Valve Revises Steam's December 2020 Linux Marketshare To 0.74%

[qarium]

Because their GPUs work, along with their drivers.

HDMI is straight up broken on Polaris (I can personally confirm it across 5 different GPUs; a GTX 1060 was fine though), and I've heard reports about HDMI issues even with RDNA2. And although this isn't specific to Linux, AMD broke Oculus Link (VR) support on Polaris and Vega GPUs (GPUs that should be more than mature by now) about 5-6 driver releases ago on Windows and doesn't seem interested in fixing it.

HDMI1.0-1.4 is broken yes. but HDMI2.0-2.2 are fine.
i fixed that problem by replacing all the TV and monitors to HDMI2.0 or higher.
i think it was 4.7rc1 kernel or 5.7rc1 who was the last kernel for the old HDMI1.X
there is a linux command who fixed it at startup but yes most people do not know this.

[Kano]

If this is your way how you solve problems then you should better improve your problem solving strategy. I understand that you most likely don't have got a direct comparision between Intel, Nvidia and AMD gfx on the same display device but if you would have got this, then it is easier to dig deeper. For some issues you can use Windows drivers to try things which are not possible (yet) with Linux drivers. Just as an example I have got an older Philips TV (1080p) which works still fine out-of-the-box with Intel and Nvidia (including nouveau) gfx drivers. With the HDMI port of my RX 570 (feel free to send me newer AMD HW) I can get a picture but sooner or later it looks like:

https://forum.kodi.tv/showthread.php?tid=105972

In this thread the solution is mentioned as well, you have to use the RGB mode. Windows drivers can be used to change this setting directly but for Linux the only way to fix this was to modify the EDID and put the fixed one into the initrd (which is no good solution for Linux live systems). As I have got lots of cables from DVI->HDMI or via DP to HDMI or DP to DVI, then DVI->HDMI I found out the connected TV via DVI->HDMI should correct color immediately (different to Windows btw.) but amdgpu does NOT support audio via DVI - this definitely works with Nvidia and Intel! This is the error report:

https://gitlab.freedesktop.org/drm/amd/-/issues/1077

So I would call this issue unsolved. I am sure no Windows user would be ever ask to implement a feature or better bug fix on its own - even if the problem itself does not happen so often. It is certainly possible to test those things very easy. Right now I found this thread, where another one is implementing a way to force the color format via kernel option:

https://gitlab.freedesktop.org/drm/amd/-/issues/476

I hope something like this will be upstreamed and a phoronix article mentions this change. Btw. if you use a DVI->HDMI cable AMD cards have got audio support - very stupid Linux restriction. But as usualy, if you have got deep pockets you can buy new hardware (display) and "solve" it that way. The mesa support is really good for AMD hardware, with some exceptions like "SEGA Mega Drive and Genesis Classics" emulator which seems to like Nvidia more. There are definitely lots of games which work with Linux, but sometimes it can be a painful experience, that's the reason why I see streaming as an option. If you use a web browser then it should have accellerated video decode. But this does not seem so easy. Chrome OS has it, Chrome does not. Firefox is progressing. With Windows even the slowest box can be used (if you forget the virus theads) - this really makes me sad.

[qarium]

If this is your way how you solve problems then you should better improve your problem solving strategy. I understand that you most likely don't have got a direct comparision between Intel, Nvidia and AMD gfx on the same display device but if you would have got this, then it is easier to dig deeper. For some issues you can use Windows drivers to try things which are not possible (yet) with Linux drivers. Just as an example I have got an older Philips TV (1080p) which works still fine out-of-the-box with Intel and Nvidia (including nouveau) gfx drivers. With the HDMI port of my RX 570 (feel free to send me newer AMD HW) I can get a picture but sooner or later it looks like:

https://forum.kodi.tv/showthread.php?tid=105972

In this thread the solution is mentioned as well, you have to use the RGB mode. Windows drivers can be used to change this setting directly but for Linux the only way to fix this was to modify the EDID and put the fixed one into the initrd (which is no good solution for Linux live systems). As I have got lots of cables from DVI->HDMI or via DP to HDMI or DP to DVI, then DVI->HDMI I found out the connected TV via DVI->HDMI should correct color immediately (different to Windows btw.) but amdgpu does NOT support audio via DVI - this definitely works with Nvidia and Intel! This is the error report:

https://gitlab.freedesktop.org/drm/amd/-/issues/1077

So I would call this issue unsolved. I am sure no Windows user would be ever ask to implement a feature or better bug fix on its own - even if the problem itself does not happen so often. It is certainly possible to test those things very easy. Right now I found this thread, where another one is implementing a way to force the color format via kernel option:

https://gitlab.freedesktop.org/drm/amd/-/issues/476

I hope something like this will be upstreamed and a phoronix article mentions this change. Btw. if you use a DVI->HDMI cable AMD cards have got audio support - very stupid Linux restriction. But as usualy, if you have got deep pockets you can buy new hardware (display) and "solve" it that way. [..]

i reported this bug i called it Yellow bug to brigman here in the forum years ago it startet on 4,7rc1 or 5,7RC1 kernel i used this old kernel longer because of this bug.
but the answer from bridgman was: this bug is a hardware bug in the display they do not want to fix it. i can use some commando on boottime to fix it. yes the commando works but there is no GUI for user to aktivate it without deep knowlege of the bug.
this hardware bug is fixed in HDMI2.0 so if you buy a new display with HDMI the bug is gone. and this is what i did more than 3 times.

the windows driver has the same problem but amd/nvidia/intel keep big databases of collecting data what display have this bug and they activate the software patch what deactivate features.

[qarium]

my RX 570 (feel free to send me newer AMD HW)

do you remember? i did send you AMD hardware in the past ;-)

i wan to buy a 6800(xt) so my Vega64 will be free. is a Vega64 a usefull upgrade for you ?

[Kano]

You know that kernel 4.7 was released 2016 and kernel 5.7 was released 2020? This is only a range of 4 years. Did you try a DVI to HDMI cable?

[qarium]

You know that kernel 4.7 was released 2016 and kernel 5.7 was released 2020? This is only a range of 4 years. Did you try a DVI to HDMI cable?

yes true i deleted the orginal yellow bug post here at phoronix. because of this i do not know the exact kernel version anymore.
but i wrote a bug report here to brigman and others long time ago.

i fixed this bug by buying HDMI2.0 monitors/TV... means i can no longer reproduce this bug.

[Kano]

Well it should not be so hard to find out the year. You could check the date when the screen was made (my TV is from 2009)

[FONT=monospace][COLOR=#000000]find /sys -name edid -exec edid-decode {} 2>/dev/null \;|grep -i -e manu -e made[/COLOR][/FONT]