| Author |
Message |
|---|
|
|
|
Using a NVIDIA Quadro M4000 (4095MB) driver: 41163 as soon as the WU downloads I get an error. Using the latest Nvidia drivers, nothing is overclocked.
Tasks:
http://www.primegrid.com/results.php?hostid=892808
|
|
|
 Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,230,186
RAC: 922,880
                               
|
Using a NVIDIA Quadro M4000 (4095MB) driver: 41163 as soon as the WU downloads I get an error. Using the latest Nvidia drivers, nothing is overclocked.
Tasks:
http://www.primegrid.com/results.php?hostid=892808
I see that your GFN tasks are also failing.
Please try running a PPS-Sieve task on the GPU. Whether that works or not will help narrow down the source of the problem. PPS-Sieve is a CUDA app, while Genefer and AP27 are OpenCL apps. If PPS-Sieve works, then we know it's a problem with the OpenCL drivers. If it doesn't work, it's the main video driver.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
10-4 running now. ETA 13min for first WU |
|
|
|
|
|
Is that enough to diagnose or should I keep it running? |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,230,186
RAC: 922,880
|
Is that enough to diagnose or should I keep it running?
Okay, PPS-sieve isn't working either, which means that either the driver is completely broken or the GPU is inaccessible.
The most likely culprit is that Windows 10 automatically updated your video driver and broke it. The remedy is to manually uninstall the video driver, and then download the video driver directly from Nvidia's website and do a complete install of the driver.
If that doesn't fix it, then it's likely something is preventing BOINC from seeing the GPU. Possible causes include:
* BOINC is running in service mode. You can't install BOINC as a service if you want to use the GPU.
* You're using Windows Remote Desktop. That won't work with a GPU.
* You've got more than one person logged in to Windows simultaneously.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Hmm strange, did a clean install from the Nvida site today. Not using remote desktop either. I'll double check the other things. Thanks. |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1957
ID: 352
Credit: 6,140,864,273
RAC: 2,278,045
                                  
|
|
SO, I'm having similar issues with new RTX 2080, driver version 411.63.
Done 48 AP27, 8 with errors.
It happens during computation, I got screen artefacts.
Eventlog says nvlddmkm stopped working (EventID 4101).
I have replaced my old 550WU with new Seasonic Prime Ultra 650 W Titanium.
Done Display Driver Uninstall in safe mode - there was AMD Fury Nano.
Done Display Driver Uninstall in safe mode for NVidia and freshly installed 411.63
Windows update did not messed-up my drivers.
No RDP.
BOINC not as a service.
Only one person logged.
Unnecessary apps like Chrome was closed.
And 5 errors overnight.
I also tryed trick with TdrDelay
https://answers.microsoft.com/en-us/windows/forum/windows_10-hardware-winpc/windows-10-display-driver-has-stopped-responding/5caff7ea-bed5-4111-baac-e72d32d4dbe2
I wonder if it is drivers version related.
____________
My stats |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1957
ID: 352
Credit: 6,140,864,273
RAC: 2,278,045
|
|
Did some reseach, 411.63 might not be the best driver...yet.
https://forums.geforce.com/default/topic/1072459/geforce-drivers/official-411-63-game-ready-whql-display-driver-feedback-thread-released-9-19-18-/
____________
My stats |
|
|
|
|
|
Honza wrote: Did some reseach, 411.63 might not be the best driver...yet.
I updated to NVidia diver version 411.63 and saw my AP27 crunch time increase by almost 30% (GTX 1060 3GB, Win 10 64 Home).
Then reinstalled version 399.24 and times went back to normal.
____________
"Accidit in puncto, quod non contingit in anno."
Something that does not occur in a year may, perchance, happen in a moment. |
|
|
|
|
|
Reverted back to Nvidia 397.93 drivers and everything seems to be working again. Michael, not sure if you want to confirm? |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,230,186
RAC: 922,880
|
Reverted back to Nvidia 397.93 drivers and everything seems to be working again. Michael, not sure if you want to confirm?
As of right now there's one completed task. It didn't find anything, but it did finish without error. There's no reason to suspect anything is wrong there.
If 397.93 works, which driver version were you using that was causing problems?
____________
My lucky number is 75898524288+1 |
|
|
|
|
Reverted back to Nvidia 397.93 drivers and everything seems to be working again. Michael, not sure if you want to confirm?
As of right now there's one completed task. It didn't find anything, but it did finish without error. There's no reason to suspect anything is wrong there.
If 397.93 works, which driver version were you using that was causing problems?
I was using 411.63 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,230,186
RAC: 922,880
|
Reverted back to Nvidia 397.93 drivers and everything seems to be working again. Michael, not sure if you want to confirm?
As of right now there's one completed task. It didn't find anything, but it did finish without error. There's no reason to suspect anything is wrong there.
If 397.93 works, which driver version were you using that was causing problems?
I was using 411.63
I'm currently running AP27 under 388.13. When my current task finishes, I'll run at least one GFN15 and PPS-Sieve task to make sure they work as well. Then I'll install the latest driver from Nvidia, presumably 411.63, and see if those tasks work with the new driver. My system is a Windows 7/GTX1060 configuration, so it's not identical to yours but probably the only thing that matters is that it's Windows and the driver version.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,230,186
RAC: 922,880
|
Reverted back to Nvidia 397.93 drivers and everything seems to be working again. Michael, not sure if you want to confirm?
As of right now there's one completed task. It didn't find anything, but it did finish without error. There's no reason to suspect anything is wrong there.
If 397.93 works, which driver version were you using that was causing problems?
I was using 411.63
I'm currently running AP27 under 388.13. When my current task finishes, I'll run at least one GFN15 and PPS-Sieve task to make sure they work as well. Then I'll install the latest driver from Nvidia, presumably 411.63, and see if those tasks work with the new driver. My system is a Windows 7/GTX1060 configuration, so it's not identical to yours but probably the only thing that matters is that it's Windows and the driver version.
I finished running all three GPU apps under 388.13, installed 411.63, and am now in the process of running the apps again. In addition to failing completely, as in your case, someone else reported a 50% reduction of speed with the 411.63 driver. I'll check for that as well. Right now I'm running the PPS-Sieve app, and so far it's not only running but is running at the expected speed. I'll post more information once I'm done testing.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,230,186
RAC: 922,880
|
Reverted back to Nvidia 397.93 drivers and everything seems to be working again. Michael, not sure if you want to confirm?
As of right now there's one completed task. It didn't find anything, but it did finish without error. There's no reason to suspect anything is wrong there.
If 397.93 works, which driver version were you using that was causing problems?
I was using 411.63
I'm currently running AP27 under 388.13. When my current task finishes, I'll run at least one GFN15 and PPS-Sieve task to make sure they work as well. Then I'll install the latest driver from Nvidia, presumably 411.63, and see if those tasks work with the new driver. My system is a Windows 7/GTX1060 configuration, so it's not identical to yours but probably the only thing that matters is that it's Windows and the driver version.
I finished running all three GPU apps under 388.13, installed 411.63, and am now in the process of running the apps again. In addition to failing completely, as in your case, someone else reported a 50% reduction of speed with the 411.63 driver. I'll check for that as well. Right now I'm running the PPS-Sieve app, and so far it's not only running but is running at the expected speed. I'll post more information once I'm done testing.
Initial observations of the 411.63 driver.
PPS-Sieve (a CUDA app) runs properly and at the correct speed.
Genefer (an OpenCL app) runs properly and at the correct speed. This was a GFN-15 task; it's possible larger GFN tasks might behave differently.
AP27 (another OpenCL app) runs properly ... but does seem slower. This task is in progress; I'll let it finish and report back. Ap27 appears to be taking about a third longer to run than it did before, but it does appear to be running correctly.
Tentatively, based on these preliminary observations, as well as others' experiences, I'd recommend NOT using the 411.63 driver.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Recently I installed the new driver in Linux (driver 410.57). So I could make a few tests.
Linux (410.57) about 2100-2200 s for AP27 with 1070/1070 TI.
Windows (388.13) about 1660 s for AP27 with a GTX 1070.
GFN-17-Mega runs well with the same speed for WIN and Linux.
No errors so far.
____________
DeleteNull |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,230,186
RAC: 922,880
|
|
Reverting back to 388.13 confirms that it's the driver causing the slowness on my system.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
So, think I've found out a bug? When I set the Nvidia driver to optimize for compute "on" I get a computation error immediately. With it off, WU proceeds fine. This is with driver 397.93. So driver 411.63 might actually not be the problem? |
|
|
|
|
|
The newer driver do no have this optimize on compute. Running 416.94 on AMD and 2080Ti.
Both genefer and AP27 runs creates an access violation at start. The sieve is running without any problem.
The errors is the same in both programs
- Unhandled Exception Record -
Reason: Access Violation (0xc0000005) at address 0x03a26204 read attempt to address 0x00000010
Any clue?
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,230,186
RAC: 922,880
|
The newer driver do no have this optimize on compute. Running 416.94 on AMD and 2080Ti.
Both genefer and AP27 runs creates an access violation at start. The sieve is running without any problem.
The errors is the same in both programs
- Unhandled Exception Record -
Reason: Access Violation (0xc0000005) at address 0x03a26204 read attempt to address 0x00000010
Any clue?
Download the driver directly from Nvidia's website. The common cause for this is that Windows installed the driver, and only installed the CUDA half but not the OLC half. GFN and AP27 are OCL apps.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Tanks, but same problem as before. I think I have to wait for a new version. [/quote] |
|
|
