Message boards :
Number crunching :
Setting up Linux to crunch CUDA90 and above for Windows users
Message board moderation
Previous · 1 . . . 42 · 43 · 44 · 45 · 46 · 47 · 48 . . . 162 · Next
Author | Message |
---|---|
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Hope Juan sees your post and chimes in. I'm here, if you need any help just ask. I will happy to help if i know how to. Yes i was able to continue continue to add my old Total credit from my old windows host to the new Linux box when i switch the OS. But be aware that was not a real merge of both hosts. just switch the old host total credit to the new host. In my case that was really true since i not change the host itself just install the new OS. if you wish to try please let me know. But before try anything first you need to put your new Linux box to work. With a very small WU cache. If possible with resource share 0 (zero) |
J. Mileski Send message Joined: 9 Jun 02 Posts: 632 Credit: 172,116,532 RAC: 572 |
Is it possible to attach to a project and have NNT set so I could change settings without trashing WU's? |
J. Mileski Send message Joined: 9 Jun 02 Posts: 632 Credit: 172,116,532 RAC: 572 |
... snip If you have Nvidia graphics cards and the minimum required Compute Level, you can use the special app with CUDA 9.0. TBar has hinted that the special app with static CUDA 9.0 also might be imminent which is even faster than the current zi3v CUDA 90 app. ... Cuda 9.0 needs CC5.0? I think I have it 1/7/2018 1:37:52 PM | | cc_config.xml not found - using defaults 1/7/2018 1:37:52 PM | | Starting BOINC client version 7.8.3 for windows_x86_64 1/7/2018 1:37:52 PM | | log flags: file_xfer, sched_ops, task 1/7/2018 1:37:52 PM | | Libraries: libcurl/7.47.1 OpenSSL/1.0.2g zlib/1.2.8 1/7/2018 1:37:52 PM | | Data directory: C:\ProgramData\BOINC 1/7/2018 1:37:52 PM | | Running under account John 1/7/2018 1:37:56 PM | | CUDA: NVIDIA GPU 0: GeForce GTX 750 (driver version 388.00, CUDA version 9.1, compute capability 5.0, 1024MB, 824MB available, 1111 GFLOPS peak) 1/7/2018 1:37:56 PM | | OpenCL: NVIDIA GPU 0: GeForce GTX 750 (driver version 388.00, device version OpenCL 1.2 CUDA, 1024MB, 824MB available, 1111 GFLOPS peak) 1/7/2018 1:37:56 PM | | Host name: DellT320 1/7/2018 1:37:56 PM | | Processor: 12 GenuineIntel Intel(R) Xeon(R) CPU E5-2420 v2 @ 2.20GHz [Family 6 Model 62 Stepping 4] 1/7/2018 1:37:56 PM | | 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 popcnt aes f16c rdrandsyscall nx lm avx vmx smx tm2 dca pbe fsgsbase smep 1/7/2018 1:37:56 PM | | OS: Microsoft Windows 10: Professional x64 Edition, (10.00.16299.00) 1/7/2018 1:37:56 PM | | Memory: 15.96 GB physical, 18.33 GB virtual 1/7/2018 1:37:56 PM | | Disk: 297.60 GB total, 263.72 GB free 1/7/2018 1:37:56 PM | | Local time is UTC +0 hours 1/7/2018 1:37:56 PM | Einstein@Home | URL http://einstein.phys.uwm.edu/; Computer ID 12596569; resource share 0 1/7/2018 1:37:56 PM | SETI@home | URL http://setiathome.berkeley.edu/; Computer ID 8335913; resource share 100 1/7/2018 1:37:56 PM | SETI@home | General prefs: from SETI@home (last modified 04-Dec-2017 23:50:09) 1/7/2018 1:37:56 PM | SETI@home | Host location: none 1/7/2018 1:37:56 PM | SETI@home | General prefs: using your defaults 1/7/2018 1:37:56 PM | | Reading preferences override file 1/7/2018 1:37:56 PM | | Preferences: 1/7/2018 1:37:56 PM | | max memory usage when active: 8169.27 MB 1/7/2018 1:37:56 PM | | max memory usage when idle: 14704.69 MB 1/7/2018 1:37:56 PM | | max disk usage: 263.16 GB 1/7/2018 1:37:56 PM | | max CPUs used: 11 1/7/2018 1:37:56 PM | | (to change preferences, visit a project web site or select Preferences in the Manager) |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Is it possible to attach to a project and have NNT set so I could change settings without trashing WU's? Yes |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Cuda 9.0 needs CC5.0? I think I have it I'm a Linux Newbie but AFAIK if you DL the app from the link posted by Keith all the required libraries are included. Your GPU need to be CC5.0 compatible if you want to run the CUDA90 builds. Check the GPU compatibility on the posted link. If no you could try the CUDA60 version. http://www.arkayn.us/forum/index.php?PHPSESSID=214df9eb79c5f82848efb28f36110e43&topic=197.msg4499#msg4499 |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Your GTX 750 is on the mentioned list with CC 5.0 level, so yes you can use the special app. Not sure what you mean by setting NNT to make changes. The app_info will call different app and plan_classes so will trash any existing work onboard. There isn't a Linux Lunatics version of the installer that converts tasks. You need to complete your work cache before converting from Windows apps to Linux apps. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
J. Mileski Send message Joined: 9 Jun 02 Posts: 632 Credit: 172,116,532 RAC: 572 |
I like to take my time when attaching to a project. I really don't want to download tasks until I am ready. I wish the default when attaching to a project was NNT. I want to be in control. I have set NNT on the computer I am going to put Debian on. When I'm ready, I'm removing the windows hard drive and installing a new drive just for Linux. I do not plan on dual boot. |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
The trouble with setting NNT and then attaching to the project is that without the project sending you work, it won't send the apps necessary to run them. This is the case for a stock project installation. Doesn't apply to an anonymous platform where the apps to run a task are likely already have been downloaded outside of the project servers and custom app_info already written to use them. In your case with changing over to Anonymous and Linux platform, you will have satisfied the #2 scenario I described. You will already have the apps and app_info installed and simply waiting for you to Allow New Tasks to be set. Shouldn't be an issue. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
I think you might have a misunderstanding about where NNT is set. NNT is set and controlled only in the Manager. It is not a setting that is kept on the SETI servers. So when you install Debian and the new Linux BOINC installation, it will be in its default configuration with Tasks Allowed. The new BOINC installation will be virgin. You will need to set NNT first in the new Linux BOINC Manager, then attach to the SETI project. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
Cuda 9.0 needs CC5.0? I think I have it . . I can confirm that Cuda 60 works on CC 3.5 cards (and above) as I have run it on my GT730 ... which is the very bottom end of cards that have sufficient CC and resources to run the special sauce apps. . . But the 550ti cards do not have enough CCeven for Cuda60, they are not suitable, the 750/750ti are CC 5.0 and the 960 is CC 5.2 so they should support Cuda90. https://developer.nvidia.com/cuda-gpus Stephen . . |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
I like to take my time when attaching to a project. I really don't want to download tasks until I am ready. I wish the default when attaching to a project was NNT. I want to be in control. . . If you go back to the beginning of this thread you will see the way I configured my systems to run Linux from a flash drive and so leaving the original Windows drive untouched and uninvolved. Stephen |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
. . @ Juan, Brent, Keith and Grant, . . Thanks for your help guys, I am back running release 104, this time with lm-sensors installed and the temps are showing normal this time. I guess I was having kittens over nothing :) Stephen :) |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Good to hear, Stephen. Are you going to load the coretemp module for cpu tasks? Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
Good to hear, Stephen. Are you going to load the coretemp module for cpu tasks? . . Que? Stephen ? ? |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
In your previous post after installing lm-sensors, you stated you answered NO to the last question sensors-detect asked. Which was to add the coretemp module to your /etc/modules directory. That will install the module to your kernel so it is always available.
If you actually did later install the module, just ignore me. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
In your previous post after installing lm-sensors, you stated you answered NO to the last question sensors-detect asked. Which was to add the coretemp module to your /etc/modules directory. That will install the module to your kernel so it is always available. . . Aha! I am withyou now. Since there was no problem reading the CPU core temps either with psensor or by typing sensors I thought it to be unnecessary. I take it you think I should ... ?? Stephen ? |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
. . I just have to say this .... . . Creditscrew is yet again demonstrating it's inconsistency and proving it's ever growing irrelevance. With the current batch of Blc tasks being very fast running, CreditScrew is taking the opportunity to cut awarded credits in half, or thereabouts. Full length non-overflow tasks instead of being awarded 100 points (yes I know, it's been a long time since that happened) or even the 80 to 90 credits that has been the norm in recent months, are now being awarded 45-55 credits. Now there's an incentive to improve the productivity of our crunchers. Not! . . With the machines slogging their guts out in shocking heatwave conditions their RACS are now lower than ever, still producing very well but getting rated like slow machines processing halflings. Stephen :( |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
In your previous post after installing lm-sensors, you stated you answered NO to the last question sensors-detect asked. Which was to add the coretemp module to your /etc/modules directory. That will install the module to your kernel so it is always available. Well since I haven't run an Intel processor in over 20 years, I could be talking out my butt, and I still don't understand how psensor was able to read sensors without lm-sensors, but I understand that Coretemp gives the most accurate and detailed temps per core on Intel processors since they have individual sensors for each core. I also don't know if Coretemp JUST reads core temperatures. The sensors-detect in my experience picks up the Super I/O monitoring chip on motherboards which provide the sensor output for the voltages, fans and temps reported. The sensors-detect script normally picks up the sensor monitor and then offers to put that into your modules directory and then does the modprobe and ksmod to get the module added to your kernel so that it is available every boot. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
. . OK I tried that but it didn't seem to find the GTX1050ti card :- Module cpuid loaded successfully. Silicon Integrated Systems SIS5595... No VIA VT82C686 Integrated Sensors... No VIA VT8231 Integrated Sensors... No AMD K8 thermal sensors... No AMD Family 10h thermal sensors... No AMD Family 11h thermal sensors... No AMD Family 12h and 14h thermal sensors... No AMD Family 15h thermal sensors... No AMD Family 15h power sensors... No AMD Family 16h power sensors... No Intel digital thermal sensor... Success! (driver `coretemp') Intel AMB FB-DIMM thermal sensor... No VIA C7 thermal sensor... No VIA Nano thermal sensor... No snip Probing for Super-I/O at 0x2e/0x2f Trying family `National Semiconductor/ITE'... Yes Found unknown chip with ID 0x1911 Probing for Super-I/O at 0x4e/0x4f Trying family `National Semiconductor/ITE'... No Trying family `SMSC'... Yes Found unknown chip with ID 0x0b00ISA slots! Do you want to scan the ISA I/O ports? (YES/no): Probing for `National Semiconductor LM78' at 0x290... No Probing for `National Semiconductor LM79' at 0x290... No Probing for `Winbond W83781D' at 0x290... No Probing for `Winbond W83782D' at 0x290... No snip Do you want to probe the I2C/SMBus adapters now? (YES/no): Sorry, no supported PCI bus adapters found. Next adapter: NVIDIA i2c adapter 4 at 1:00.0 (i2c-0) Do you want to scan it? (yes/NO/selectively): yes Next adapter: NVIDIA i2c adapter 6 at 1:00.0 (i2c-1) Do you want to scan it? (yes/NO/selectively): yes Next adapter: NVIDIA i2c adapter 7 at 1:00.0 (i2c-2) Do you want to scan it? (yes/NO/selectively): yes snip To load everything that is needed, add this to /etc/modules: #----cut here---- # Chip drivers coretemp #----cut here---- If you have some drivers built into your kernel, the list above will contain too many modules. Skip the appropriate ones! Do you want to add these lines automatically to /etc/modules? (yes/NO)yes Successful! snip Monitoring programs won't work until the needed modules are loaded. You may want to run 'service kmod start' to load them. snip stephen@Mi-Burrito:~$ sudo service kmod start kmod stop/waiting . . Que?? snip stephen@Mi-Burrito:~$ sensors coretemp-isa-0000 Adapter: ISA adapter Core 0: +32.0°C (high = +76.0°C, crit = +100.0°C) Core 1: +32.0°C (high = +76.0°C, crit = +100.0°C) . . Arrrghh! It does not seem to recognise ANY sensors except the C2D core temps. Psensors alone does better than that !! . . ???? Stephen ? |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
OK, you definitely have another sensor besides the coretemp one. Probing for Super-I/O at 0x2e/0x2f Trying family `National Semiconductor/ITE'... Yes Found unknown chip with ID 0x1911 Probing for Super-I/O at 0x4e/0x4f Trying family `National Semiconductor/ITE'... No Trying family `SMSC'... Yes Found unknown chip with ID 0x0b00ISA slots! and also another chip via the ISA slots. So I would say you have a ITE Super I/O chip along with a SMSC chip. The ITE one should be easy to add, the SMSC one is problematic with what I have found so far on the web. But I really need to see what the ITE chip is identified as. But first lets get the coretemp module properly loaded. Type service module-init-tools startin Terminal and hit enter. That should unload modules and then reload them. Then try sensors again and see if you are now picking up the coretemp module. If it still doesn't see the module try sudo modprobe coretemp I would also have you look in two directories /etc/modprobe.d and /etc/modules-load.d. You should have a file named coretemp.conf in both directories. You should also find the driver in ~~"your kernel version number"/kernel/drivers/hwmon/"coretemp.ko" (guessing this would likely be the driver name" I would not think you need to force the coretemp module as it it a long known and familiar one and should be instantly recognized and loaded. I would like to probe for the other Super I/O ITE chip, but am unsure what name it is from your post. If I could figure out what the driver name is, then we could force it at address 0x1911 Anyway after you try the modprobe for coretemp, post back the results. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
©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.