Message boards :
Number crunching :
Running SETI@home on an nVidia Fermi GPU
Message board moderation
Previous · 1 . . . 8 · 9 · 10 · 11 · 12 · 13 · 14 . . . 15 · Next
Author | Message |
---|---|
Speedy Send message Joined: 26 Jun 04 Posts: 1643 Credit: 12,921,799 RAC: 89 |
fermi app is running incredibly slow, well at least i think. take a look. I'm running the same driver & I get same read out as above. 4/07/2010 9:57:57 a.m. NVIDIA GPU 0: GeForce GTX 470 (driver version 25721, CUDA version 3010, compute capability 2.0, 1248MB, 726 GFLOPS peak) Boinc .56 64 bit http://setiathome.berkeley.edu/show_host_detail.php?hostid=5417843 |
Rich-E Send message Joined: 23 Feb 01 Posts: 41 Credit: 1,685,487 RAC: 0 |
Re-posting, by popular request, from CUDA MB V12b rebuild supposed to work with Fermi GPUs. Good morning. I just picked up a GTX480 to try the thing out on SETI. The BOINC manager gets stuck at ~4.4% while downloading cufft32_30_14.dll. At the advice of another thread, I came here and attempted to download the files above by hand. All but cufft32_30_14.dll will download. It gets to about 370K and then stalls. Is there an alternate location for this .dll or is there some setting in BOINC or SETI that I can set that might be interfering with the simple download? I would prefer not running optimized in the beginning, so that I can see what success looks like. But for now that is where I am stuck. Any help would be appreciated. In case it helps: Motherboard: ASUS P6T7 GPU: EVGA GTX480 BOINC Message Info: 7/10/2010 6:39:38 AM Starting BOINC client version 6.10.56 for windows_x86_64 7/10/2010 6:39:38 AM log flags: file_xfer, sched_ops, task 7/10/2010 6:39:38 AM Libraries: libcurl/7.19.7 OpenSSL/0.9.8l zlib/1.2.3 7/10/2010 6:39:38 AM Data directory: D:\ProgramData\BOINC 7/10/2010 6:39:38 AM Running under account SeyfertR 7/10/2010 6:39:38 AM Processor: 8 GenuineIntel Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz [Family 6 Model 26 Stepping 5] 7/10/2010 6:39:38 AM Processor: 256.00 KB cache 7/10/2010 6:39:38 AM Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 syscall nx lm vmx tm2 popcnt pbe 7/10/2010 6:39:38 AM OS: Microsoft Windows 7: Ultimate x64 Edition, (06.01.7600.00) 7/10/2010 6:39:38 AM Memory: 5.99 GB physical, 11.98 GB virtual 7/10/2010 6:39:38 AM Disk: 558.76 GB total, 496.75 GB free 7/10/2010 6:39:38 AM Local time is UTC -4 hours 7/10/2010 6:39:38 AM NVIDIA GPU 0: GeForce GTX 480 (driver version 25721, CUDA version 3010, compute capability 2.0, 1503MB, 778 GFLOPS peak) Take care and have a good day. Rich |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
I just picked up a GTX480 to try the thing out on SETI. The BOINC manager gets stuck at ~4.4% while downloading cufft32_30_14.dll. At the advice of another thread, I came here and attempted to download the files above by hand. All but cufft32_30_14.dll will download. It gets to about 370K and then stalls. Try it from the Beta project: http://boinc2.ssl.berkeley.edu/beta/download/cufft32_30_14.dll or the Main Project: http://boinc2.ssl.berkeley.edu/sah/download_fanout/cufft32_30_14.dll Claggy Edit: I've just downloaded it succesfully from the Beta project using Getright, it'll resume downloads if they stall. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14677 Credit: 200,643,578 RAC: 874 |
Good morning. Polite questions get quick answers! It should be possible to get the CUDA DLLs by simple browser download, either from the main project download fanout (which is where BOINC will be looking), or from the Beta site I linked in that opening post. If it doesn't work try, try, and try again. Alternatively, the DLLs are in the NVidia CUDA 3.0 toolkit (big download, but a more reliable server). I think this is the one: http://www.nvidia.com/object/thankyou.html?url=/compute/cuda/3_0/toolkit/cudatoolkit_3.0_win_32.exe |
Rich-E Send message Joined: 23 Feb 01 Posts: 41 Credit: 1,685,487 RAC: 0 |
Success! Thank-you both very much for your help. I must have a Windows issue, as all three links that Claggy proposed, did not complete a download; however, the NVidia link worked without issues and I was able to located the necessary files. I have to run to work now, but will put it all together tonight and hopefully start exercising this GPU. Take care and thanks again. Rich |
sarmitage Send message Joined: 2 Dec 09 Posts: 56 Credit: 1,123,857 RAC: 0 |
Is there a way to allow the CPU->GPU rescheduling if you don't have an app_info.xml file? Can you create a minimal app_info.xml that contains only the one alias, or do I need to somehow get the entries for the rest of the (stock) apps as well? -Scott |
sarmitage Send message Joined: 2 Dec 09 Posts: 56 Credit: 1,123,857 RAC: 0 |
Is there a way to allow the CPU->GPU rescheduling if you don't have an app_info.xml file? Can you create a minimal app_info.xml that contains only the one alias, or do I need to somehow get the entries for the rest of the (stock) apps as well? I'm going to just use the manual method for now, one-off. Get the app_info stuff sorted out tomorrow. -S |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14677 Credit: 200,643,578 RAC: 874 |
If you have already (in the past) downloaded CPU work direct from Berkeley, you should have the stock CPU application and information about how to use it (<app_version>) already on your machine. In that case, ReSchedule should work. But if you have only ever downloaded GPU work since the last time you installed/reset SETI, BOINC won't know what to do with the rescheduled work, and will ditch it. Most people round here who get this interested have already installed an optimised CPU app, because it makes a huge difference (improvement) in processing efficiency. So they have an app_info file already, and the question has never really arisen before. You could be the first to try it! |
Rich-E Send message Joined: 23 Feb 01 Posts: 41 Credit: 1,685,487 RAC: 0 |
Good morning again. Everything seems to be running well as of this morning. I had to do a dance to get the system to see that the file was there, without it restarting the download and over-writing the file. I also upgraded to 6.10.58. I finished my first GTX 480 CUDA task about 2 hrs ago. They are completing in about 8 min and 46 sec! Is that about right or are these the same error'ed results I have been reading about? If the results are legit I may need to pick up an AMD board and see how it compares. I must say that that is a pretty fine performance increase. Take care, have a good day and thanks for the help getting over the hurdle. Rich Happy Fermi Cruncher |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Good morning Rich-E, Everything looks fine as far as your Fermi card. You had one -9 that I saw but it was legitimate, the rest I checked seemed to be good to go. One thing I would like to recommend is to get your I7 optimized. This is your speed now..18,321.07 17,911.41 for a .43 midrange WU. My little E5400 turns in times like this...10,372.31 9,205.35 on a .426 angle range. Once you get optimized you should leave me in the dust. PROUD MEMBER OF Team Starfire World BOINC |
sarmitage Send message Joined: 2 Dec 09 Posts: 56 Credit: 1,123,857 RAC: 0 |
Good morning Rich-E, There seem to be many different threads on this out there, and I don't want to mess up my cache with the long outage coming up. Could you point me to a canonical README thread that goes over all the ups and downs that may be encountered when installing optimized apps? (I like to be prepared). Thanks! -Scott |
sarmitage Send message Joined: 2 Dec 09 Posts: 56 Credit: 1,123,857 RAC: 0 |
If you have already (in the past) downloaded CPU work direct from Berkeley, you should have the stock CPU application and information about how to use it (<app_version>) already on your machine. In that case, ReSchedule should work. It looks like the info that would get stored in app_info.xml (if I had optimized apps) is currently in client_state.xml. I plan to copy the follow app_version section, changing only the plan_class to read "cuda" instead of "cuda_fermi". Look OK to everyone? <app_version> <app_name>setiathome_enhanced</app_name> <version_num>610</version_num> <platform>windows_intelx86</platform> <avg_ncpus>0.714854</avg_ncpus> <max_ncpus>0.714854</max_ncpus> <flops>267081085618.626400</flops> <plan_class>cuda_fermi</plan_class> <api_version>6.3.22</api_version> <file_ref> <file_name>setiathome_6.10_windows_intelx86__cuda_fermi.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>libfftw3f-3-1-1a_upx.dll</file_name> <open_name>libfftw3f-3-1-1a_upx.dll</open_name> </file_ref> <file_ref> <file_name>seti_609.jpg</file_name> <open_name>seti_logo</open_name> </file_ref> <file_ref> <file_name>cudart32_30_14.dll</file_name> <open_name>cudart.dll</open_name> <copy_file/> </file_ref> <file_ref> <file_name>cufft32_30_14.dll</file_name> <open_name>cufft.dll</open_name> <copy_file/> </file_ref> <file_ref> <file_name>setiathome-6.10_cuda_AUTHORS</file_name> <open_name>setiathome-6.10_cuda_AUTHORS</open_name> </file_ref> <file_ref> <file_name>setiathome-6.10_cuda_COPYING</file_name> <open_name>setiathome-6.10_cuda_COPYING</open_name> </file_ref> <file_ref> <file_name>setiathome-6.10_cuda_COPYRIGHT</file_name> <open_name>setiathome-6.10_cuda_COPYRIGHT</open_name> </file_ref> <file_ref> <file_name>setiathome-6.10_cuda_README</file_name> <open_name>setiathome-6.10_cuda_README</open_name> </file_ref> <coproc> <type>CUDA</type> <count>1.000000</count> </coproc> <gpu_ram>209715200.000000</gpu_ram> </app_version> -Scott |
Speedy Send message Joined: 26 Jun 04 Posts: 1643 Credit: 12,921,799 RAC: 89 |
If you have already (in the past) downloaded CPU work direct from Berkeley, you should have the stock CPU application and information about how to use it (<app_version>) already on your machine. In that case, ReSchedule should work. Follow message 1006793 it worked for me. |
Rich-E Send message Joined: 23 Feb 01 Posts: 41 Credit: 1,685,487 RAC: 0 |
Good morning Rich-E, Good morning. I took a look at the Lunatic's site and I'm willing to put in the time. One overall question though before I start drilling: If I want to run optimized CPU and Fermi, am I loading the Unified Installer and then adding the CUDA Multi-beam, OR is Fermi now included in the Unified Installer OR is Fermi a manual process defined by this thread. I read the FAQ section but I seemed to be having a senior moment this morning. Thanks and have a good day. Rich |
TheFreshPrince a.k.a. BlueTooth76 Send message Joined: 4 Jun 99 Posts: 210 Credit: 10,315,944 RAC: 0 |
Good morning Rich-E, As far as I know you still have to modify the app_info.xml after you used the Lunatics unified-installer if you have a Fermi card. You also need to download the files that are described in the app_info.xml. Rig name: "x6Crunchy" OS: Win 7 x64 MB: Asus M4N98TD EVO CPU: AMD X6 1055T 2.8(1,2v) GPU: 2x Asus GTX560ti Member of: Dutch Power Cows |
TheFreshPrince a.k.a. BlueTooth76 Send message Joined: 4 Jun 99 Posts: 210 Credit: 10,315,944 RAC: 0 |
Good morning Rich-E, As far as I know you still have to modify the app_info.xml after you used the Lunatics unified-installer if you have a Fermi card. You also need to download the files that are described in the app_info.xml. For the files, follow the link (message in this thread): http://setiathome.berkeley.edu/forum_thread.php?id=60079&nowrap=true#1014270 Rig name: "x6Crunchy" OS: Win 7 x64 MB: Asus M4N98TD EVO CPU: AMD X6 1055T 2.8(1,2v) GPU: 2x Asus GTX560ti Member of: Dutch Power Cows |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
I took a look at the Lunatic's site and I'm willing to put in the time. One overall question though before I start drilling: If I want to run optimized CPU and Fermi, am I loading the Unified Installer and then adding the CUDA Multi-beam, OR is Fermi now included in the Unified Installer OR is Fermi a manual process defined by this thread. I read the FAQ section but I seemed to be having a senior moment this morning. The Lunatics installer does not yet have a CUDA Fermi application. As you've been running the stock Fermi application you already have the files needed, so you can ignore well-meant advice on where to download them. You will need the other information from the second post in this thread, specifically the information to be added to app_info.xml. At the beginning of the outage it seems you had 23 CUDA tasks in progress, so I'll assume those have been completed and are ready to upload when the outage ends. That should simplify the transition a bit. First, shut down BOINC completely and use Task Manager to be sure boinc.exe is no longer running. Then you can use the Lunatics installer and select only the appropriate AK_v8b CPU S@H Enhanced application plus the Astropulse CPU application. After the install finishes, add the Fermi information to app_info.xml. The needed files should still be in place, just check that every filename in the app_info.xml has a matching file. Then you can restart BOINC. Joe |
hbomber Send message Joined: 2 May 01 Posts: 437 Credit: 50,852,854 RAC: 0 |
Any alternative download location for setiathome_6.10_windows_intelx86__cuda_fermi.exe? |
TheFreshPrince a.k.a. BlueTooth76 Send message Joined: 4 Jun 99 Posts: 210 Credit: 10,315,944 RAC: 0 |
Besides my GTX470 I have also a 9800GTX+ that's unused at the moment. Does anyone know if I can ad the 9800GTX+ to my rig and still use the opt apps? What changes should be made to the app_info.xml? I do not have the right PSU at the moment but if you say it should work, I could go out and buy a 850Watt Corsair. If you say it doesn't work, I could go out and buy a GTX460 for some extra crunching power :P Rig name: "x6Crunchy" OS: Win 7 x64 MB: Asus M4N98TD EVO CPU: AMD X6 1055T 2.8(1,2v) GPU: 2x Asus GTX560ti Member of: Dutch Power Cows |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14677 Credit: 200,643,578 RAC: 874 |
Besides my GTX470 I have also a 9800GTX+ that's unused at the moment. I don't actually know: it would be interesting to find out. My rigs are all on cheap motherboards, bought before CUDA came on the scene, and they each only have a single full-size PCIe slot, so I can't do the full test. In theory, the v6.10 Fermi application is backwards-compatible and should run on the 9800GTX+. That bit I can test later this afternoon, if you want, before you spend any money. If not, there's certainly a Lunatics application in Beta testing that will work. So no change needed in app_info. One thing you would need to do is to create/modify a simple cc_config.xml file. That's been covered many times - see message 1015382 for the most recent example. |
©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.