Yeap BIOS is up to date…
I dont know how to post all these in text. Is there any log that has all these errors logged?
Also on this thread I found the workaround solution to my problem
(added the “pci=nommconf” to resolve these type of errors)
Hi All, we were able to use an acceptable work around for our set up. We figured out that there is a conflict between Nvidia driver and time out period due to the PEX/PCie sleep state. To get around this turn off ASPM in your BIOS, However…
You also need to disable ASPM in the GRUB
" pcie_aspm=off "
For us we saw this issue using the GTX 1080’s. But this was still technically a workaround and even the BIOS released merely gave a more permanent means of a work around. Nvidia basically just gave us the cold shoulder.
First, I apologize I realized I poorly wrote my last response… I should have explained our exact issue.
Are you talking about the “pci=nommconf” and similar for shutting off the AER reporting? Those workaround are simply a mask covering the issue up in the end all this did was turn off the AER reporting service which is ok for some but ads complexity for others. The AER message we were receiving were due to a conflict between the Nvidia driver attempting to communicate with the video card and timing out due to lack of response because the PCIe was in ASPM L0/L1 (asleep essentially)
In the end the reasoning was due to ASPM being in L0/L1 and taking too long to come live again, which inturn has to communicate with the PEX again and then send data to the Nvidia card. When we disable ASPM permanently via OS GRUB (then in the BIOS only after BIOS update) the ASPM would not drop to L0 so there was no delay due to sleep, thus no time out and no AER reported.
Technically it’s a workaround solution, and why all of a sudden we started seeing it I’m not 100% sure, whether it’s from the 1080 drivers or maybe a Intel microcode change we’re not sure.
I found also another answer that he is saying that
Links states dropping to PCIe 1.1 is standard practice of newer GPUs and they only do this while not under load, when under load they ramp back up to full link state
AFAIK there is no way to change this behaviour
Well even with that it needs 7sec to boot
[wolfyrion@WPC ~]$ systemd-analyze
Startup finished in 5.910s (kernel) + 873ms (userspace) = 6.784s
[wolfyrion@WPC ~]$ systemd-analyze blame
Well thats all for now about my research on these errors…