Nouveau Linux 4.8 + Mesa 12.1-dev vs. NVIDIA OpenGL Performance

[pal666]

We sponsor that manufacturer because

you like https://en.wikipedia.org/wiki/Dancing_pigs

we don't have time to play with our operating system, we want to get things done.

windows is on the other side

You call that "linux-hostile", I call that excellent linux support.

it's not me, it's linus torvalds. i give his opinion on this subject slightly more weight than yours

[dungeon]

are you member of church of nvidia?

You sounds like a guy who hate churches

[artivision]

I've got the idea that Nvidia has superior software (but not hardware) and they try defend their advantage improving the closed source driver and not helping open source knowledge sharing

Nvidia's hardware is the scientifically correct to do, small MIMD units like neurons, close to memory as possible, no real optimizations needed, very efficient. But the ASIC way is to pack the maximum you can for games, like heavy SIMD, not the scientific way. So the TitanX had 6.5TFlops but it was efficient regardless of the task wile FuryX had packed 8.5TFlops and gave the same result. But now it is in your hand to make your game efficient with asynchronous shaders and a FuryX can beat a GTX1080. So i will say a flop is always a flop and I'm sorry for those who bought the less flops.

Also superior software is not the problem here. First because there isn't and second because I spoke just for open power management, so I can run whatever I want to my ASIC like Gallium Nine. Anyway If Nvidia thinks they are right, then lets put it in Wikipedia under the terminology NSIC and not ASIC, if it is to run only what Nvidia likes.

[Amarildo]

Superior software is a problem for AMD, because their drivers introduce a lot of overhead while the NVIDIA's driver doesn't. That's why when AMD cards get the games closer to the "bare metal" with Vulkan they see a large improvement, while NVIDIA's cards don't see much (because their drivers are close to optimal already).