein.
Everything else in your log snippet pertaining to downloading of tasks and data would be for CPU tasks only.
Thank you both very much for the quick heads-up. Just too bad about the 512Mb requirement...
Sorry if I went a bit around the bush here by the way. I did try to find more information on this but failed to find the info that you just provided. There is probably another reason then for the earlier posters issue(s). I guess you won't be needing my host :-)
I had forgotten what the GPU limits were here, From the host id you posted in the other thread, i looked up your last scheduler contact which says the same
(It looks like a 'resend lost tasks' scheduler contact, But some work wasn't resent because of a lack of disk space allocated):
Quote:
2012-12-02 16:54:29.6248 [PID=14391] [send] [HOST#6237669] Sending app_version einstein_S6LV1 6 113 SSE2; 4.86 GFLOPS
2012-12-02 16:54:29.6253 [PID=14391] [send] est. duration for WU 139377392: unscaled 13340.98 scaled 13492.94
2012-12-02 16:54:29.6253 [PID=14391] [HOST#6237669] Sending [RESULT#324410466 h1_0429.65_S6GC1__1510_S6LV1B_0] (est. dur. 13492.94 seconds)
2012-12-02 16:54:31.6392 [PID=14391] [send] est. duration for WU 139376176: unscaled 13340.98 scaled 13492.94
2012-12-02 16:54:31.6392 [PID=14391] [send] [WU#139376176] meets deadline: 53971.75 + 13492.94 < 1209600
2012-12-02 16:54:31.6398 [PID=14391] [send] [HOST#6237669] Sending app_version einstein_S6LV1 6 113 SSE2; 4.86 GFLOPS
2012-12-02 16:54:31.6407 [PID=14391] [send] est. duration for WU 139376176: unscaled 13340.98 scaled 13492.94
2012-12-02 16:54:31.6407 [PID=14391] [HOST#6237669] Sending [RESULT#324406832 h1_0429.70_S6GC1__804_S6LV1B_1] (est. dur. 13492.94 seconds)
2012-12-02 16:54:31.6465 [PID=14391] [send] est. duration for WU 139376177: unscaled 13340.98 scaled 13492.94
2012-12-02 16:54:31.6465 [PID=14391] [send] [WU#139376177] meets deadline: 67464.69 + 13492.94 < 1209600
2012-12-02 16:54:31.6470 [PID=14391] [send] [HOST#6237669] Sending app_version einstein_S6LV1 6 113 SSE2; 4.86 GFLOPS
2012-12-02 16:54:31.6477 [PID=14391] [send] est. duration for WU 139376177: unscaled 13340.98 scaled 13492.94
2012-12-02 16:54:31.6477 [PID=14391] [HOST#6237669] Sending [RESULT#324406834 h1_0429.70_S6GC1__803_S6LV1B_1] (est. dur. 13492.94 seconds)
2012-12-02 16:54:31.6521 [PID=14391] [send] est. duration for WU 139377393: unscaled 13340.98 scaled 13492.94
2012-12-02 16:54:31.6522 [PID=14391] [send] [WU#139377393] meets deadline: 80957.62 + 13492.94 < 1209600
2012-12-02 16:54:31.6526 [PID=14391] [send] [HOST#6237669] Sending app_version einstein_S6LV1 6 113 SSE2; 4.86 GFLOPS
2012-12-02 16:54:31.6532 [PID=14391] [send] est. duration for WU 139377393: unscaled 13340.98 scaled 13492.94
2012-12-02 16:54:31.6532 [PID=14391] [HOST#6237669] Sending [RESULT#324410468 h1_0429.70_S6GC1__802_S6LV1B_0] (est. dur. 13492.94 seconds)
2012-12-02 16:54:31.6579 [PID=14391] [send] est. duration for WU 139377395: unscaled 13340.98 scaled 13492.94
2012-12-02 16:54:31.6579 [PID=14391] [send] [WU#139377395] meets deadline: 94450.56 + 13492.94 < 1209600
2012-12-02 16:54:31.6584 [PID=14391] [send] [HOST#6237669] Sending app_version einstein_S6LV1 6 113 SSE2; 4.86 GFLOPS
2012-12-02 16:54:31.6589 [PID=14391] [send] est. duration for WU 139377395: unscaled 13340.98 scaled 13492.94
2012-12-02 16:54:31.6589 [PID=14391] [HOST#6237669] Sending [RESULT#324410472 h1_0429.70_S6GC1__801_S6LV1B_0] (est. dur. 13492.94 seconds)
2012-12-02 16:54:33.6701 [PID=14391] [send] stopping work search - insufficient disk space
2012-12-02 16:54:33.6701 [PID=14391] [send] stopping work search - insufficient disk space
2012-12-02 16:54:33.6701 [PID=14391] [send] stopping work search - insufficient disk space
2012-12-02 16:54:33.6904 [PID=14391] [version] Checking plan class 'BRP4SSE'
2012-12-02 16:54:33.6905 [PID=14391] [version] plan class ok
2012-12-02 16:54:33.6905 [PID=14391] [version] Checking plan class 'BRP4cuda32OSX'
2012-12-02 16:54:33.6905 [PID=14391] [version] parsed project prefs setting 'gpu_util_brp': 0.000000
2012-12-02 16:54:33.6905 [PID=14391] [version] GPU RAM required max: 314572800.000000, supplied: 265945088.000000
2012-12-02 16:54:33.6905 [PID=14391] [version] Checking plan class 'opencl-ati-lion'
2012-12-02 16:54:33.6905 [PID=14391] [version] OS version required min: 110000, supplied: 100800
2012-12-02 16:54:33.6905 [PID=14391] [version] Best version of app einsteinbinary_BRP4 is ID 311 (3.04 GFLOPS)
2012-12-02 16:54:33.6906 [PID=14391] [send] [HOST#6237669] [WU#139382154 p2030.20111212.G194.13-01.78.N.b1s0g0.00000_592] WU is infeasible: Not enough disk
2012-12-02 16:54:33.6962 [PID=14391] Sending reply to [HOST#6237669]: 8 results, delay req 60.00
2012-12-02 16:54:33.6970 [PID=14391] Scheduler ran 6.237 seconds
RE: I am not getting that
)
You wont be getting any GPU tasks because your GPUs don't meet the documented requirements. You need 512 MB cards to have a chance of meeting the min memory requirements.
SSE and SSE2 are CPU capabilites required by CPU only tasks. If you see either of these in the app name you know you are dealing with a CPU task. With NVIDIA GPUs, if you see 'cuda' as part of the app name, you know you are dealing with a GPU task. All the current app names are listed here. When a GPU task is running you will see something like (0.20CPUs + 1.00 NVIDIA GPUs) if you are running the GPU tasks one at a time. With modern GPUs and sufficient GPU memory, you can run multiple tasks concurrently.
These lines tell you that BOINC has recognised your GPUs and that they may be of use to other projects but unfortunately not for Einstein.
Everything else in your log snippet pertaining to downloading of tasks and data would be for CPU tasks only.
Cheers,
Gary.
Hiya Gary & Claggy Thank
)
Hiya Gary & Claggy
Thank you both very much for the quick heads-up. Just too bad about the 512Mb requirement...
Sorry if I went a bit around the bush here by the way. I did try to find more information on this but failed to find the info that you just provided. There is probably another reason then for the earlier posters issue(s). I guess you won't be needing my host :-)
Cheers
J
I had forgotten what the GPU
)
I had forgotten what the GPU limits were here, From the host id you posted in the other thread, i looked up your last scheduler contact which says the same
(It looks like a 'resend lost tasks' scheduler contact, But some work wasn't resent because of a lack of disk space allocated):
Claggy