Maybe its time to put this in the 1st post here.
 
Currenty there are only 2 confirmed workarounds:
 
1. Disable Hardware Acceleration in your browser. This causes performance loss with demanding sites, and videos, streams, so it comes at a price. TDR's still might happen in other applications that are graphics demanding, but not games (graphics programs, wallpaper engine etc)
2. Set the nvidia driver to Prefer Maximum Performance (instead of Optimal). This wont let your card downclock in idle, so it will use more power, and cause more heat than usual. But this entirely prevents TDR's.
 
+ You can try a hybrid solution, and set the Prefer Maximum Performance on a per application base in the nvidia driver. For example for your browser's exe. Has to be yet confirmed, if this is a reliable solution.

I just tried it doesnt seem to be effective, plus you cant set it on firefox and that also TDRs a lot..

jamesleroux
I just tried it doesnt seem to be effective, plus you cant set it on firefox and that also TDRs a lot..

Tried what?

Dont think my blackscreen is temporary, I spent a few hours troubleshooting and have gone through all the usual steps complete with testing with onboard GPU and a 2070 super but looks like my 3090 is kaput. Waiting for response to my service ticket so I can begin the RMA process. 
 
Was a good streak, till now I dont think I have had to RMA a graphics card since the 500 series. 

nice

I have the same issues

this is happening recently to me as well after updating to precision x1 1.1.3 and updating the card's firmware (3080). screen would go black for a few seconds and apps like chrome would crash (i've noticed it's mainly just chrome?) and then afterburner would say that connection to GPU1 is not found. event viewer says its either a dwm.exe issue or paging file too small error. super irritating. chrome's hardware acceleration option is disabled and my power setting is set to optimal power

Ive tracked the error in chrome, firefox, and brave browsers...  Right now ive simply disabled HW acceleration and am seeing if that helps. I dont like the card running hot all the time. so i put the power management back to "normal", and re-enabled PCIe link state management since that setting did nothing.

Asrock 570 phantom gaming, Evga 850 platinum power supply with all 8 pins from VGA 1,2&3 so no daisy chain cable 32Gb ram Corsair AMD 3600 and 3090 ftw3 ultra, Well after about 15hours the card is dead black screen disconnected it removed and reseated (6/7 attempt) got it working tried to load cyber punk and it crashed again with fans ramping 100% then after that never got the Windows screen again, I know the computer has booted into windows as the hue plus lighting is doing what I set it to in window, normally at power loss the hue plus results to constant lighting until it boots into windows. Not a happy chappy 

Just as a heads up, today there is a new acknowledged issue at the current game ready drivers topic at nvidia.
The wallpaper engine crashes peoples machines also with nvlddmkm error.
I mean, this gives some hope, because hat is also something running while the card is idle. Maybe if they can fix that...