Message boards :
Number crunching :
NVIDIA GeForce GTX 1060 3GB
Message board moderation
Author | Message |
---|---|
Donald Bosso-Locquiao Send message Joined: 22 Dec 01 Posts: 18 Credit: 132,836 RAC: 0 |
I have a NVIDIA GeForce GTX 1060 3GB. Where in settings do I tell BOINC to use the CUDA to munch on Seti@HOME projects? Thanks :) |
betreger Send message Joined: 29 Jun 99 Posts: 11358 Credit: 29,581,041 RAC: 66 |
You should use SOG with that card. It will crunch a lot more work. |
Donald Bosso-Locquiao Send message Joined: 22 Dec 01 Posts: 18 Credit: 132,836 RAC: 0 |
SOG? |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Signals on GPU...It's the short way of saying the OpenCl application for the GPU. There are several different applications that will run on your graphic card, there are cuda applications and there are openCl application. I think they were saying, you should look at installing the lunatics optimized applications and select SoG in the options during the installation http://mikesworld.eu/download.html 0.45 Beta 6 is the last updated version. During the installation, you need to make sure the selections are correct. If you don't, it will install cuda 32 for you which is not what you want. This is an advance method of controlling your GPU and you deciding which applications run on your computer Zalster |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
I have a NVIDIA GeForce GTX 1060 3GB. Where in settings do I tell BOINC to use the CUDA to munch on Seti@HOME projects? My previous post was in response to your last. As far as your original question. The server will send work to your computer and test the different applications on your computer (as long as you have selected use GPU in your preference from your member page, not on the BOINC Manager). It will go through all the versions running multiple work units until it decides which application it thinks is the fastest for you system. Zalster |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13161 Credit: 1,160,866,277 RAC: 1,873 |
I have a NVIDIA GeForce GTX 1060 3GB. Where in settings do I tell BOINC to use the CUDA to munch on Seti@HOME projects? Which could take months for the servers to decide. It takes 11 validated, non-overflow tasks per application before the servers can decide which application MIGHT be fastest on your host. Fastest way to get your host producing most efficiently is to install the Lunatics application and make sure to choose the SoG app on the Nvidia application chooser page. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
I have a NVIDIA GeForce GTX 1060 3GB. Where in settings do I tell BOINC to use the CUDA to munch on Seti@HOME projects? Also make sure that you have a CPU core reserved to feed that GPU, both my dual 3GB 1060 rigs reserve 2 CPU cores to support both GPU's. It's also more helpful not to hide your computers as more help can be had when we can see them, take a look at mine and you'll see that no private details are shown. ;-) Cheers. |
Donald Bosso-Locquiao Send message Joined: 22 Dec 01 Posts: 18 Credit: 132,836 RAC: 0 |
I am not hiding it. I am 90% blind (low vision) and cannot figure out how to unhide it. |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
I am not hiding it. I am 90% blind (low vision) and cannot figure out how to unhide it. Go to https://setiathome.berkeley.edu/home.php Click on seti@home preferences in the center of the page In the primary (default) settings look for Should SETI@home show your computers on its web site? Click edit at the bottom of this set of preferences and click the box next to show your computers. Save It may take up to 1 hour for the changes to occur. |
betreger Send message Joined: 29 Jun 99 Posts: 11358 Credit: 29,581,041 RAC: 66 |
Which could take months for the servers to decide. It takes 11 validated, non-overflow tasks per application before the servers can decide which application MIGHT be fastest on your host. Keith a dedicates 1060 would figure that out in a few days or so. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
Which could take months for the servers to decide. It takes 11 validated, non-overflow tasks per application before the servers can decide which application MIGHT be fastest on your host. It takes just over a week, but you will still get the others sent once a week after that to be sure that the servers picked the right app and you won't get the AVX CPU app running stock. ;-) Cheers. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13720 Credit: 208,696,464 RAC: 304 |
Which could take months for the servers to decide. It takes 11 validated, non-overflow tasks per application before the servers can decide which application MIGHT be fastest on your host. Depending on the work mix, the Manager can still select the wrong application. On Beta, my GTX 1070s ended up with CUDA42 as it mostly had Arecbo WUs where as SoG & CUDA50 had mostly GBT. Grant Darwin NT |
Donald Bosso-Locquiao Send message Joined: 22 Dec 01 Posts: 18 Credit: 132,836 RAC: 0 |
Okay. I have listed my pc as directed above. Any further suggestions?> it's a gaming PC but I got it just to run Set@Home faster as my Xmass present to myself. Thanks. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13720 Credit: 208,696,464 RAC: 304 |
Task manager in Win 7 shows all 4 cores loaded evenly. 3 processing CPU WUs and 1 supporting the GPU gives 4 cores total. Grant Darwin NT |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13720 Credit: 208,696,464 RAC: 304 |
Okay. I have listed my pc as directed above. As others have said, the Lunatics Beta 6 installer will let you run the AVX application for the CPU, which will give a 30%+ boost in CPU output. As things are, running stock, if you put these command line values in the mb_cmdline-8.22_windows_intel__opencl_nvidia_SoG.txt file found in the C:\ProgramData\BOINC\projects\setiathome.berkeley.edu folder -hp -period_iterations_num 30 -high_perf -sbs 1024 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64 you will get a significant boost in GPU output (or use Wiggos's values that he uses for his GTX 1060s if he posts them). If your screen becomes laggy, characters go missing when typing etc , then increase the -period_iterations_num from 30 to 50 & see if that helps, and/or remove the -high_perf option. There is no need to update or exit/restart for the values to take effect. When the next WU starts processing, it will pick up the current values in that file. Also if you put a file called app_config.xml in the same folder, and put the following settings in there, it will reserve 1 CPU core for each GPU WU being processed. If you run out of GPU work (pretty much every weekly outage) then that core will just switch back to processing CPU work. <app_config> <app> <name>setiathome_v8</name> <gpu_versions> <gpu_usage>1.00</gpu_usage> <cpu_usage>1.00</cpu_usage> </gpu_versions> </app> <app> <name>astropulse_v7</name> <gpu_versions> <gpu_usage>0.5</gpu_usage> <cpu_usage>1.0</cpu_usage> </gpu_versions> </app> </app_config> You need to exit/restart BOINC, or in the Manager select Options, Read config files, for these settings to take effect. Grant Darwin NT |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
I use, -sbs 1024 -period_iterations_num 10 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64 in my mb_cmdline_win_x86_SSE3_OpenCL_NV_SoG.txt file and -unroll 10 -oclFFT_plan 256 16 512 -ffa_block 12288 -ffa_block_fetch 6144 in my ap_cmdline_win_x86_SSE2_OpenCL_NV.txt file. Cheers. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13720 Credit: 208,696,464 RAC: 304 |
Task manager in Win 7 shows all 4 cores loaded evenly.That is correct BUT if you then snooze the GPU , all 4 cores are still evenly loaded. But that is what was suggested would happen if you run out of GPU work. I'll try Grants App_config.xml. It will give the same result. The Core that was supporting the GPU will start to process CPU work. When the GPU restarts, that CPU WU (or one of the others) will pause when the core supports the GPU again. Grant Darwin NT |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13720 Credit: 208,696,464 RAC: 304 |
What seems to be happening is that if you snooze the GPU or it runs out of work, Boinc adds an extra CPU task to the mix until GPU work comes back. Is it supposed to do that? Well, yeah. You've set a maximum % for CPU usage. When the GPU stops processing, the CPU usage drops below that setting, so it picks up another CPU Work Unit to meet your settings. I'd suggest using Process Explorer (or similar) to see what's happening as it groups all related processes together making it easier to see what's using what & how much. With Task Manager, they tend to be spread out making it harder to see the overall picture. Grant Darwin NT |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13720 Credit: 208,696,464 RAC: 304 |
Process explorer? Ok, thanks I'll give it a try. Yep, best to get it from the source. https://docs.microsoft.com/en-us/sysinternals/downloads/process-explorer Grant Darwin NT |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
is there a script to force a CPU core to a GPU? If I have an 4 core CPU but I only run 3 CPU tasks That in theory should leave 1 core free for a GPU except it doesn't. Task manager in Win 7 shows all 4 cores loaded evenly. So the GPU gets a bit of each of the cores not one to itself. You can modify the app_config to include CPU usage, then to a max concurrent to 3 that will limit only 3 CPU work units and leave 1 cpu core for the GPU. Of course, you need to change use at most back up to 100% I have it somewhere in an old folder, will need to dig it up for you. |
©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.