Trying To Run The AMDGPU Driver With A Hawaii GPU On Linux 4.5

[duby229]

Since I can run this R9 Fury at full-speed for hours without issues on Catalyst with no signs of thermal problemes, I'm not going to bother trying to clean the heatsink.

Oh well then. Thee number one source of human error, refusal to check possible scenarios. It wouldn't surprise one bit of Catalyst runs only because it throttles properly. The performance of OpenGL in Catalyst is so low anyway that may well be the case. Trial and error, or in other words process of elimination.

[Michael]

I do understand that this is new code that may be the cause of the problem, however I've seen these exact same symptoms get resolved a few times by simply cleaning the heasink and dies, and then apply new thermal compound and reseat the heatsink. It's not hard to do most times, and it would help eliminate a possible scenario.

Since I can run this R9 Fury at full-speed for hours without issues on Catalyst with no signs of thermal problemes, I'm not going to bother trying to clean the heatsink.

[duby229]

Yet the performance is great and no thermal problems when using Catalyst on the same system?

I do understand that this is new code that may be the cause of the problem, however I've seen these exact same symptoms get resolved a few times by simply cleaning the heasink and dies, and then apply new thermal compound and reseat the heatsink. It's not hard to do most times, and it would help eliminate a possible scenario.

[bridgman]

Yet the performance is great and no thermal problems when using Catalyst on the same system?

Yep, definitely sounds like heatsink

Seriously, if the problem is something like bad power tables in the VBIOS then probably best not to run it forced to high for long.

[Michael]

That sounds just like a heatsink that needs reseated.

Yet the performance is great and no thermal problems when using Catalyst on the same system?

[duby229]

Also, separately, trying my R9 Fury with the high performance setting to see if it makes a difference... So far the apparent difference is the card seems to lockup or overheat at least once per benchmark. Lose the display and the fan maxes to 100% noisy setting (haven't SSH'ed in), but then reboot, and can run another benchmark before the same repeats.

That sounds just like a heatsink that needs reseated.

[Michael]

Excellent

Also, separately, trying my R9 Fury with the high performance setting to see if it makes a difference... So far the apparent difference is the card seems to lockup or overheat at least once per benchmark. Lose the display and the fan maxes to 100% noisy setting (haven't SSH'ed in), but then reboot, and can run another benchmark before the same repeats.

[bridgman]

Excellent

[Michael]

After disabling AMDGPU PowerPlay and then messing around with the firmware files some more, looks like I have the R9 290 now playing fine with AMDGPU... Should have some Catalyst vs. Radeon vs. AMDGPU for R9 290 this weekend.

[agd5f]

It worked OK apart from VA-API hardware video acceleration. I presume mesa has yet to implement this for amdgpu (as opposed to radeonsi).

radeonsi is the mesa driver and supports both the amdgpu and radeon kernel drivers. Same with the vdpau, omx, and vaapi mesa drivers.