Rx 470 disaster troubleshooting

[Bacon1]

Could be a monitor issue, does the signal come back after unplugging and re-plugging in the cable? But yeah I'd say sell it on ebay for a nice profit at this point. Miners won't care about a display issue as those boxes are headless anyway and its likely a combo with some other hardware that is causing the issue, as you stated the RMA place tested it for hours w/o reproducing the issue.

 

[bononos]

Try using hdmi instead of display port. Or try changing the scaling mode to full panel in the settings. It seems to solve issues with random black screen screens that happen.
https://www.bleepingcomputer.com/forums/t/630830/rx480-random-black-screen-during-gaming/

 

[cen1]

Still here, still with the same problem with rx 470 or perhaps the whole system, things are turning more bizzare by the minute.

I can mine cryptocurrency for days 24/7, I can leave unigine running for hours, aida64 stability test was running now for an hour, system works perfectly. Desktop works absolutely no problem.

As soon as I start specific games whole system shuts down in minutes.

Specifically today, I managed to reproduce a crash (system just "shuts down" completely!) three times with this:
1. Load up Witcher 1
2. Load up a specific save file
3. Rest at campfire, get out of meditation, system shuts down.

GPU is now in secondary PCIe slot, when it was in first slot it only caused the screen to go dark. Second PCI slot causes full system shutdown. CS:GO experiences same problem, Fifa16 no crash for hours. So far it seems that very specific games or scenarios cause the crash.

These are AIDA logs, I don't see anything wrong with the temps or anything else really. GPU goes to 99% load, that is about it.
https://drive.google.com/open?id=0B0glACajmi8tMU54Z3B6ZUdhRHM
https://drive.google.com/open?id=0B0glACajmi8tYTY5MTdHdzRmVms

Is there any other diagnostic tool that could gather more info about the system at the time of the crash? It seems that now I have decently reproducible scenario so it would be worth investigating further.

I'd like to try some more stuff with this Witcher scenario if people here have some smart ideas, otherwise my only course of action is to send the whole thing to service (but who knows if it will fail with standard tests?) or find/replace parts one by one (mobo, cpu, ram).

 

[Shmee]

To start with, I would say to run basic tests on other components. Run memtest on your RAM overnight for starters. I recommend the UEFI version of passmark memtest free v7.3. Also, what CPU do you have? It could be the CPU, believe it or not, as it contains the PCIE controller. I have seen this before in certain systems, where the CPU passes p95 etc, but when stressing the GPU in a certain way, system fails. Replace the CPU and everything is fine.

Do you have a secondary card to test the system with? If it is fine with another GPU, we know it really is the 470. If not, then suspect motherboard or CPU.

 

[cen1]

It's i5-6600. I'll try to get an nvidia card somewhere, if that works it should shift the blame to mobo/cpu/ram.

At this point it's hard to say that this would be power or temperature related since all heavy benchmarks run without a problem, I even left furmark running for half an hour.

 

[VirtualLarry]

Try bumping the uncore voltage a tad, and shutting off "C-States".

 

[altfive]

Interesting enough I got a very similar setup with the exact same problem (I've confirmed watching the video).

- Z170A M5
- i5 7600
- XFX RS RX 470 (Black Edition)
- EVGA 750W GQ (80 gold)

Worth bringing up that the whole PC is brand new.

Also worth mentioning that the problem happens any time, not just gaming. That counts idle or mining.

I've already tried:

- All sorts of driver versions
- Underclocking/Overclocking the GPU
- Other PCI slots

Try bumping the uncore voltage a tad, and shutting off "C-States".

Thanks for mentioning "C-states", sounds plausible - but it hasn't worked either. I'll try this uncore voltage.

I really don't know what to do at this point. RMA the card will be a painfull process, but if there's no other way..

@cen1 You had success with your's? Would really appreciate a follow-up here!

Edit: I still can hear the sound of music and other things on the PC. But any other process using the VGA apparently just crashes/dies.

 

[altfive]

Noticed that even if the PC is idle and the problem happens, the coolers turn on.

 

[Chicken76]

cen1, what model is your monitor?