Hi folks,
I'm posting this up to see if anyone else is having this issue (or to perhaps make those aware who have these issues that this could be the cause). I have already filed a support ticket for it.
To get started, I have the following setup;
AMD Ryzen 5800X CPU
Asus ROG STRIX X570-F Gaming Mobo (running at 1900 MHz Infinity Fabric)
Corsair Vengeance DDR4-4000 (running XMP and clock slowed to 3800 MHz)
Asus ROG STRIX LC 6800 XT GPU
Corsair MP600 1TB NVMe SSD
Windows 10 x64 with all the latest updates, drivers, firmware and BIOS installed
Aida64 6.32.5600 with the sensor polling set to a conservative 2000 ms
Odospace plugin enabled with a Samsung Galaxy Tab2 for the display
When my system is at idle or near idle (surfing, Youtube, Spotify, Remote Desktop etc) it will hang at random times (roughly every 1-2 hours on average). This hang is a complete system deadlock, requiring holding the power button down or waiting for the motherboard watchdog timer to hard reboot the system).
If the system (crucially, the GPU) is under load, the system is stable. I can game for hours with no issues. But as soon as it's idle again, it's a ticking time bomb. I tried all sorts of things such as BIOS and firmware updates, intentionally overclocking the CPU, disabling power level states, raising CPU and SoC voltages in the BIOS, lowering RAM speeds, resetting BIOS to optimised defaults, even reinstalling Windows on a new drive.
Through days and days of trial and error, after I reinstalled Windows on a fresh drive I slowly reinstalled things and I narrowed this down to the Odospace plugin. As soon as I disabled this, my system became rock solid stable again. I went back to my main SSD / OS and disabled the Odospace plugin. No crashes for days, even after reboots and power cycles.
It's worth noting that whilst idle, the 6800 XT GPU SoC voltage and Wattage meters disappear and reappear every few seconds. My gut says there's some kind of race-effect where the sensor poll happens at the moment these sensor values "disappear".
Oddly though, with the Odospace plugin disabled, if I thrash the SMBUS by setting the poll rate to 10x speed (200 ms) and watch the sensor page, Aida64 and the system as a whole is still rock solid stable. I would have expected hammering the SMBUS this hard would have caused a lock up.
Another thing to note, a couple of seconds before my system hangs, any sound playing and the mouse cursor all start to skip / judder. Moments later, it hangs completely. No blue screen of death, no logs in Event Viewer, just a complete system hang.
It's my hope that by posting this up, some other poor soul who is experiencing these issues may be guided to temporarily disabling the Odospace plugin until a fix can be found.
@odospace Please note the above - you may need to get involved with the Aida64 team to patch this bug