Strange full-system blackscreens when under mild GPU load with an AMD iGPU

[Slipfox]

Morning, folks. I just built a PC yesterday, for all intents and purposes it went well for my first build, but I am currently struggling with an issue where, when doing anything that even remotely uses my GPU (i.e playing a game) my system will sometimes just go completely black and unresponsive, even to Win+Ctrl+Shift+B.

This happens across multiple OSes (Windows and Linux), so I believe it to be a hardware issue, and it doesn't happen when I put the machine under a synthetic CPU load (i.e infinite loop batch file on a handful of command prompts), so I think it's an issue with either the graphics chips themselves or something related to them in the BIOS.

I don't think it's a temp issue, either, booting into BIOS directly after one of these instances shows CPU temps are good. Has anyone else had a similar issue with AMD integrated graphics?

 

 

CPU: AMD Ryzen 5 5600G 3.9 GHz 6-Core Processor
Motherboard: Asus PRIME B550-PLUS ATX AM4 Motherboard)
Memory: Corsair Vengeance LPX 16 GB (2 x 8 GB) DDR4-3200 CL16 Memory
Storage: Samsung 970 Evo Plus 1 TB M.2-2280 NVME Solid State Drive
Case: Corsair 4000D Airflow ATX Mid Tower Case
Power Supply: Corsair RMx (2018) 750 W 80+ Gold Certified Fully Modular ATX Power Supply
Wireless Network Adapter: TP-Link Archer TX50E 802.11a/b/g/n/ac/ax PCIe x1 Wi-Fi Adapter
 

[Bombastinator]

1 hour ago, Slipfox said:

Morning, folks. I just built a PC yesterday, for all intents and purposes it went well for my first build, but I am currently struggling with an issue where, when doing anything that even remotely uses my GPU (i.e playing a game) my system will sometimes just go completely black and unresponsive, even to Win+Ctrl+Shift+B.

This happens across multiple OSes (Windows and Linux), so I believe it to be a hardware issue, and it doesn't happen when I put the machine under a synthetic CPU load (i.e infinite loop batch file on a handful of command prompts), so I think it's an issue with either the graphics chips themselves or something related to them in the BIOS.

I don't think it's a temp issue, either, booting into BIOS directly after one of these instances shows CPU temps are good. Has anyone else had a similar issue with AMD integrated graphics?

 

 

CPU: AMD Ryzen 5 5600G 3.9 GHz 6-Core Processor
Motherboard: Asus PRIME B550-PLUS ATX AM4 Motherboard)
Memory: Corsair Vengeance LPX 16 GB (2 x 8 GB) DDR4-3200 CL16 Memory
Storage: Samsung 970 Evo Plus 1 TB M.2-2280 NVME Solid State Drive
Case: Corsair 4000D Airflow ATX Mid Tower Case
Power Supply: Corsair RMx (2018) 750 W 80+ Gold Certified Fully Modular ATX Power Supply
Wireless Network Adapter: TP-Link Archer TX50E 802.11a/b/g/n/ac/ax PCIe x1 Wi-Fi Adapter
 

Could also be a memory issue.  IGPs use system memory, but often the higher registers aren’t used for anything unless a gpu is put under load.  

[Slipfox]

6 minutes ago, Bombastinator said:

Could also be a memory issue.  IGPs use system memory, but often the higher registers aren’t used for anything unless a gpu is put under load.  

Just ran the Windows Memory Diagnostic and nothing bad was reported.

[Slipfox]

Update, I've tried fully removing and reinserting both CPU and RAM, no dice. I'm honestly at my wit's end here, I really don't want to deal with RMA right now since I'm about to go on a trip but I can't think of any reason for my GPU to be failing like this besides a hardware fault.

[Bombastinator]

1 hour ago, Slipfox said:

Just ran the Windows Memory Diagnostic and nothing bad was reported.

Sadly the windows memory stuff has a terrible rep. Pulls a lot of false negatives.  It’s so bad it defaults to run itself twice. There’s a freeware program called memtest86 which is considered more reliable, especially if run alll the way through more than once.  It isn’t particularly fast though.

[VladTheImpaler]

Agree with Bombastinator, memtest86 is much better..

 

Some random ideas...

 

Have you tried a Bios update?

If you're using XMP/DOCP have you tried with it off?

Is your memory on the motherboards certified list? If not check timings or manually set them.. (unlikely but possible even sub timings could cause a problem)

Check temps in real time with HWMonitor or HWInfo.. ( look for VRM temps/GPU temps, not just CPU )

Unlikely a problem, but if you have another PSU maybe test with it..

 

Maybe disable stuff like onboard audio and lan, any extra ports.. In essence try to cripple the system to isolate the CPU/GPU.. I'd hate for you to RMA the CPU to later learn it was a motherboard problem..

 

Edit: Yes you have motherboard heaksinks.. Deleted related

 

 

 

 

 

[Slipfox]

2 hours ago, Bombastinator said:

Sadly the windows memory stuff has a terrible rep. Pulls a lot of false negatives.  It’s so bad it defaults to run itself twice. There’s a freeware program called memtest86 which is considered more reliable, especially if run alll the way through more than once.  It isn’t particularly fast though.

Did two passes of memtest86, nothing failed.

1 hour ago, VladTheImpaler said:

Have you tried a Bios update?

If you're using XMP/DOCP have you tried with it off?

Just tried those, I wasn't on the latest BIOS version but updating to latest and disabling DOCP did nothing.

1 hour ago, VladTheImpaler said:

Is your memory on the motherboards certified list? If not check timings or manually set them.. (unlikely but possible even sub timings could cause a problem)

Just checked the part number, shows on the mobo support list.

1 hour ago, VladTheImpaler said:

Maybe disable stuff like onboard audio and lan, any extra ports.. In essence try to cripple the system to isolate the CPU/GPU.. I'd hate for you to RMA the CPU to later learn it was a motherboard problem..

I guess that'll be next on my testing list.

[Slipfox]

Hm, alright, so I did some testing, and I *may* be wrong? It looks like drivers are only crashing when I play TF2? I say this because when I play a much more intensive game (like ArmA 3 or Fortnite (yes, i know, bad game, but it was free and modern so good for testing)) It doesn't crash as far as i could tell.

[Slipfox]

Nope, was right the first time. drivers just crashed *again* watching a youtube video.

[Bombastinator]

21 hours ago, Slipfox said:

Did two passes of memtest86, nothing failed.

Just tried those, I wasn't on the latest BIOS version but updating to latest and disabling DOCP did nothing.

Just checked the part number, shows on the mobo support list.

I guess that'll be next on my testing list.

So it’s probably not a memory problem.   TF2 specific implies a software problem.  May want to try Re downloading an reinstalling.

[Slipfox]

20 minutes ago, Bombastinator said:

So it’s probably not a memory problem.   TF2 specific implies a software problem.  May want to try Re downloading an reinstalling

As stated in previous reply, I just discovered that it's actually *not* tf2 specific and I was incorrect. Looking like it's gotta be a hardware/driver problem.

[Bombastinator]

2 hours ago, Slipfox said:

As stated in previous reply, I just discovered that it's actually *not* tf2 specific and I was incorrect. Looking like it's gotta be a hardware/driver problem.

So not tf2 specific, then tf2 specific then not again.  OK. 

[Slipfox]

5 minutes ago, Bombastinator said:

So not tf2 specific, then tf2 specific then not again.  OK. 

I had originally presumed it to be a hardware/driver issue, then i thought that a BIOS update and disabling DOCP fixed it for everything *but* TF2, then I realized by doing more testing that no, i was wrong. It is, in fact, a hardware/driver issue, i just made a dumb guess based on little evidence.

[Slipfox]

Update: I went out and bought a dedicated graphics card, which led me to the discovery that this entire issue has been caused by faulty graphics cores in the CPU. Now runs flawlessly.