Southern Islands Support Will Come To AMDGPU On Linux 4.9

[pq1930562]

But short hair and different font matters i guess from marketing POV...

Yes. That is absolutely correct sir. And different letters.

Anyway, just realized that none of the animations and sounds that have been shown in this thread so far are as cool as this sound and animation:

[dungeon]

Anyway, just realized that none of the animations and sounds that have been shown in this thread so far are as cool as this sound and animation:

Better listen this sound, but avoid looking at facial expression

[dungeon]

Indeed, my card doesn't work with radeon nor amdgpu.
Just for the heck of it I tried Ubuntu's 16.04 livedvd, result below lol

That is scrambled image developers enjoy to see It is probably just one harvy quirk needed somewhere

[pq1930562]

Better listen this sound, but avoid looking at facial expression

What are you getting at with that video?

First of all, it's geo-blocked by YouTube, so can not watch it. But found another video of that song on YouTube. And it doesn't seem to have anything to do with this thread at all.

[bridgman]

Indeed, my card doesn't work with radeon nor amdgpu. Just for the heck of it I tried Ubuntu's 16.04 livedvd, result below lol

Yeah, it's not supposed to look like that. Just checking, but does the card work with some other driver (eg Windows or Linux Catalyst or amdgpu-pro) ?

If it works with Windows but not with the open source Linux drivers, was it ever reflashed with VBIOS from a different card at any point ? I'm asking because current Windows drivers (and to a certain extent recent amdgpu-pro kernel drivers) rely a bit less on VBIOS than the open source drivers do.

[starshipeleven]

the problem i talk about has nothing to do mit opensource drivers.

this problem for example the usage of "OpenGL-Compatibility-profile" is inside of the workstation applications.

That was fixed with Vulkan, now games can have an API that evolves fast as GPUs, while workstation applications can finally have an OpenGL that remains set in stone for the next millenia.

Or I'm still not understanding.

[bridgman]

It works just fine on windows and linux (with catalyst, as that is what I'm currently using on Arch). Is it supported by amdgpu-pro (didn't know that it was since it's GCN 1.0)

Not supported by amdgpu-pro yet but there are a number of different franken-stacks floating around and being tested... sometimes useful info comes out of that.

Now that you mentioned the VBIOS, the only thing that I did to it was to use this tool https://www.techpowerup.com/forums/t...-cards.189089/ to raise the TDP limit.

Shouldn't necessarily cause a problem (editors tend to be lower risk than flashing VBIOS from a different card) but might be related. If editing & reflashing caused the problem then editing out the changes you made wouldn't necessarily help - do you happen to have to original VBIOS image saved away ? Don't flash it back yet until we have a chance to hear if this is a problem anyone else has seen with stock cards.

User dungeon mentioned you commented in bug tickets previously... do you have the ticket #'s handy in case there is any useful info there ?

[juno]

Shouldn't necessarily cause a problem (editors tend to be lower risk than flashing VBIOS from a different card) but might be related. If editing & reflashing caused the problem then editing out the changes you made wouldn't necessarily help - do you happen to have to original VBIOS image saved away ? Don't flash it back yet until we have a chance to hear if this is a problem anyone else has seen with stock cards.

Wouldn't it make sense to have defaults in the driver, in case the vbios is "corrupt"? For voltages, clocks, memory timings, temperature and power limits? This would maybe make factory or manually overclocked and tweaked cards worse, but at least ensure functionality.

The whole BIOS thing is a mess anyway, imho. At least it is not necessary anymore for Windows-users w/ the latest cards as they got WattMan

I've tried my Hawaii card last weekend on Linux again and was not able to go >60 Hz (monitor supports 144) without getting flickering and corrupted images. It works on Windows. I lowered the ridiculously high voltages long time ago by editing the vbios. So this might actually cause the corruption?

Idk if I still have the old VBIOS, will look for it tomorrow.
Here's the bug ticket: https://bugs.freedesktop.org/show_bug.cgi?id=60879

Techpowerup maintains a VBIOS database, if you are lucky, you might find your card.

[juno]

Guest If you have dual bios, there should be a backup or differently configured second bios stored on the card. There will be a switch on the side of the pcb to select which one to use. Sometimes one is a write protected backup, sometimes there are quiet vs. boost versions to choose from. Either way, if you move the switch you should be back to a original vbios without the need to flash again.
Assuming sapphire did this right, not like Asus with my card.

[bridgman]

Wouldn't it make sense to have defaults in the driver, in case the vbios is "corrupt"? For voltages, clocks, memory timings, temperature and power limits? This would maybe make factory or manually overclocked and tweaked cards worse, but at least ensure functionality.

Yeah, the problem is "defaults for what ?".

VBIOS also controls how the chip identifies itself to driver/OS, so you can't assume the chip ID the driver reads actually matches the silicon on the board. I don't know if we print enough info on the front of the chip to reliably identify both chip and revision, but I imagine we do... but then the heatsink etc.. needs to be pulled off which brings its own problems.

I guess the driver could include default values for a number of different revisions & batches for each chip and let the user try each one via boot parm or something until they find one that works, but that's the kind of hassle VBIOS is there to avoid.