please confirm the fix release when available here, so we can post reports with our configurations.
corruption still present with driver 460.73.01 released 11MAY2021 , kernel 5.4.0-73-generic #82-Ubuntu
EDIT: seems the 80.01 is not in the ubuntu 20.01 repos yet. Will try that when it is available.
You should try the 465 version, for me it fixed the screen corruption issue but it only enables 1 monitor.
I use 2 monitors for work ;-)
But I only ever had corruption on the 2nd monitor, so I guess this is a solution of sorts!
recent driver update to 460.80 fixed the issue for me. Thanks to the devs for quick fix!
Bug still present in 460.80 , 5.4.0-73-generic #82-Ubuntu
Has Nvidia found the reason for this? It appears the screenshot doesn’t see this corruption, so only a user see’s it when a photo is taken.
I am running kernel 5.8 and for me the problem is gone, but it will only work if I have one monitor connected. If I connect the second monitor, it all goes black.
any word from Nvidia when this will be fixed for a kernel >= 5.4?
bug still present with Nvidia driver 460.84, linux 5.4.0-74-generic #83-Ubuntu
bug still present in 5.8.0-55-lowlatency #62~20.04.1-Ubuntu , nvidia 460.84.
@amrits Any further guidance on an expected working configuration?
The 5.8 kernel has both monitors working (manual install of evdi from Added support for 5.10 by applying @abucodonosor's patch found in #234. by sickcodes · Pull Request #237 · DisplayLink/evdi · GitHub ) but severe corruption on the second monitor occurs when filling visual buffer of konsole (kde xterm)
A refinement. Doing some checking of logs and I came across this in :
grep -i --color memory /var/log/Xorg.0.log
[ 21957.531] (–) NVIDIA(0): Memory: 4194304 kBytes
[ 21957.757] (II) NVIDIA: Using 24576.00 MB of virtual memory for indirect memory
[ 21958.060] (==) NVIDIA(0): Disabling shared memory pixmaps
[260495.749] (WW) NVIDIA(0): Failed to allocate 3840x2160+0+0 head surface: out of memory.
Is this relevant?
Hi @amrits , it’s been a few months and we still have this problem.
Any further information on which combination of drivers and kernel solve this?
Bug still present in drivers 465.31 with kernel 5.8.0-55-lowlatency #62~20.04.1-Ubuntu SMP
did some digging over the weekend and found some more details.
I noticed that in the dithering section of nvidia settings, there were no values set for the 2nd monitor - perhaps this is an oversight?
I set the values to be the same as the first monitor controls->(dithering, auto, 8bpc).
The static corruption goes away now. However when we get corruption, it now flashes at approx 30hz (used my Pixel to slow motion and count). It is possible it is aliasing as monitors are 60Hz.
Does this help anyone isolate this problem? Clearly if parameters can change this , it suggests the bug is somewhere in the driver memory management, or something to do with the Displaylink
This is a business product, I really need it to work as such!!!
I solved the white noise by placing a DP-> HDMI adapter. See
I am using 2xDP , one of the monitors has HDMI (the one with the noise), I’ll see if I can get an HDMI cable (as monitor has native HDMI) and see if it solves this.
Nvidia needs to up their game with bug-fixes. This black hole for a business product is a problem.
Ok I tried using an HDMI cable to the remote (2nd) monitor and it made it worse. Specifically the screen bottom left had bits of the left screen overlap, misaligned and also it was really bright.
Back on 2xDP - perhaps I need to find a decent adapter in case it’s a problem with HDMI cable.
This is an LG 4K monitor and also has UsbC - but the lenovo doesn’t support daisy chaining like Apple…
My monitors don’t have colordepth indications - is there an xrandr way of looking at this?
ok nvidia settings says both monitors are 24 bit, so must be something else. However, perhaps the DP->HDMI adapter is doing something? I ordered the same model, will report back.
I can confirm this works - I am surprised but it works. Using both Display ports on the lenovo docking station the second 4K screen is stable. A side point, I had to set the monitor to sRGB mode, otherwise it was super bright - I wonder if anyone knows the details of this subtlety.
@amrits @nvidia277 Could we have some feedback from Nvidia as to why this workaround is necessary. The exact same cable being used with this adapter does not work using HDMI from the docking station. If this really is a 30 bit limitation, perhaps prod lenovo…
In my case, I was able to compare the OSD info of monitors when windows(24bit) and linux(30bit) are booted.The monitors EDID report they can support 30bit, but I guess the nvidia driver doesn’t respect the bandwidth limts due the middlemen(dockingstation etc) - see also DeepColourDepthSupportPlan - Ubuntu Wiki
On Linux it has clearly shown 30bit, and nvidiasettings shows 24bit.
So, instead of the adapter, you can use custom edid (see my askubuntu post), it worked for me to limit it also and prevent the screen corruption.
But it is very unflexible because of my daily docking/undocking cycles with different monitors between home<->office1<->office2.
Anyway, I can’t use Linux due unreliable docking/undocking (sporadic X11 restart/crash after resume). Really sad, not able to use the expensive mobile RTX 5000 in a reliable way in Linux without X11 crashes. Leason learned ;-)