I recently snagged a 3080 XC3 Ultra and put it into my main rig, which is several years old at this point, upgrading from 1080. Everything seems to be running fine, but while in games I'll get random USB dropouts every 10-15 minutes that last ~10 seconds. All USB disconnects, including the internal headers are being used to fan RGB control and PSU monitoring. Nothing else appears to be impacted.
 
I normally would suspect PSU, but I'm using a HX750i that should be good enough. When the dropouts occur the PSU is sending ~350W, which isn't even peak when gaming (425-450W). This is also well within what the PSU can handle. The dropouts also happen within like 50W of what the 1080 pulls, so it's not like it's a crazy difference here. I did also confirm that swapping back to 1080 corrects the problem.
 
My main suspect is the motherboard simply because it's fairly old at this point and is pretty heavily loaded between the 3080 and 2x NVMe drives.
 
My full specs are:

• CPU: Intel 6700k
• Mobo: Gigabyte Z170x Gaming 6 (lastest BIOS)
• PSU: Corsair HX750i
• 2x16gb Corsair CMK32GX4M2A2666C16
• NVMe 1: HP EX950 2TB
• NVMe 2: Intel SSDPEKKW512G7
• HDD: Western Digital Caviar Black 2 TB

 
My original plan was to upgrade the rest of the system to Alder Lake once it drops, but I may need to revisit that plan. Does my hunch that the mobo is the culprit rather than the PSU make sense? Or should I really try replacing the PSU first?

I was not running an overclock, and I did attempt resetting the mobo to defaults (which should have also exercised fast boot disabled). I am running latest Windows 10 and definitely rebooted quite a few times while debugging.
 
For RGB it's controlled by Corsair iCUE, but I did test with it closed and the services stopped.
 
The thing that's really throwing me is that only USB is impacted. When it happens CPU and GPU clock is relatively stable.
 
I should also note that there's nothing in Event Viewer about the dropout, only entries are from when everything reconnects. Also almost all USB devices are connected via a powered USB hub. Other than the internal headers, the only things connected directly are the Logitech Unifying Receiver and a Yubikey... not exactly high powered devices.
 
Edit: the only other reason I don't suspect software issues is due to the USB dropouts only occuring when the GPU is under some amount load. If I idle at the desktop or do some web browsing it'll be fine, but if I idle in a game menu it'll occur eventually.

It seems a PSU problem to me. Before getting a new PSU,  Try updating the nvidia driver first.  Clean install, using ddu to uninstall the current driver.  Install the latest driver.    

Corsair iCUE has been reported many times in the past as the root cause of issues ... I do not use it, so I can't test it
 
 
could it be a damaged cord on Any USB device ? ... test with new mouse & KB ,,, unplug all other USB hubs & devices

Okay so I went a bit nuclear and:

1. Reset BIOS to default again
2. Reinstalled windows (though I did reinstall iCUE)
3. Removed the USB hub and plugged everything directly into USB ports on the mobo

I could still reproduce the issue. I started removing USB devices until I got down to mouse, keyboard, unifying receiver, Xbox wireless adapter, and Yubikey. Still happened.
 
I then removed the Xbox wireless adapter and uninstalled iCUE. After that I was able to avoid a dropout for ~40 minutes, which is pretty good! Will need to spend more time with it to see if it has stabilized, and then figure out which of those two was the culprit.
 
Though it's still confusing to me why the 3080 is causing a difference here.
 
Edit: regarding temps, everything seemed fine. The hottest anything got was the 3080 at a peak of 79. I did also disable USB suspend in the power plan settings. I didn't disable per device power management this time, but I did try that before I wiped windows.

devrand
Okay so I went a bit nuclear and:

1. Reset BIOS to default again
2. Reinstalled windows (though I did reinstall iCUE)
3. Removed the USB hub and plugged everything directly into USB ports on the mobo

I could still reproduce the issue. I started removing USB devices until I got down to mouse, keyboard, unifying receiver, Xbox wireless adapter, and Yubikey. Still happened.
 
I then removed the Xbox wireless adapter and uninstalled iCUE. After that I was able to avoid a dropout for ~40 minutes, which is pretty good! Will need to spend more time with it to see if it has stabilized, and then figure out which of those two was the culprit.
 
Though it's still confusing to me why the 3080 is causing a difference here.
 
Edit: regarding temps, everything seemed fine. The hottest anything got was the 3080 at a peak of 79. I did also disable USB suspend in the power plan settings. I didn't disable per device power management this time, but I did try that before I wiped windows.

It is impossible to say why exactly 3080 is making a difference but theory that it is because of its higher power requirements is a plausible one. When you removed other devices you reduced power requirements and when you disabled USB selective suspend you told Windows not to try to preserve power that goes to USB devices, which doesn't always work correctly, see here What is USB Selective Suspend feature? How to enable or disable it? (thewindowsclub.com) Nvidia cards have USB ports and their own USB drivers so it can be a conflict.
 
Keep in mind disabling USB power cut just through selective suspend in plan does not work all the time everywhere so you might end up having to disable it in device manager too.
 
It can also be iCue interfering with devices / device drivers. iCue is not exactly having the reputation of being headache free software. Keep in mind you need to do following manual steps in order to completely uninstall iCue Perform a clean reinstallation of the Corsair Utility Engine (iCUE) – Corsair
 

Okay, so after some more experimenting after work today I found it's a combination of iCUE and the HX750i's USB header. With either one of those removed (or both) the issue seems to go away. Not entirely sure why this only reproduced with the 3080, but things seemed to have stabilized now.
 
Thank you all for the input!

ZoranC

devrand
Okay, so after some more experimenting after work today I found it's a combination of iCUE and the HX750i's USB header. With either one of those removed (or both) the issue seems to go away. Not entirely sure why this only reproduced with the 3080, but things seemed to have stabilized now.

 
Are you saying you had HX750i connected with USB cable to motherboard and had issues until you disconnected it or removed iCue? If yes that is VERY interesting because I had experienced iCue freezing on me when I had my AX1600i connected to USB and my motherboard is X299 dark and video card is EVGA 2080 Super. After that experience I "banned" iCue from my systems.

Correct. I confirmed that the dropouts stopped if I removed iCUE or unplugged the PSU's USB cable. It currently seems to be stable with iCUE running and the PSU unplugged.

I was going to ban iCUE but I decided to give it a chance and turned on debug logging and waited for a dropout. Obviously there's a bunch of noise in the logs when the dropout occurs, however a second or two before that there were a bunch of communication errors with the PSU. That's what ultimately got me to test with the PSU's USB unplugged.

I uninstalled iCue a while back.  It was using 5% of my CPU all by itself.