7900 XTX - Monitor continually wakes up when on DisplayPort and HWiNFO64 is running

[AMD718]

I pretty much have HWiNFO64 running at all times. Was trying everything to figure out why my LG 27GL850 would continually wake and then shut off, rinse repeat, when it should be in a steady sleep state. Determined that it does not do it when connected via HDMI, but this particular monitor is limited in its capability when connected via HDMI, so that's not an option. Was about to chalk it up to a VBIOS or early 7900 XTX driver bug when I thought to see if it still occurs when HWiNFO64 is not running. Turns out normal operational behavior (monitor turns off and stays off until an HID interrupt wakes it) returns. Unfortunately, due to this issue, I'll need to switch my usage with respect to HWiNFO64 and only use it when I specifically need it, instead of having it running all the time.

 

[Martin]

I'm quite sure this isn't a bug or issue in HWiNFO, but the GPU drivers.
Some polling of GPU parameters is triggering activity in the GPU which is preventing entering a lower power state.

 

[AMD718]

I'm quite sure this isn't a bug or issue in HWiNFO, but the GPU drivers.
Some polling of GPU parameters is triggering activity in the GPU which is preventing entering a lower power state.

Thanks, Martin. In that case, I'll submit this as a bug to AMD.

 

[Martin]

I will let them know as well, after holidays.

 

[AMD718]

I will let them know as well, after holidays.

Merry Christmas, Martin! Another, even more interesting development. Not only does the display continuously wake up when on DisplayPort and hwinfo running, but eventually the display driver will crash. I spent the last couple days trying to figure out why my 7900 XTX would crash when it was idle (never when playing games or in actual use). Would come back and the display driver crash had been logged in event viewer. Had to restart (via Chrome Remote Desktop as display driver did not recover properly) PC to restore normal GPU/display operations. At first I thought it was an unstable undervolt. Tested various undervolt values, RAM overclocks, PPT limits, fan curves, and nothing prevented the issue from occurring. Then I tried disabling all wattman tuning, so 100% stock, and still had the issue. At this point I figured my 7900 XTX was defective and was considering returning / exchanging it. Was very strange to me as I could game 100% stable with 1050mV and 2750@fast +15% PPT, yielding a very performant reference 7900 XTX vs stock, while tjunction was always below 96c and gpu core always below 75c. Once I found that keeping HWiNFO64 running caused the issue that is the subject of this thread, naturally I stopped running it until AMD can fix the issue. Well, guess what, no more idle driver crashes. In fact, I was able to put my undervolt/overclock/ppt/fan tweaks back in place and it's still 100% stable at idle. I guess this is even more important for AMD to fix than the DisplayPort monitor sleep state cycling. Cheers, and have a Happy New Year!

 

[Martin]

Merry Christmas to you too!
Well, I'm not surprised about this as this GPU family is quite a big leap from previous generations. It will take AMD some time until bugs are root caused and fixed. I believe we'll see lots of fixes in next driver releases...

 

[AMD718]

Hello, Martin. Happy New Year. I have an update for you. Neither of these issues appear to be actually related to HWiNFO64. There is a thread about it between myself and an AMD RTG engineer here. In the end, the issue appears to be Win11 "Settings -> System -> Power & Battery -> Best power efficiency" power setting, which seems to reduce idle stability on the GPU.

 

[Martin]

Thanks for the update, glad to know that it's not caused by HWiNFO and you found a solution.

 

[AMD718]

Thanks for the update, glad to know that it's not caused by HWiNFO and you found a solution.

Another update, for your own edification (in case you're still interested, even though this is not an hwinfo bug) or in case any other users stumble upon this thread. It turns out that "Power & Battery -> Balanced" does not resolve the idle crashes completely. It does significantly reduce the frequency of them occurring, but ultimately not a solid fix. Additionally, it does nothing for the monitor sleep/wake cycling. However, I've found from a reddit user's comment within one of my threads that settings EnableUlps to 0, for all occurrences in the registry, does 100% fix the monitor sleep/wake cycling. I am optimistic that it is the true root cause of both the idle crashes and the monitor power state cycling. With EnableUlps set to 1, which is what it was set to in all of my previous testing, the 7900 XTX would turn off, even the RGB lighting, and then turn back on, coinciding with the monitor sleep/wake cycling, when the system was idle. Now, with EnableUlps set to 0, the RGB lighting never turns off (which is what I would normally expect), and the monitor never wakes up when it should be sleeping. Hope this helps someone. I won't call a victory on the idle crashes until it's been a few weeks without a single occurrence though.

 

[Martin]

That's interesting as ULPS used to be a feature for Dual-GPUs which was completely switching off the secondary GPU when not used (to preserve power).

 

[AMD718]

That's interesting as ULPS used to be a feature for Dual-GPUs which was completely switching off the secondary GPU when not used (to preserve power).

I wonder if it was turning off my 7900 XTX because it thinks it's a secondary GPU (even though I only have one physical GPU) since I have a Citrix virtual display adapter device installed by the Citrix receiver client. Maybe that throws it off.

 

[Martin]

Well, that would be quite possible...