Thread 'Ubuntu 14.04 - Nvidia GT740

Author	Message
David Lewis Send message Joined: 7 Mar 16 Posts: 7	Message 68171 - Posted: 7 Mar 2016, 22:07:04 UTC Hi, I run Einstein@home, can't seem to get GPU's to work for processing: test@test:~$ lspci \| grep -i vga 01:00.0 VGA compatible controller: NVIDIA Corporation GK107 [GeForce GT 740] (rev a1) test@test:~$ boinc 07-Mar-2016 16:56:15 [---] cc_config.xml not found - using defaults 07-Mar-2016 16:56:15 [---] Starting BOINC client version 7.2.42 for x86_64-pc-linux-gnu 07-Mar-2016 16:56:15 [---] log flags: file_xfer, sched_ops, task 07-Mar-2016 16:56:15 [---] Libraries: libcurl/7.35.0 OpenSSL/1.0.1f zlib/1.2.8 libidn/1.28 librtmp/2.3 07-Mar-2016 16:56:15 [---] Data directory: /home/david execv: No such file or directory 07-Mar-2016 16:56:15 [---] GPU detection failed. error code 512 07-Mar-2016 16:56:15 [---] No usable GPUs found 07-Mar-2016 16:56:15 [---] Host name: badass 07-Mar-2016 16:56:15 [---] Processor: 4 GenuineIntel Intel(R) Core(TM) i5-4690K CPU @ 3.50GHz [Family 6 Model 60 Stepping 3] 07-Mar-2016 16:56:15 [---] 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 ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm ida arat pln pts dtherm tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm xsaveopt 07-Mar-2016 16:56:15 [---] OS: Linux: 3.19.0-25-generic 07-Mar-2016 16:56:15 [---] Memory: 7.75 GB physical, 7.70 GB virtual 07-Mar-2016 16:56:15 [---] Disk: 102.33 GB total, 59.95 GB free 07-Mar-2016 16:56:15 [---] Local time is UTC -5 hours 07-Mar-2016 16:56:15 [---] No general preferences found - using defaults 07-Mar-2016 16:56:15 [---] Preferences: 07-Mar-2016 16:56:15 [---] max memory usage when active: 3967.71MB 07-Mar-2016 16:56:15 [---] max memory usage when idle: 7141.87MB 07-Mar-2016 16:56:15 [---] max disk usage: 59.85GB 07-Mar-2016 16:56:15 [---] don't use GPU while active 07-Mar-2016 16:56:15 [---] suspend work if non-BOINC CPU load exceeds 25% 07-Mar-2016 16:56:15 [---] (to change preferences, visit a project web site or select Preferences in the Manager) dir_open: Could not open directory 'slots' from '/home/david'. ^F07-Mar-2016 16:56:15 [---] GUI RPC bind to port 31416 failed: 98 07-Mar-2016 16:56:45 gstate.init() failed Error Code: -180 Any help would be much appreciated. Thanks. ID: 68171 ·

Agentb Send message Joined: 30 May 15 Posts: 265	Message 68172 - Posted: 7 Mar 2016, 22:40:25 UTC - in response to Message 68171. Last modified: 7 Mar 2016, 22:55:32 UTC Hi, I run Einstein@home, can't seem to get GPU's to work for processing: David hi Does this mean you have it working for CPU tasks? How have you installed nvidia drivers? Post nvidia-smi output if you can. Did you install boinc from repository? I'll assume so, if not then what i write below is off course. Assuming you did install from repo, then sudo /etc/init.d/boinc-client restart should restart. This test@test:~$ boinc 07-Mar-2016 16:56:15 [---] Data directory: /home/david execv: No such file or directory is not the optimal way to start boinc on a standard ubuntu install, boinc is complaining here it can not find it's data directory, let alone start looking for GPUs. If CPU boinc is working, the first 30 lines of the event log should help. (it will show the data directory etc) good luck. ID: 68172 ·

David Lewis Send message Joined: 7 Mar 16 Posts: 7	Message 68173 - Posted: 7 Mar 2016, 23:01:16 UTC - in response to Message 68172. I have working boinc cpu tasks, I am using boinc-manager to start it normally, not the boinc command. I just tried the boinc command becaues I knew it would give output about the gpu, I can't figure out how to see from the gui if it is using the gpu. I installed boinc packages from the repos. test@test:~$ nvidia-smi Mon Mar 7 18:00:35 2016 +------------------------------------------------------+ \| NVIDIA-SMI 352.63 Driver Version: 352.63 \| \|-------------------------------+----------------------+----------------------+ \| GPU Name Persistence-M\| Bus-Id Disp.A \| Volatile Uncorr. ECC \| \| Fan Temp Perf Pwr:Usage/Cap\| Memory-Usage \| GPU-Util Compute M. \| \|===============================+======================+======================\| \| 0 GeForce GT 740 Off \| 0000:01:00.0 N/A \| N/A \| \| 30% 34C P8 N/A / N/A \| 355MiB / 2047MiB \| N/A Default \| +-------------------------------+----------------------+----------------------+ +-----------------------------------------------------------------------------+ \| Processes: GPU Memory \| \| GPU PID Type Process name Usage \| \|=============================================================================\| \| 0 Not Supported \| +-----------------------------------------------------------------------------+ ID: 68173 ·

Agentb Send message Joined: 30 May 15 Posts: 265	Message 68174 - Posted: 7 Mar 2016, 23:37:23 UTC - in response to Message 68173. I have working boinc cpu tasks, I am using boinc-manager to start it normally, not the boinc command. I just tried the boinc command becaues I knew it would give output about the gpu, I can't figure out how to see from the gui if it is using the gpu. I installed boinc packages from the repos. OK goodo -that makes it a bit easier. Did you try a restart as below? From the manager <Ctrl><Shift>E should show the event log. The standard install has the data directory at /var/lib/boinc-client You will find a few things there. The important thing here is the event log is also stored in a file named stdoutdae.txt It will be revealing, please post the first 30-40 lines. ID: 68174 ·

David Lewis Send message Joined: 7 Mar 16 Posts: 7	Message 68175 - Posted: 7 Mar 2016, 23:55:42 UTC - in response to Message 68174. I restarted boinc as specified below. Looks like logs are in /var/lib/boinc-client. Here is the david@badass:/var/lib/boinc-client$ tail -n 60 stdoutdae.txt 07-Mar-2016 14:23:19 [---] Suspending computation - user request 07-Mar-2016 15:23:18 [---] Resuming computation 07-Mar-2016 16:44:20 [---] Suspending computation - user request 07-Mar-2016 16:44:47 [---] Resuming computation 07-Mar-2016 16:47:48 [---] Received signal 15 07-Mar-2016 16:47:49 [---] Exit requested by user 07-Mar-2016 16:48:17 [---] Starting BOINC client version 7.2.42 for x86_64-pc-linux-gnu 07-Mar-2016 16:48:17 [---] log flags: file_xfer, sched_ops, task 07-Mar-2016 16:48:17 [---] Libraries: libcurl/7.35.0 OpenSSL/1.0.1f zlib/1.2.8 libidn/1.28 librtmp/2.3 07-Mar-2016 16:48:17 [---] Data directory: /var/lib/boinc-client 07-Mar-2016 16:48:17 [---] No usable GPUs found 07-Mar-2016 16:48:17 [---] Host name: badass 07-Mar-2016 16:48:17 [---] Processor: 4 GenuineIntel Intel(R) Core(TM) i5-4690K CPU @ 3.50GHz [Family 6 Model 60 Stepping 3] 07-Mar-2016 16:48:17 [---] 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 ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm ida arat pln pts dtherm tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm xsaveopt 07-Mar-2016 16:48:17 [---] OS: Linux: 3.19.0-25-generic 07-Mar-2016 16:48:17 [---] Memory: 7.75 GB physical, 7.70 GB virtual 07-Mar-2016 16:48:17 [---] Disk: 102.33 GB total, 59.96 GB free 07-Mar-2016 16:48:17 [---] Local time is UTC -5 hours 07-Mar-2016 16:48:17 [---] Config: GUI RPCs allowed from: 07-Mar-2016 16:48:17 [Einstein@Home] URL http://einstein.phys.uwm.edu/; Computer ID 12211950; resource share 100 07-Mar-2016 16:48:17 [---] No general preferences found - using defaults 07-Mar-2016 16:48:17 [---] Reading preferences override file 07-Mar-2016 16:48:17 [---] Preferences: 07-Mar-2016 16:48:17 [---] max memory usage when active: 3967.71MB 07-Mar-2016 16:48:17 [---] max memory usage when idle: 7141.87MB 07-Mar-2016 16:48:17 [---] max disk usage: 59.88GB 07-Mar-2016 16:48:17 [---] don't use GPU while active 07-Mar-2016 16:48:17 [---] suspend work if non-BOINC CPU load exceeds 25% 07-Mar-2016 16:48:17 [---] (to change preferences, visit a project web site or select Preferences in the Manager) 07-Mar-2016 16:48:17 [---] gui_rpc_auth.cfg is empty - no GUI RPC password protection 07-Mar-2016 16:48:17 [---] Not using a proxy 07-Mar-2016 16:48:17 Initialization completed 07-Mar-2016 16:51:13 [---] Received signal 15 07-Mar-2016 16:51:13 [---] Exit requested by user 07-Mar-2016 16:51:26 [---] Starting BOINC client version 7.2.42 for x86_64-pc-linux-gnu 07-Mar-2016 16:51:26 [---] log flags: file_xfer, sched_ops, task 07-Mar-2016 16:51:26 [---] Libraries: libcurl/7.35.0 OpenSSL/1.0.1f zlib/1.2.8 libidn/1.28 librtmp/2.3 07-Mar-2016 16:51:26 [---] Data directory: /var/lib/boinc-client 07-Mar-2016 16:51:26 [---] No usable GPUs found 07-Mar-2016 16:51:26 [---] Host name: badass 07-Mar-2016 16:51:26 [---] Processor: 4 GenuineIntel Intel(R) Core(TM) i5-4690K CPU @ 3.50GHz [Family 6 Model 60 Stepping 3] 07-Mar-2016 16:51:26 [---] 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 ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm ida arat pln pts dtherm tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm xsaveopt 07-Mar-2016 16:51:26 [---] OS: Linux: 3.19.0-25-generic 07-Mar-2016 16:51:26 [---] Memory: 7.75 GB physical, 7.70 GB virtual 07-Mar-2016 16:51:26 [---] Disk: 102.33 GB total, 59.95 GB free 07-Mar-2016 16:51:26 [---] Local time is UTC -5 hours 07-Mar-2016 16:51:26 [---] Config: GUI RPCs allowed from: 07-Mar-2016 16:51:26 [Einstein@Home] URL http://einstein.phys.uwm.edu/; Computer ID 12211950; resource share 100 07-Mar-2016 16:51:26 [---] No general preferences found - using defaults 07-Mar-2016 16:51:26 [---] Reading preferences override file 07-Mar-2016 16:51:26 [---] Preferences: 07-Mar-2016 16:51:26 [---] max memory usage when active: 3967.71MB 07-Mar-2016 16:51:26 [---] max memory usage when idle: 7141.87MB 07-Mar-2016 16:51:26 [---] max disk usage: 59.87GB 07-Mar-2016 16:51:26 [---] don't use GPU while active 07-Mar-2016 16:51:26 [---] suspend work if non-BOINC CPU load exceeds 25% 07-Mar-2016 16:51:26 [---] (to change preferences, visit a project web site or select Preferences in the Manager) 07-Mar-2016 16:51:26 [---] gui_rpc_auth.cfg is empty - no GUI RPC password protection 07-Mar-2016 16:51:26 [---] Not using a proxy 07-Mar-2016 16:51:26 Initialization completed Thank you so much for responding so fast. ID: 68175 ·

Agentb Send message Joined: 30 May 15 Posts: 265	Message 68176 - Posted: 8 Mar 2016, 1:27:03 UTC - in response to Message 68175. 07-Mar-2016 16:48:17 [---] No usable GPUs found Try running sudo nvidia-modprobe first then a restart again. Then if that fails it's needing a few libraries i suspect. I don't think these are installed by default Ubuntu install. I normally get these from here ldconfig -p \| grep libOpenCL ldconfig -p \| grep libcuda should show where the host thinks these important libraries are. ID: 68176 ·

David Lewis Send message Joined: 7 Mar 16 Posts: 7	Message 68177 - Posted: 8 Mar 2016, 2:51:02 UTC - in response to Message 68176. I guess I don't have nvidia-modprobe as it couldn't be found. Looks like I have the libraries though right? david@badass:/var/lib/boinc-client$ ldconfig -p \| grep -i libopencl libOpenCL.so.1 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libOpenCL.so.1 libOpenCL.so (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libOpenCL.so david@badass:/var/lib/boinc-client$ sudo ldconfig -p \| grep libcuda libcuda.so.1 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libcuda.so.1 libcuda.so.1 (libc6) => /usr/lib/i386-linux-gnu/libcuda.so.1 libcuda.so (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libcuda.so libcuda.so (libc6) => /usr/lib/i386-linux-gnu/libcuda.so david@badass:/var/lib/boinc-client$ sudo nvidia-modprobe sudo: nvidia-modprobe: command not found [/code] ID: 68177 ·

Agentb Send message Joined: 30 May 15 Posts: 265	Message 68189 - Posted: 8 Mar 2016, 19:30:37 UTC - in response to Message 68177. I guess I don't have nvidia-modprobe as it couldn't be found. I guess you didn't try installing it? ID: 68189 ·

David Lewis Send message Joined: 7 Mar 16 Posts: 7	Message 68192 - Posted: 8 Mar 2016, 20:35:03 UTC - in response to Message 68189. Sorry, due to my ignorance I didn't think to just install it, lol. I guess this means it is working now? 08-Mar-2016 15:31:12 [---] Received signal 15 08-Mar-2016 15:31:13 [---] Exit requested by user 08-Mar-2016 15:31:26 [---] Starting BOINC client version 7.2.42 for x86_64-pc-linux-gnu 08-Mar-2016 15:31:26 [---] log flags: file_xfer, sched_ops, task 08-Mar-2016 15:31:26 [---] Libraries: libcurl/7.35.0 OpenSSL/1.0.1f zlib/1.2.8 libidn/1.28 librtmp/2.3 08-Mar-2016 15:31:26 [---] Data directory: /var/lib/boinc-client 08-Mar-2016 15:31:26 [---] CUDA: NVIDIA GPU 0: GeForce GT 740 (driver version unknown, CUDA version 7.5, compute capability 3.0, 2047MB, 1696MB available, 763 GFLOPS peak) 08-Mar-2016 15:31:26 [---] OpenCL: NVIDIA GPU 0: GeForce GT 740 (driver version 352.63, device version OpenCL 1.2 CUDA, 2047MB, 1696MB available, 763 GFLOPS peak) 08-Mar-2016 15:31:26 [---] Host name: badass 08-Mar-2016 15:31:26 [---] Processor: 4 GenuineIntel Intel(R) Core(TM) i5-4690K CPU @ 3.50GHz [Family 6 Model 60 Stepping 3] 08-Mar-2016 15:31:26 [---] 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 ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm ida arat pln pts dtherm tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm xsaveopt 08-Mar-2016 15:31:26 [---] OS: Linux: 3.19.0-25-generic 08-Mar-2016 15:31:26 [---] Memory: 7.75 GB physical, 7.70 GB virtual 08-Mar-2016 15:31:26 [---] Disk: 102.33 GB total, 59.84 GB free 08-Mar-2016 15:31:26 [---] Local time is UTC -5 hours 08-Mar-2016 15:31:26 [---] Config: GUI RPCs allowed from: 08-Mar-2016 15:31:26 [Einstein@Home] URL http://einstein.phys.uwm.edu/; Computer ID 12211950; resource share 100 08-Mar-2016 15:31:26 [---] No general preferences found - using defaults 08-Mar-2016 15:31:26 [---] Reading preferences override file 08-Mar-2016 15:31:26 [---] Preferences: 08-Mar-2016 15:31:26 [---] max memory usage when active: 3967.71MB 08-Mar-2016 15:31:26 [---] max memory usage when idle: 7141.87MB 08-Mar-2016 15:31:26 [---] max disk usage: 59.76GB 08-Mar-2016 15:31:26 [---] don't use GPU while active 08-Mar-2016 15:31:26 [---] suspend work if non-BOINC CPU load exceeds 25% 08-Mar-2016 15:31:26 [---] (to change preferences, visit a project web site or select Preferences in the Manager) 08-Mar-2016 15:31:26 [---] gui_rpc_auth.cfg is empty - no GUI RPC password protection 08-Mar-2016 15:31:26 [---] Not using a proxy 08-Mar-2016 15:31:26 Initialization completed 08-Mar-2016 15:31:26 [---] Suspending GPU computation - computer is in use What is nvidia-modprobe and why did that allow this to work? Just trying to figure out so that I will know for next time. Thanks. ID: 68192 ·

David Lewis Send message Joined: 7 Mar 16 Posts: 7	Message 68193 - Posted: 8 Mar 2016, 20:37:49 UTC - in response to Message 68192. Also, my only other questions is about this: 08-Mar-2016 15:31:26 [---] Suspending GPU computation - computer is in use Is there a good way to tell boinc to use GPU even if computer is in use until I tell it not to? Thanks. ID: 68193 ·

Juha Volunteer developer Volunteer tester Help desk expert Send message Joined: 20 Nov 12 Posts: 801	Message 68194 - Posted: 8 Mar 2016, 20:44:46 UTC - in response to Message 68192. Last modified: 8 Mar 2016, 20:45:39 UTC What is nvidia-modprobe and why did that allow this to work? It loads NVIDIA kernel modules and does other necessary setup work on behalf of the CUDA/OpenCL runtime. The runtime does not have the required privileges to do that by itself. ID: 68194 ·

Juha Volunteer developer Volunteer tester Help desk expert Send message Joined: 20 Nov 12 Posts: 801	Message 68195 - Posted: 8 Mar 2016, 20:48:57 UTC - in response to Message 68193. Is there a good way to tell boinc to use GPU even if computer is in use until I tell it not to? Did you try unticking the "Suspend GPU computing when computer is in use" checkbox in Computing preferences? ID: 68195 ·

David Lewis Send message Joined: 7 Mar 16 Posts: 7	Message 68196 - Posted: 8 Mar 2016, 21:19:22 UTC - in response to Message 68195. Is there a good way to tell boinc to use GPU even if computer is in use until I tell it not to? Did you try unticking the "Suspend GPU computing when computer is in use" checkbox in Computing preferences? I was able to find the option in the menu in the 'boinctui' program. Not sure if this is the normal way to do it or if there is a more standard way. I am happy about getting this working because I run Einstein@home, now using GPU, and GIMPS prime search using CPU, so they aren't fighting for CPU resources now. Thanks for all of the help. ID: 68196 ·

Juha Volunteer developer Volunteer tester Help desk expert Send message Joined: 20 Nov 12 Posts: 801	Message 68199 - Posted: 8 Mar 2016, 22:00:19 UTC - in response to Message 68196. Well, there's BOINC Manager. BOINC TUI is about the same as Manager expect it's Text UI and made by 3rd party. ID: 68199 ·

Agentb Send message Joined: 30 May 15 Posts: 265	Message 68202 - Posted: 8 Mar 2016, 22:37:17 UTC - in response to Message 68194. What is nvidia-modprobe and why did that allow this to work? It loads NVIDIA kernel modules and does other necessary setup work on behalf of the CUDA/OpenCL runtime. The runtime does not have the required privileges to do that by itself. Thanks Juha, I'll add $ man nvidia-modprobe ID: 68202 ·

Agentb Send message Joined: 30 May 15 Posts: 265	Message 68203 - Posted: 8 Mar 2016, 22:46:16 UTC - in response to Message 68196. I am happy about getting this working because I run Einstein@home, now using GPU, and GIMPS prime search using CPU, so they aren't fighting for CPU resources now. Thanks for all of the help. Thats good. Just be mindful that GPU tasks do need CPU resources to "feed" them. A rule of thumb is allow one free core per GPU card for good performance. You can get an idea of how fast the GPU can run by not running CPU tasks at all, then add CPU tasks. At E@H you can safely run two or three GPU tasks concurrently. Each task will run slower but the number of tasks overall should increase. Probably best to post E@H questions over on their forums. ID: 68203 ·

Agentb Send message Joined: 30 May 15 Posts: 265	Message 68266 - Posted: 13 Mar 2016, 8:47:40 UTC There has been a couple of bugs reported on this issue nvidia-graphics-drivers-* should recommend nvidia-modprobe and recently a fix has been realease on Trusty (14.04) here Always load all the nvidia modules ID: 68266 ·

Juha Volunteer developer Volunteer tester Help desk expert Send message Joined: 20 Nov 12 Posts: 801	Message 68270 - Posted: 13 Mar 2016, 20:08:13 UTC - in response to Message 68266. Excellent! Now we only need something similar for fglrx and Beignet. The fix is in 16.04 but I don't think it has made it 14.04 just yet, it's still listed as In progress for 14.04. ID: 68270 ·

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.

Thread 'Ubuntu 14.04 - Nvidia GT740 - Cuda Issue'