Valve Developed An Intel Linux Vulkan GPU Driver

[gens]

Lol, Mantle, Vulkan and co are all about the CPU becoming the bottleneck, which is not really an issue for Intel CPU/GPUs yet.

if they share memory, then it is

[eydee]

It is very simple to detect if Optimus is available. In case of Linux glxinfo would not show anything from Intel if "xrandr --setprovideroutputsource" is used. Can't be hard to check on Windows.

We are talking about Steam. It won't even detect VRAM, not to mention optimus. It actually has zero optimus support, that's why people have to put bumblebee (or whatever it is) commands into the launch options game by game. Even though Steam is patched up and ported to linux, its heart is still written for windows 95 and multi-GPU systems were only a dream back then.

[rikkinho]

??

We are talking about Steam. It won't even detect VRAM, not to mention optimus. It actually has zero optimus support, that's why people have to put bumblebee (or whatever it is) commands into the launch options game by game. Even though Steam is patched up and ported to linux, its heart is still written for windows 95 and multi-GPU systems were only a dream back then.

what? steam don't need to have optimus support, optimus is problem from drivers, you can launch steam with bumblebee or nvidia prime, and it will assume the nvidia card... the problem is with x, nvidia and linux not valve

[Kano]

@eydee

It would be better it you think before you write. Bumblebee works with an on-demand started 2nd xserver as the Nvidia hardware is switched oft with bbswitch while not used. That's undetectable via glxinfo without preloader, but the officially supported variant has the Nvidia driver active all the time and can not save much energy while you don't run games, vsync is currently not supported as well. But the speed you can get is much higher than via Bumblebee. Btw. the xrandr code that Nvidia is using does not only work with modesetting, but with Intel and even radeon as output device as well. I used a HD 4550 and GT 630 OEM in a Q67 Intel board to check.

[rikkinho]

?

@eydee

It would be better it you think before you write. Bumblebee works with an on-demand started 2nd xserver as the Nvidia hardware is switched oft with bbswitch while not used. That's undetectable via glxinfo without preloader, but the officially supported variant has the Nvidia driver active all the time and can not save much energy while you don't run games, vsync is currently not supported as well. But the speed you can get is much higher than via Bumblebee. Btw. the xrandr code that Nvidia is using does not only work with modesetting, but with Intel and even radeon as output device as well. I used a HD 4550 and GT 630 OEM in a Q67 Intel board to check.

with ubuntu 12.04.5, 14.04 14.10 you can switch cards (some type of fixed mode) and with intel nvidia card is off. nvidia prime don thave sync with monitor and is horrible to use, i have 2 optimus laptops, 1 intel/amd, 1 with intel, 1 nvidia, 2 desktops (nvidia and amd) and a laptop with apu with windows 7/8, ubuntu, manjaro and one arch

[mmstick]

So much for the anti-Valve haters on here that would consistently state that Valve never did anything good for FOSS and the FOSS community.

[haagch]

Obligatory comment that optimus is only a problem because of nvidia's closed source driver.

With amd + PRIME the problem has had a solution for some time now where you can set up in driconf what gpu a given binary should be rendered with:

http://lists.freedesktop.org/archives/mesa-dev/2014-May/060131.html

Ehm ... yeees?

It's written to large parts by full time employees of amd and commited to mesa with accounts with their company mail addresses. If that's not official, what is?

[Marc Driftmeyer]

Take notice to many who think Intel is some sort of great supporter of Linux. Valve has far fewer resources to leverage and manage to do this work.

[kruger]

[liam]

Not sure what you are trying to say.

I was responding to

Although, Vulkan on Intel seems kinda pointless, will be GPU limited anyways.

in particular.
The point was pedantic if I understood you to be saying that Intel GPUs aren't worth the effort of supporting vulkan because of their, relatively, low ceiling performance. I was simply saying that all GPUs will be GPU limited in scenarios where the bottleneck isn't the CPU... but in a more succinct way