Folding Forum

Folding GPU Project 10493 (8,6,150) first froze the computer and had to do a hard restart, BSD crashed then froze again after startup, restarted the computer for the 3th time and stopped the GPU from folding and computer ran fine all running 'GPU driver 16.6.1'. I did a GPU driver update that popped up in the Windows GPU update notification and restarted and tried to fold again on 'GPU driver update 16.5.1' and now I got a notification that the graphics drive crashed and recovered. Project never went past .34% without crashing. I just paused it just to get the project number.

I decided to do a manual update the GPU and now it want's to go back again to 16.3.2 (WHQL) version. So I did that and restarted. Unpaused GPU folding and it I get a notification again on crashed and recovered but the folding kept going? I decided just to cancel that project and have it try another project. It's want's to download but I'll wait until it does and keep you updated.

Windows 10
Radeon HD 8280 APU with driver v16.3.2 (WHQL)
FAH v 7.4.4

Just to note: When I did a manual update it shows in the "Radeon Update Settings" new update is v16.3.1 and clicked on it to do the update. I go back to the "Radeon Update Setting" and it shows v16.5.1 now and there is still a v16.3.2 update.
Right now it's on v16.5.1. Not v16.3.2. I'm reporting that as a bug to AMD.

There are some (not really very many) reports of FAH running on APUs. Some folks have been successful and some have not. This may be associated with a particular driver version so reporting it to AMD is a good start.

FAH uses OpenCL so it may be useful to run test programs that confirm that OpenCL is working as expected. Here's another topic that discusses such testing:Subject: Folding on AMD Radeon 8610G- Driver Problems

FAH also puts a heavy load on th APU by way of OpenCL so it's possible that your hardware designer didn't provide a suitable plan for dissipating whatever heat can be generated at whatever clock-rate your APU is running. If temporarily reducing the clock rate stops the system hang behavior, then you've likely identified all or part of the problem as being associated with heat dissipation.

Let us know what AMD says.

Got new GPU work with no crash. Been running for over half an hour. Work may have been corrupted upon download.

bruce wrote:There are some (not really very many) reports of FAH running on APUs. Some folks have been successful and some have not. This may be associated with a particular driver version so reporting it to AMD is a good start.

FAH uses OpenCL so it may be useful to run test programs that confirm that OpenCL is working as expected. Here's another topic that discusses such testing:Subject: Folding on AMD Radeon 8610G- Driver Problems

FAH also puts a heavy load on th APU by way of OpenCL so it's possible that your hardware designer didn't provide a suitable plan for dissipating whatever heat can be generated at whatever clock-rate your APU is running. If temporarily reducing the clock rate stops the system hang behavior, then you've likely identified all or part of the problem as being associated with heat dissipation.

Let us know what AMD says.

Thanks for the advice. But for one if it was overheating the computer it world turn off. I know this because that fan for the CPU quit on me and had to get a new one. And the bug I'm referring to is the way that AMD updater does it's updates.

Not that it matters in your case: Yes, if a fan fails, the computer will overheat rapidly and any CPU or GPU made in the last 20 years will shut down to protect itself. If partial cooling is present, it may never reach the emergency shutdown temperature, but it still may still exceed thermal limits of stable operation, resulting in software crashes or sometimes just incorrect computations.

bruce wrote:Not that it matters in your case: Yes, if a fan fails, the computer will overheat rapidly and any CPU or GPU made in the last 20 years will shut down to protect itself. If partial cooling is present, it may never reach the emergency shutdown temperature, but it still may still exceed thermal limits of stable operation, resulting in software crashes or sometimes just incorrect computations.

That's true.