V520 causes intermittent blanking on daisychained DP monitor, v515 ok

In Fedora linux with a GTX 1650. Symptom: the daisychained DP monitor works fine for 1 to 15 seconds, then goes black for 1 to 8s (usually around 2s), then comes back on, goto step 1. Never stops.

Upgrading to akmod-nvidia-3:520.56.06-1.fc35.x86_64 introduced this new monitor blanking-out / flickering bug. It only affects one of my 3 monitors, which is plugged in as the final monitor on a 2-monitor DP daisy chain.

akmod-nvidia-3:515.65.01-1.fc35.x86_64 does not have this problem. I dnf downgraded to 515 and the bug went away. Bug is definitely caused by some change between 515 and 520. Reproduced this many times. 515 installed/booted: no bug; 520 installed/booted: bug.

I tried putting in the latest nvidia-gpu-firmware from testing first, as that upgraded at the same time, but it didn’t affect anything. Bug persists with any version of nvidia-gpu-firmware.

Versions of packages that work:
akmod-nvidia-3:515.65.01-1.fc35.x86_64
nvidia-settings-3:515.65.01-1.fc35.x86_64
xorg-x11-drv-nvidia-3:515.65.01-1.fc35.x86_64
xorg-x11-drv-nvidia-kmodsrc-3:515.65.01-1.fc35.x86_64
xorg-x11-drv-nvidia-libs-3:515.65.01-1.fc35.i686
xorg-x11-drv-nvidia-libs-3:515.65.01-1.fc35.x86_64

Versions broken:
akmod-nvidia-3:520.56.06-1.fc35.x86_64
kmod-nvidia-5.18.11-100.fc35.x86_64-3:520.56.06-1.fc35.x86_64
nvidia-settings-3:520.56.06-1.fc35.x86_64
xorg-x11-drv-nvidia-3:520.56.06-1.fc35.x86_64
xorg-x11-drv-nvidia-kmodsrc-3:520.56.06-1.fc35.x86_64
xorg-x11-drv-nvidia-libs-3:520.56.06-1.fc35.i686
xorg-x11-drv-nvidia-libs-3:520.56.06-1.fc35.x86_64

will attach nvidia-bug-report.log.gz in a sec

nvidia-bug-report.log.gz (122.9 KB)

Forced to upgrade to Fedora 36. Bug persists in F36 with the 520 (520.56.06) driver. I cannot seem to get 510 or 515 to compile with rpmfusion akmods on F36 so now I am stuck with the flickering 3rd monitor. I just turn it off now – it’s unusable with the screen blanking bug.

I can report that I ran F35 with the 515 driver for the 2 months since I reported this bug. As I said before 515 does not have this problem and my setup worked flawlessly the whole time. 520, no matter the OS/kernel version is faulty and has this bug.

I have found similar me-toos in the newer bug reports. To consolidate them, I note them here:

Maybe the 3 of us together can discover all the commonalities. There must be something unique to our setups because it’s clear most people do not have this problem. However, the problem is real, and it’s not a hardware problem as everyone has reported that they can fix the bug by using 515 or Windows.

I want to note that this cannot be put down to a cabling issue as often seems to be the case in blanking problems reported on the net. The reason is that when I was able to switch between drivers version 515 and 520 at will, 515 was 100% bug free and 520 always had the bug. Zero change in hardware or cabling.

This jives with the other reports here of similar bugs happening in linux but not Windows. Again, with no change in cabling or hardware.

One thing I am doing a bit different from what most people would be doing is my computer->monitor1 DP cable is 10’ because I have my computers in a different room than my monitors (noise free!). Then the daisychain cable is the 6’ one that came with the monitor (all Dell, 2 US series, one P series). But again, works 100% fine with older nvidia drivers going way back a few years. But 520 changes something, and that causes this bug.

I remembered I had a free HDMI port on the video card (was only using DP and DVI so far). So I thought I’d try running a 10’ HDMI cable just for the blanking monitor. Thus eliminating the DP daisychain.

Got it done and the monitor now works perfectly fine, bug free. No blanking.

So the problem is definitely a DP, or a DP-daisychaining issue. Waiting to see if the other guys report using DP daisychain or not.

Note, this doesn’t solve the bug, since I’ll be able to recreate it just by plugging the DP-daisychain cable back in. But it does get me back to 3 monitors in the meantime. The reason I was using the daisychain was that: a) I eventually wanted a 4th monitor, and now with the HDMI port taken I cannot do that; and b) less cables going through the hole in the wall.

When a new version (525?) comes out on fedora rpmfusion, I will be able to test this bug again just by swapping cables, and I will do so and report back.