nvidia-bug-report.log (1.5 MB)
"No devices were found" after installing CUDA 10.2 on Ubuntu 18.04.4 Server with AMD Ryzen 3970x and RTX 2080 SUPER™ GAMING OC 8G
Both gpus can’t be initialized:
Jun 04 20:51:24 anyone kernel: NVRM: GPU 0000:4b:00.0: RmInitAdapter failed! (0x26:0x65:1227) Jun 04 20:51:24 anyone kernel: NVRM: GPU 0000:4b:00.0: rm_init_adapter failed, device minor number 1 Jun 04 20:51:45 anyone kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x24:0x65:1185) Jun 04 20:51:45 anyone kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
- please update system bios
- remove both cards and then plug only one back in, check if they’re running alone
- if not, check if it works in another system
I don’t have an Aorous, but a Designare, but I don’t think it’s that different.
Secure Boot should be in the “Boot” section of the “Advanced/Detailed” view.
Thanks for your reply. Yes, it should be, but I checked very carefully and did not find it in the Boot section.
Secure boot is turned off according to the log.
Thank you guys for the prompt replies!
After installing CUDA 10.2 previously, I simply removed all the two RTX 2080 SUPER GAMING OC 8G cards, and inserted only one RTX 2060 Super without any change in the system, then it immediately detected the 2060 card by ‘nvidia-smi’.
nvidia-bug-report_2060super.log (2.1 MB)
Original report log nvidia-bug-report.log (1.5 MB)
(I am not sure why my post was flagged as spam. I sincerely followed the guidelines for debugging and asked for help specific to the topic.)
Logs for each card alone
Can’t be initialized, you can only check if any of both work in another system and RMA them otherwise.
Thanks for the reply! I don’t have Windows in hand so it would take some time to do that.
Not another OS, in another computer, i.e. another mainboard.
Remembering the distortions in bios, though, I’d say the gpus are broken.
We have two machines in total with exactly the same configurations, i.e., two motherboards, each of which has two GPU cards. Both have seen the same issues that ‘No devices were found’ after installing CUDA, except that I only tested the bug report script with one machine and posted the reports here.
Also, one thing to emphasize is that after swapping RTX 2080 with a RTX 2060 directly, ‘nvidia-smi’ shows the RTX 2060 info.
I checked the way to update BIOS for this motherboard and what I found is that first I need a Windows system to download it to a USB Drive.
What is ‘RMA’ exactly? Is it ‘Return Merchandise Authorization’?
Thanks for pointing out!
I don’t have a different mainboard in hand, it may take some time to try that.
Looks like a bios issue, then, or a general incompatibility with the mainboard.
Regarding RMA, while for real it’s just a number used in warranty process, it’s often used as a sloppy synonym/abbreviation for ‘return to manufacturer or merchant according to warranty conditions’.