Linux 5.8 To Better Deal With Critical Thermal Faults For Radeon GPUs

[phoronix]

Phoronix: Linux 5.8 To Better Deal With Critical Thermal Faults For Radeon GPUs

AMD on Tuesday sent in another batch of feature updates for Linux 5.8 with the cut-off for new material upon us with this next kernel cycle expected to begin in early June...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-5.8-AMDGPU-Critical-Therm

[remenic]

What genius thought it might be a good idea to just signal the OS when it overheats, so it can receive a signal back to shut down, in stead of just shutting down and preventing damage? Are there use cases where you simply would want to keep the card going until it dies? Why is this controlled in software???

[ssokolow]

What genius thought it might be a good idea to just signal the OS when it overheats, so it can receive a signal back to shut down, in stead of just shutting down and preventing damage? Are there use cases where you simply would want to keep the card going until it dies? Why is this controlled in software???

Is there information I'm missing? To me, it looked more like "Let's no longer have the hardware safety cut-offs pull the rug out from under the GUI by warning it as we're getting close so it can take mitigating action."

[Brisse]

remenic ssokolow Why not just have both? A hardware cut-off at a critical temperature, then an additional software cut-off signal just below the critical temperature? I wouldn't be surprised if that's what they have, and in that case it makes sense to try a graceful shutdown first.

[remenic]

remenic ssokolow Why not just have both? A hardware cut-off at a critical temperature, then an additional software cut-off signal just below the critical temperature? I wouldn't be surprised if that's what they have, and in that case it makes sense to try a graceful shutdown first.

Not saying there shouldn't be any alerts when the device is overheating, that should always happen so the software can avoid a complete crash. But from what I gather, without this driver update the device would break from overheating.

UPDATE:
Upon reading the article again, it seems there are both software and hardware CTF events, and hardware CTF events were already acted upon, I suppose. So, I'm spreading FUD. Shame on me.

[FireBurn]

Hmm, will this improve things on PRIME laptops? Sometimes when I'm gaming the laptop overheats (usually because I'm blocking the vents playing in bed) I can usually alt-tab out of the game and the system is kinda usable, but I can't do anything with the discreet graphics until a hard reboot - is there a way to manually reset the card, power it off and back on without the hard reboot?

[ix900]

Hmm, will this improve things on PRIME laptops? Sometimes when I'm gaming the laptop overheats (usually because I'm blocking the vents playing in bed)

The technically shouldn't fix incorrect use situations in any way at all though if they shut it down on you they are doing you a favor since you're incapable of correct use.

[Danny3]

Nothing about RDNA 2 ?
Will they release it this year or not?

[andrei_me]

Nothing about RDNA 2 ?
Will they release it this year or not?

If RDNA2 is as different as they say it is, probably the patches will come later, close to launch so they don't reveal cards spec before the time.

It happened with Vega AFAIR, it was same day git support 🧐

[xcom]

Please help me in some question.
A 210 W Vega 56 card, why has 180 watt powercap in sysfs?