NVIDIA Rolls Out 370 Linux Driver Beta, Adds Pascal Under/Over-Clocking

[omer666]

Ironically, we won't get such things as 2D VSync or complete Optimus implementation before having Wayland support.

Indeed, open source software can do proper GPU switch, but using alternatives to nVidia's technology. And nVidia is not willing to maintain too much platform-specific code. It is for this very same reason that we're not getting Wayland with their proprietary driver anytime soon.

I hope we will have a clear statement in September about all this mess.

[Passso]

first thing: untrue, (if you think that an always one dedicated gpu means "working optimus" I don't want you ever be part of a QA team. Don't lie to yourself)
second thing: also untrue. (yeah, and that's why open drivers support that already )

... and both things you say: untrue

I use Primusrun for the last 5 years on 3 different laptops, and can dynamically switch to the GPU when some apps require it.

[karolherbst]

... and both things you say: untrue

I use Primusrun for the last 5 years on 3 different laptops, and can dynamically switch to the GPU when some apps require it.

and primusrun isn't prime offloading at all. it is in application memory copies between two OpenGL context running on different X server. I know it works, but that ain't a good thing, cause the pcie overhead is there and it is big. Everything above fullhd@60fps is too much.

[Passso]

and primusrun isn't prime offloading at all. it is in application memory copies between two OpenGL context running on different X server. I know it works, but that ain't a good thing, cause the pcie overhead is there and it is big. Everything above fullhd@60fps is too much.

Yup, but whatever the technical trick used, it is the same result and it works well enough.
Sure you sacrifice a few fps (or far more if you go 4k for laptop gaming... which isn't really smart!)

Anyway, it is great to see real prime offloading.

[karolherbst]

Anyway, it is great to see real prime offloading.

you missed the point. Nvidia doesn't do prime offloading.