Author |
Message |
5potSend message
Joined: 8 Mar 12 Posts: 411 Credit: 2,083,882,218 RAC: 0 Level
Scientific publications
|
Figured I would start a beta thread. Would like to see how times are, in comparison, especially with old vs. new GPUs. Whoever runs Betas, if you could post runtimes and cards used, would love to see how it's going. RUNTIMES, and not issues/problems. That's in a different thread.
Thanks |
|
|
nenymSend message
Joined: 31 Mar 09 Posts: 137 Credit: 1,308,230,581 RAC: 0 Level
Scientific publications
|
My the first successful run of a 6.47 beta NOELIA task. I have changed nothing during last week on the host (XP 64bit, GTX 560Ti CC2.1, core 6.10.58). Only a task changed: trypsin_lig_100 -> trypsin2com.
http://www.gpugrid.net/results.php?hostid=128242&offset=0&show_names=1&state=0&appid=21 |
|
|
|
GTX 670 Win7 x64 301.42 failed all 3 betas is was sent. This same machine has runs all LONG WUs sucessfully to completion.
None of these were run to successful completion by anyone
run2-NOELIA_LIE_trypsin_lig_1000ligand-0-1-RND3135_5
run2-NOELIA_LIE_trypsin_lig_100complex-0-1-RND6701_5
run2-NOELIA_LIE_trypsin_lig_1005complex-0-1-RND9180_5
____________
Thanks - Steve |
|
|
|
I just received a beta for app 6.47 (CUDA 4.2) running a NOELIA run4 and so far everything looks good as my GTX 670 (Win7 x64, BOINC 7.0.25, driver 301.42) crunched it to completion without error. Is this a beta squared ??? (app and wu type are both new)
- GPU utilization was between 77%-79%: Lower than average ~90%
- CPU usage held steady about 3% - 4% (i7 920 w/ HT on). Much better than using a full core like all of the LONG WU types do on Win7. Perhaps this will change when moved from beta to LONG but that's how it behaved during this run.
^^^I put those two pieces of information together because I believe they are related. ^^^
- GPU mem usage at 21-22%: Lower than average ~30%
- CPU mem working set is at 119k and page faults at 67k.
I don't normally look at CPU memory usage as it doesn't matter much to me because I have plenty of RAM in this machine.
GPU runtime = 3016 sec.
CPU runtime = 913 sec
Time per step (avg over 525000 steps): 5.740 ms
Total points = 3000
Upload file size = 6 MB (meaningless as this is beta)
All in all looks pretty good to me and most importantly they (app 6.47 and NOELIA run4) work well together so we can now get you the results you need! <applause>
____________
Thanks - Steve |
|
|
|
just received and failed a NOELIA run1
<core_client_version>7.0.25</core_client_version>
<![CDATA[
<message>
- exit code 98 (0x62)
</message>
<stderr_txt>
ERROR: file deven.cpp line 1106: # Energies have become nan
called boinc_finish
</stderr_txt>
]]>
____________
Thanks - Steve |
|
|
nenymSend message
Joined: 31 Mar 09 Posts: 137 Credit: 1,308,230,581 RAC: 0 Level
Scientific publications
|
Hi hardly testing NOELIA.
Xeon 2,83GHz + GTX560Ti CC2.1 driver 301.42, XP 64bit, 6.10.58 ID 12842
"short" task: high CPU load cca 50-60% of core, low GPU load cca 70%
"long" task: middle CPU load cca 20 - 30% of core, GPU load 85 - 90%
PAOLA & NATHAN longrun tasks makes GPU load 99%.
Notice: Tasks failed immediately on 9600GT and GTX260 rev A.2, as I have expekted. |
|
|
nenymSend message
Joined: 31 Mar 09 Posts: 137 Credit: 1,308,230,581 RAC: 0 Level
Scientific publications
|
Some "long" tasks erroed out for high OC (890MHz), but it is stable for actual longrun tasks. Run1 errored out immediately. By my point of view low GPU load and high sensitivity for OC is not a good combination. |
|
|
|
Been getting beta's good part of the day and so far they seem to be running well.
I am running two GTX570's on a 6 core.
They seem to be taking just over an hour to run.
Seen a cuda 40 come through too. Ran well.
Aaron |
|
|