Message boards :
Number crunching :
AK V8 + CUDA MB team work mod
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 . . . 12 · Next
Author | Message |
---|---|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Not quite true. Lesson is: you should balance project shares and work cache length to get correct combination of running apps. And I curious how you avoid idle core situation with Einstein + CUDA MB stock in BOINC 6.4.5 ? I end up or with idle core (3+1) or with 5+1 combinations (last more optimal actually than first). And even more important lesson - we deadly need project pairing ability from BOINC that still is absent :/ |
RandyC Send message Joined: 20 Oct 99 Posts: 714 Credit: 1,704,345 RAC: 0 |
Quite simple...E@H uses both cpus and CUDA uses gpu. As I said, I've got cc_config set up for ncpus=3. Long Term debt value goes WAY up for Einstein and I occasionally reset it to zero using BoincDV. Not sure BoincDV's still available though...a quick Google shows the DL page is gone. |
Westsail and *Pyxey* Send message Joined: 26 Jul 99 Posts: 338 Credit: 20,544,999 RAC: 0 |
Maybe try for roll back to 180.48? I use that driver on all cards and the 9500's will only work with that or 178.xx, everyting work fine then. Not sure if the problem but I know I can't use anything above 180. Good luck "The most exciting phrase to hear in science, the one that heralds new discoveries, is not Eureka! (I found it!) but rather, 'hmm... that's funny...'" -- Isaac Asimov |
Paul D Harris Send message Joined: 1 Dec 99 Posts: 1122 Credit: 33,600,005 RAC: 0 |
Thanks I made the file like you said and I quite BOINC and Thanks I got it to working now! |
Paul D Harris Send message Joined: 1 Dec 99 Posts: 1122 Credit: 33,600,005 RAC: 0 |
Thanks I made the file like you said and I quite BOINC and How do you set up the cc_config.xml file for a core i7 chip? I have one coming next week. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Interesting question indeed. It depends of how good AK v8 and AP will utilize HT. In HT P4 days pair of opt app could perform worse than single opt app with HT disabled. If app is highly optimized it tends to use all available hardware resources of core - there is too low amount of idle core parts that can benefit from HyperThreading. So I consider it as open field for experimentation on this stage. |
Paul D Harris Send message Joined: 1 Dec 99 Posts: 1122 Credit: 33,600,005 RAC: 0 |
So do I set it for 4 cores+1 or 8 cors+1 |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14649 Credit: 200,643,578 RAC: 874 |
As your tag says, you're the one who has volunteered to test it! Run it for a day or two at 4+1, then switch and run it for a day or two at 8+1: compare how much work you got through with each setting, and come back and tell us the result. |
Björn Send message Joined: 29 Jul 01 Posts: 22 Credit: 1,058,001 RAC: 0 |
Had one WU that got stuck at 0,000% and 1 second so far. Oddly enough the 2 cores shared 3 WUs after I cancelled that WU (http://setiathome.berkeley.edu/workunit.php?wuid=403063999). Seems to be working smooth now. Kudos, Raimster! Edit: had another one right after I posted. http://setiathome.berkeley.edu/workunit.php?wuid=403063986 Again, my two cores worked on 3 tasks simultaneously, leaving the GPU in idle. I'm using a X2 4800+ and a 9800GT with your latest SSE3 package, if that helps. Edit 2: After restarting Boinc, everything works again as it should, with 2 CPU and 1 GPU task. |
Byron S Goodgame Send message Joined: 16 Jan 06 Posts: 1145 Credit: 3,936,993 RAC: 0 |
|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Hi Raistmer, Ah, sure. I have another enhanced AK v8 version for AMD chips, it should work little faster even than released x86 AK_v8 SSE3 and SSE2 variants for AMD chips. As testing showed it still true not only for earlier AMD Athlon 64 and X2 chips but for more new Phenom chips too. Don't know about Phenom 2 though, it would be nice to have some tester with Phenom 2 chip onboard... (will prepare new release and post links after) |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
|
Golden_Frog Send message Joined: 28 Oct 99 Posts: 27 Credit: 1,650,057 RAC: 0 |
I am running your v7 app and running non-BOINC projects on the CPU. The non-Boinc project is starving out the GPU. What can I do to make the SetiCuda client run high Priority? |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
I am running your v7 app and running non-BOINC projects on the CPU. The non-Boinc project is starving out the GPU. What can I do to make the SetiCuda client run high Priority? How exactly it starving out GPU ? What project? What symptoms ? CUDA MB is running on higher priority than other projects already. |
Westsail and *Pyxey* Send message Joined: 26 Jul 99 Posts: 338 Credit: 20,544,999 RAC: 0 |
I think what he means is this. When I tried running other projects there becomes a situation where due to tdf all cpu cores are running other projects so after CUDA completes a unit no new one load until BOINC finally decide to start another Seti unit. Currently the problem is being worked around by manual intervention for run other projects. When GPU go idle just restart Boinc client then suspend all but seti. Cuda start crunch again and everything good. Usually...usually..if only one non seti task at a time then everything work fine. Haven't had the scrtom to see how v8 and gpugrid play. I know with v7, obviously, it works fine. Good luck! "The most exciting phrase to hear in science, the one that heralds new discoveries, is not Eureka! (I found it!) but rather, 'hmm... that's funny...'" -- Isaac Asimov |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
Someone know when the official SETI@home CUDA >= V6.09 will be published? BUG-free and speed optimized [for VLARs or all ARs] ? |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
I think what he means is this. When I tried running other projects there becomes a situation where due to tdf all cpu cores are running other projects so after CUDA completes a unit no new one load until BOINC finally decide to start another Seti unit. Currently the problem is being worked around by manual intervention for run other projects. When GPU go idle just restart Boinc client then suspend all but seti. Cuda start crunch again and everything good. Usually...usually..if only one non seti task at a time then everything work fine. Haven't had the scrtom to see how v8 and gpugrid play. I know with v7, obviously, it works fine. Good luck! Yes... that's why I hope BOINC will do its job eventually, cause it's BOINC's job to do correct scheduling of all available apps... (including still missing project pairing ability and CPU + GPU app for the same tasks support ability). |
eaglescouter Send message Joined: 28 Dec 02 Posts: 162 Credit: 42,012,553 RAC: 0 |
One more time (for record ;) ): SSE3 version for AMD chips is updated. Here is example of minimal cc_config.xml file you need <cc_config> <options> <ncpus>enter_number_of_cores+1_value_here</ncpus> </options> </cc_config> You should put it in BOINCdata folder. |
Fred W Send message Joined: 13 Jun 99 Posts: 2524 Credit: 11,954,210 RAC: 0 |
Just had my first error using the V8 + CUDA. The sterr is: stderr out <core_client_version>6.6.0</core_client_version> <![CDATA[ <message> - exit code -12 (0xfffffff4) </message> <stderr_txt> setiathome_CUDA: Found 2 CUDA device(s): Device 1 : GeForce GTX 295 totalGlobalMem = 939524096 sharedMemPerBlock = 16384 regsPerBlock = 16384 warpSize = 32 memPitch = 262144 maxThreadsPerBlock = 512 clockRate = 1242000 totalConstMem = 65536 major = 1 minor = 3 textureAlignment = 256 deviceOverlap = 1 multiProcessorCount = 30 Device 2 : GeForce GTX 295 totalGlobalMem = 939524096 sharedMemPerBlock = 16384 regsPerBlock = 16384 warpSize = 32 memPitch = 262144 maxThreadsPerBlock = 512 clockRate = 1242000 totalConstMem = 65536 major = 1 minor = 3 textureAlignment = 256 deviceOverlap = 1 multiProcessorCount = 30 setiathome_CUDA: No device specified, determined to use CUDA device 1: GeForce GTX 295 SETI@home using CUDA accelerated device GeForce GTX 295 CPU-GPU team mod by Raistmer Priority normalize modification by Raistmer Priority of worker thread rised successfully Priority of process adjusted successfully Total GPU memory 939524096 free GPU memory 873117696 setiathome_enhanced 6.02 Visual Studio/Microsoft C++ libboinc: 6.3.22 Work Unit Info: ............... WU true angle range is : 2.575230 Optimal function choices: ----------------------------------------------------- name ----------------------------------------------------- v_BaseLineSmooth (no other) v_GetPowerSpectrum 0.00015 0.00000 v_ChirpData 0.01018 0.00000 v_Transpose4 0.00293 0.00000 FPU opt folding 0.00175 0.00000 SETI@home error -12 Unknown error cudaAcc_find_triplets erroneously found a triplet twice in find_triplets_kernel File: d:/BTR/seticuda/seti_boinc/client/cuda/cudaAcc_pulsefind.cu Line: 236 </stderr_txt> ]]> Not a complaint - so far this must be about a 0.01% failure rate (or even less). F. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
SETI@home error -12 Unknown error Thanks for report. It's known bug for 6.08 CUDA MB code, I passed bug report to CUDA dev already. This bug was observed with 6.08 stock too (as should be in most cases because CUDA-related codebase for stock and mods is the same). Keep posting if you will notice some another CUDA error, please. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.