Setting up a Linux machine to crunch CUDA80 for Windows users

Message boards : Number crunching : Setting up a Linux machine to crunch CUDA80 for Windows users
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 45 · 46 · 47 · 48 · 49 · 50 · 51 . . . 59 · Next

AuthorMessage
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1916462 - Posted: 2 Feb 2018, 10:14:07 UTC - in response to Message 1916385.  

I agree with your assessment TBar. I thought Stephen had already performed the obvious and suggested. I would have reinstalled by now instead of beating my head against the wall.


. . Well maybe I am slow but I like to eliminate the easiest solutions before resorting to acts of desperation.

. . I was about to edit app_info and wipe out the section on running tasks, which I believe was intended to get the app to see the new -sbs value of 256, when I rechecked stderr and saw it already had. So I asked the question, that being the case is there any point in doing that part. Which either no one has read or simply not bothered to respond to. If there is any other purpose in that step then I will do that before going to the next phase of replacing all the AP related files. Apart from hopefully avoiding the more deperate steps it might give more insight into the original cause. Do I have corruption issues on the drive? Did I cause the problem with a change I may have made (despite not remembering it) or was there some other cause. The simplest cure would be to abort the 5 tasks and let some other host have them (I can see the queue forming for that) but that would not eliminate the problem at all, only leave it ro recur if I were to receive any other AP tasks.

Stephen

! !
ID: 1916462 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 4964
Credit: 318,860,332
RAC: 798,713
United States
Message 1916510 - Posted: 2 Feb 2018, 18:18:48 UTC - in response to Message 1916462.  

Still a good idea to delete the running tasks from app_info, even if the -sbs change seems to have occurred. You need to get BOINC to look at the tasks as brand new, never been seen.

See if that changes the results. Then if still the same, I would unpack the original package and copy and paste the files into your folder.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1916510 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 4964
Credit: 318,860,332
RAC: 798,713
United States
Message 1916517 - Posted: 2 Feb 2018, 19:26:54 UTC - in response to Message 1916510.  

I meant delete the tasks from client_state.xml.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1916517 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1916550 - Posted: 3 Feb 2018, 0:11:17 UTC - in response to Message 1916510.  

Still a good idea to delete the running tasks from app_info, even if the -sbs change seems to have occurred. You need to get BOINC to look at the tasks as brand new, never been seen.

See if that changes the results. Then if still the same, I would unpack the original package and copy and paste the files into your folder.


. . Yep, I did that this morning, and made sure the existing slots were all cleared but the behaviour remains the same.

. . I guess it has reached defcon desperate ....

Stephen

:(
ID: 1916550 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1916551 - Posted: 3 Feb 2018, 0:12:49 UTC - in response to Message 1916517.  

I meant delete the tasks from client_state.xml.


. . No probs, I understood. TBar's instructions were very clear on that point.

Stephen

:)
ID: 1916551 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1916774 - Posted: 4 Feb 2018, 1:20:23 UTC
Last modified: 4 Feb 2018, 1:30:31 UTC

. . Well looking back to the original downloads has proven TBar was correct that I must have edited the command line text file, not just because -sbs was 512 rather than the original 256, but because unroll was set to 13 for the 970s, not 12 as in the download. But that does explain why I sought out and edited that file, to change the unroll. It also pins it to a time, when I originally installed the app. I guess the devil made me change -sbs to 512 though I am surprised that I did not use 1024. Not surprisingly, changing the values back to defaults did not fix anything. So after removing running tasks from client_state and returning those values to default, there is no change with the problem which pretty much proves that was not the issue no matter how much TBar likes to disparage people. This is no surprise to me because those changes would have been done before I received and successfully ran AP tasks.

. . So desperation measures. I copied all the AP related files, including app_info, from the downloads to the project directory. I may have transgressed against the Magnificant TBar though as I did re-instate the -nobs and -pfb 32 parameters in app_info. But thankfully this has not fixed the problem either, so at least I do not have corruption issues on the drive.

. . So far 0 for 4. Anyone else have any theories? I am sure TBar wii find some new reason to denigrate me but hey, maybe he will have a solution next time.

. . I am toying with the idea it may have been caused by the failed attempt to update Linux to release 104. Each of the 3 attempts to update since release 96 have resulted in being unable to boot Linux from the flashdrive, but thankfully returning to release 96 each time has been successful. But there may have been some common files that were changed and remain so despite the fallback. Such as Firefox which was updated and remains so, sadly, as the new version is a bit of a CPU hog. Could such a change have eliminated some dependency that affects the AP app?

Stephen

? ?
ID: 1916774 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 4964
Credit: 318,860,332
RAC: 798,713
United States
Message 1916777 - Posted: 4 Feb 2018, 1:35:21 UTC - in response to Message 1916774.  

Sorry to hear you still don't have any success Stephen. OK, here is something to try that hasn't been suggested yet. Have you checked the dependencies on the AP executable yet? I'm talking about the astropulse_7.08_x86_64-pc-linux-gnu__opencl_nvidia_100 application file. Do this:

Open a Terminal session and type

ldd (and a space)


Now open up the BOINC/projects/setiathome.berkeley.edu directory in Nautilus and find the astropulse_7.08_x86_64-pc-linux-gnu__opencl_nvidia_100 file. Now click on that file and drag and drop it into the Terminal session. Now make the Terminal session have focus by clicking on the Terminal title bar and then hit Enter on the keyboard. The ldd program will print out the dependencies that the program needs to run. Any missing dependencies will show up at the top of the printout. Notice the libOpenCL.so.1 => /usr/lib/x86_64-linux-gnu/libOpenCL.so.1 (0x00007fe425cfc000) file is listed in this output from my system.

keith@Darksider:~$ ldd '/home/keith/Desktop/BOINC_7.8/projects/setiathome.berkeley.edu/astropulse_7.08_x86_64-pc-linux-gnu__opencl_nvidia_100' 
	linux-vdso.so.1 =>  (0x00007ffecd7df000)
	libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fe426124000)
	libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fe425f07000)
	libOpenCL.so.1 => /usr/lib/x86_64-linux-gnu/libOpenCL.so.1 (0x00007fe425cfc000)
	libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007fe425ae2000)
	libnsl.so.1 => /lib/x86_64-linux-gnu/libnsl.so.1 (0x00007fe4258c9000)
	libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fe425547000)
	libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fe42523e000)
	libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fe424e74000)
	/lib64/ld-linux-x86-64.so.2 (0x00007fe426328000)
	libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fe424c5e000)
keith@Darksider:~$ 


Since you are having issue compiling the wisdom file from the opencl source file, I wonder if you have a missing dependency. This might show an issue or show nothing out of ordinary. Something new to check and try. YMMV.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1916777 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 4138
Credit: 249,136,934
RAC: 334,823
United States
Message 1916779 - Posted: 4 Feb 2018, 1:58:31 UTC

I looks as though you missed this part again;
If that doesn't solve it, you probably have another corrupt AP file in the setiathome.berkeley.edu folder. Meaning you need to remove every file with AP in the name and copy the AP files from the Download again.
There are a number of files in the setiathome.berkeley.edu folder with AP or AstroPulse in the name, such as,
AP_clFFTplan_GeForceGTX...
AstroPulse_Kernels_r2751.cl_GeForceGTX...
It is common TroubleShooting 101 to remove those files generated by the App so it can build new ones, just in case one becomes corrupt.
It doesn't sound like you removed them.
TroubleShooting 102 goes into replacing the OS, or parts of it. I'm going to let someone else endure that since you believe 101 to be desperate.
My suggestion is to first try reinstalling the GPU driver.
Good Luck.

BTW, did you ever identify the Files you claimed were missing back here?
They try to run but immediately halt and the status changes to "waiting to run". I have checked app_info.xml and there is section for the AP app which has the right app files in the folder. There was also a second section referring to an older version of AP which does not have the appropriate apps in the folder.
ID: 1916779 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1916813 - Posted: 4 Feb 2018, 4:37:13 UTC - in response to Message 1916777.  

. . Thanks for that Keith,

. . Sadly I think this might also be a dead end ... my results are much like yours

stephen@Albert-E18790314:~$ ldd '/home/stephen/BOINC/projects/setiathome.berkeley.edu/astropulse_7.08_x86_64-pc-linux-gnu__opencl_nvidia_100'
linux-vdso.so.1 => (0x00007fff0c39d000)
libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f9f2c280000)
libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f9f2c062000)
libOpenCL.so.1 => /usr/lib/x86_64-linux-gnu/libOpenCL.so.1 (0x00007f9f2be59000)
libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007f9f2bc40000)
libnsl.so.1 => /lib/x86_64-linux-gnu/libnsl.so.1 (0x00007f9f2ba26000)
libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007f9f2b722000)
libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f9f2b41c000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f9f2b053000)
/lib64/ld-linux-x86-64.so.2 (0x00007f9f2c484000)
libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007f9f2ae3d000)


. . But I am not sure why you think it is not creating the wisdom file, this is what is in the folder ...

r2751_IntelRCoreTMi56600CPU330GHz_x64.wisdom


. . Is that the file you are talking about? There is also ...

r3305_IntelRCoreTMi56600CPU330GHz_x64.wisdom


. . But the latter is from the stock apps that originally kicked off.

Stephen

? ?
ID: 1916813 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1916821 - Posted: 4 Feb 2018, 5:03:24 UTC - in response to Message 1916779.  

I looks as though you missed this part again;
If that doesn't solve it, you probably have another corrupt AP file in the setiathome.berkeley.edu folder. Meaning you need to remove every file with AP in the name and copy the AP files from the Download again.
There are a number of files in the setiathome.berkeley.edu folder with AP or AstroPulse in the name, such as,
AP_clFFTplan_GeForceGTX...
AstroPulse_Kernels_r2751.cl_GeForceGTX...
It is common TroubleShooting 101 to remove those files generated by the App so it can build new ones, just in case one becomes corrupt.
It doesn't sound like you removed them.


. . OK I am presuming this does not include the .wu files with AP in them. But I am now guessing I should have removed the wisdom file that I discussed with Keith in the previous message even though there is no AP in the filename. Are there any other files that should be removed that might also not have AP in their name or is it only the app files (OpenCL files) with an AP in them? I definitely got the one in orange but I may have missed the first one, I was looking for AP and CL or openCL and of course AstroPulse but I don't remember seeing the cIFFTplan part.

TroubleShooting 102 goes into replacing the OS, or parts of it. I'm going to let someone else endure that since you believe 101 to be desperate.
My suggestion is to first try reinstalling the GPU driver.
Good Luck.


. . I consider anything that requires me to meddle with parts of file systems whose function is unknown to me or is foundation level software to be desperate. The higher the possibility of catastrophic failure the more desperate it is.

. . But re-installing the video driver I have done before and can tackle again.

BTW, did you ever identify the Files you claimed were missing back here?
They try to run but immediately halt and the status changes to "waiting to run". I have checked app_info.xml and there is section for the AP app which has the right app files in the folder. There was also a second section referring to an older version of AP which does not have the appropriate apps in the folder.


. . The second AP section referred to an app r2728 but there were no files with r2728 where the r2751 files were. Since "commenting" them out made no difference I "uncommented" that section so the file was back to original state. The whole thing has now been replaced so it is a moot point.

. . BTW thanks for further assistance ...

Stephen

:)
ID: 1916821 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 4964
Credit: 318,860,332
RAC: 798,713
United States
Message 1916829 - Posted: 4 Feb 2018, 5:19:03 UTC - in response to Message 1916813.  

. . Thanks for that Keith,

. . But I am not sure why you think it is not creating the wisdom file, this is what is in the folder ...

r2751_IntelRCoreTMi56600CPU330GHz_x64.wisdom

Stephen

Because this is the original error message in your post for help.

ERROR: OpenCL kernel/call 'clCreateContext' call failed (999) in file ../../src/GPU_lock.cpp near line 1003.


That is an error in either accessing the AP gpu application or in creating or reading the wisdom file associated with the gpu, driver version or application.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1916829 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 4138
Credit: 249,136,934
RAC: 334,823
United States
Message 1916833 - Posted: 4 Feb 2018, 6:02:00 UTC - in response to Message 1916829.  

Well, I've never heard of a .wisdom file causing the App not to start, but, I suppose anything is possible. It certainly won't hurt to remove the wisdom file.
If you go back to version 2710 of the GPU_lock.cpp file that line is talking about device IDs, so, that might be driver related. Drivers are attached to kernels, which he has changed a couple of times.
The files that need to be removed are the ones created by the App;
AP_clFFTplan_GeForceGTX...
AstroPulse_Kernels_GeForceGTX...
and .wisdom for good measure.
ID: 1916833 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1916852 - Posted: 4 Feb 2018, 12:22:18 UTC - in response to Message 1916833.  

Well, I've never heard of a .wisdom file causing the App not to start, but, I suppose anything is possible. It certainly won't hurt to remove the wisdom file.
If you go back to version 2710 of the GPU_lock.cpp file that line is talking about device IDs, so, that might be driver related. Drivers are attached to kernels, which he has changed a couple of times.
The files that need to be removed are the ones created by the App;
AP_clFFTplan_GeForceGTX...
AstroPulse_Kernels_GeForceGTX...
and .wisdom for good measure.


. . Well it was another (remote) possibility eliminated. After removing all of those files it still will not run, and it has not re-created the wisdom file so Keith is right about that error message. Not sure which to try first, re-installing the video drivers or trying some sort of recovery of the OS files/setup. Maybe I should just abort the wu's and set AP work to NO.

Stephen

:(
ID: 1916852 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 4138
Credit: 249,136,934
RAC: 334,823
United States
Message 1916872 - Posted: 4 Feb 2018, 15:04:32 UTC - in response to Message 1916852.  
Last modified: 4 Feb 2018, 15:18:23 UTC

Well, I've never heard of a .wisdom file causing the App not to start, but, I suppose anything is possible. It certainly won't hurt to remove the wisdom file.
If you go back to version 2710 of the GPU_lock.cpp file that line is talking about device IDs, so, that might be driver related. Drivers are attached to kernels, which he has changed a couple of times.
The files that need to be removed are the ones created by the App;
AP_clFFTplan_GeForceGTX...
AstroPulse_Kernels_GeForceGTX...
and .wisdom for good measure.


. . Well it was another (remote) possibility eliminated. After removing all of those files it still will not run, and it has not re-created the wisdom file so Keith is right about that error message. Not sure which to try first, re-installing the video drivers or trying some sort of recovery of the OS files/setup. Maybe I should just abort the wu's and set AP work to NO.

Stephen

:(

The App has to Run to create a .wisdom file. The App is Not running, hence No Wisdom file is being created. The Error message refers to Device IDs, look it up, https://setisvn.ssl.berkeley.edu/trac/browser/branches/sah_v7_opt/src/GPU_lock.cpp?rev=2751#L1003 Device IDs are handled by the driver, Not the wisdom file. Do what you want to. I'm done here.
ID: 1916872 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 4964
Credit: 318,860,332
RAC: 798,713
United States
Message 1916896 - Posted: 4 Feb 2018, 18:27:39 UTC - in response to Message 1916872.  

Thanks for the snippet of code of where the error occurs. So, the device isn't getting a device id correctly. I have found device id enumeration bothersome in both Windows and Linux.

I would pull all the graphics cards out and boot the system with one graphics card. Then remove the video drivers and shut it down. Then reinstall the graphics cards and reinstall the video drivers to force the system and video drivers to re-enumerate the graphics cards and sort out the device id's. Then unpack the BOINC package from scratch and reinstall.

That is what I would do. I wouldn't waste any more time in trying to fix a broken system. Just start from scratch.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1916896 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1916932 - Posted: 4 Feb 2018, 21:38:37 UTC - in response to Message 1916896.  

Thanks for the snippet of code of where the error occurs. So, the device isn't getting a device id correctly. I have found device id enumeration bothersome in both Windows and Linux.

I would pull all the graphics cards out and boot the system with one graphics card. Then remove the video drivers and shut it down. Then reinstall the graphics cards and reinstall the video drivers to force the system and video drivers to re-enumerate the graphics cards and sort out the device id's. Then unpack the BOINC package from scratch and reinstall.

That is what I would do. I wouldn't waste any more time in trying to fix a broken system. Just start from scratch.


. . Hi Keith,

. . If the only problem is not being able to get a valid ID for the GPUs then would that not also disable to the CUDA80 app which surely must need that as well? Or is it only OpenCL apps that need that information?

. . Well I guess I video drivers it is ...

Stephen

<shrug>
ID: 1916932 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1916935 - Posted: 4 Feb 2018, 21:45:55 UTC - in response to Message 1916872.  


The App has to Run to create a .wisdom file. The App is Not running, hence No Wisdom file is being created. The Error message refers to Device IDs, look it up, https://setisvn.ssl.berkeley.edu/trac/browser/branches/sah_v7_opt/src/GPU_lock.cpp?rev=2751#L1003 Device IDs are handled by the driver, Not the wisdom file. Do what you want to. I'm done here.


. . Yes that was the point that Keith made. Well thanks for the attempt. Since there will be very little AP work in future (or so it seems) I will give up and say no to AP work. There are lots of cunchers who will be happy to lose a bit of competition for them :)

. . I had a look at that code but do not comprehend the variables for IDs and fake IDs. When I get some other things sorted I will probably rebuild that rig with later versions of OS, BOINC and Special sauce (CUDA90).

Stephen

PS Sorry you find it so frustrating that other users aren't programmers or Linux savants.

:(
ID: 1916935 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 4964
Credit: 318,860,332
RAC: 798,713
United States
Message 1916937 - Posted: 4 Feb 2018, 21:47:08 UTC - in response to Message 1916932.  

Well the video drivers have to supply two very different components. One for CUDA and one for OpenCL. Obviously, the OpenCL component is munged up. I agree time for complete removal of the video drivers and a new install of the the drivers is indicated.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1916937 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1917442 - Posted: 7 Feb 2018, 14:20:49 UTC - in response to Message 1916937.  
Last modified: 7 Feb 2018, 14:28:22 UTC

Well the video drivers have to supply two very different components. One for CUDA and one for OpenCL. Obviously, the OpenCL component is munged up. I agree time for complete removal of the video drivers and a new install of the the drivers is indicated.


. . Of course, they function independently, so card detection is different for each.

. . Well first I ran package manager and looked for loose ends, and tried to tidy them up, but no change.

. . Now I have uninstalled 384.111 and then re-installed it. Still no change. Nothing to do but dump these suckers and set preferences to not accept AP work. Whatever has changed it is beyond my doing or undoing.

. . I have grown sick of seeing them sitting there, mocking me, saying "waiting to run" ...

. . One thing that just occurred to me. Maybe something was altered somewhere when I removed the GTX1050 from the mix. Since, for whatever reason, BOINC decided it was the most powerful GPU could it have disabled something when it detected that it was no longer there? That happened after I had the last successful AP tasks and before this problem appeared. It is the only significant change that has taken place in a long time.

Stephen

<sigh>
ID: 1917442 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3535
Credit: 75,428,059
RAC: 114,100
Australia
Message 1917451 - Posted: 7 Feb 2018, 14:43:02 UTC

. . Just for the record ...


Thu 08 Feb 2018 00:59:08 AEDT | | Data directory: /home/stephen/BOINC
Thu 08 Feb 2018 00:59:08 AEDT | | CUDA: NVIDIA GPU 0: GeForce GTX 970 (driver version unknown, CUDA version 9.0, compute capability 5.2, 4032MB, 3830MB available, 6510 GFLOPS peak)
Thu 08 Feb 2018 00:59:08 AEDT | | CUDA: NVIDIA GPU 1: GeForce GTX 970 (driver version unknown, CUDA version 9.0, compute capability 5.2, 4037MB, 3965MB available, 6510 GFLOPS peak)
Thu 08 Feb 2018 00:59:08 AEDT | | OpenCL: NVIDIA GPU 0: GeForce GTX 970 (driver version 384.111, device version OpenCL 1.2 CUDA, 4032MB, 3830MB available, 6510 GFLOPS peak)
Thu 08 Feb 2018 00:59:08 AEDT | | OpenCL: NVIDIA GPU 1: GeForce GTX 970 (driver version 384.111, device version OpenCL 1.2 CUDA, 4037MB, 3965MB available, 6510 GFLOPS peak)


. . That was the same before re-installing the drivers. So I don't think detecting the GPU is the issue.

Stephen
ID: 1917451 · Report as offensive     Reply Quote
Previous · 1 . . . 45 · 46 · 47 · 48 · 49 · 50 · 51 . . . 59 · Next

Message boards : Number crunching : Setting up a Linux machine to crunch CUDA80 for Windows users


 
©2018 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.