Linus Torvalds (torvalds@linux-foundation.org) on 10/2/09 wrote:
---------------------------
>a reader (a@b.c) on 10/2/09 wrote:
>---------------------------
>>>Or you can get a GMA500-based thing, which apparently will
>>>not run Windows to satisfaction, and which has a horrible
>>>driver situation under Linux.
>>
>>aren't you the gatekeeper of that sort of stuff?
>
>Undocumented chipsets and 3D graphics drivers? No. The
>kernel does some of it (memory management, mode setting
>etc), but we need docs for that part too.
>
>And most of the 3D graphics pipeline is in user space OpenGL
>libraries, and in X etc.
>
>>linux is not doing well in netbook. don't you kernel guys
>>have some levers to pull the get a decent powervr driver?
>>it's *widely* used. so might worth the trouble?
>
>Oh, it would be worth some trouble, but it needs to be
>done by Intel. Get the docs out and preferably writing
>at least reasonable starting-point open source drivers
>too. But the OpenVR core isn't even Intel IP, and apparently
>they screwed up when licensing it, so they can't do it.
>
>Yet. I'm hoping/assuming that they are working on it.
>Intel certainly needs some kind of low-power
>graphics solution, and Larrabee isn't it, and they seem
>to be talking about shutting down their traditional IGP
>work.
>
>We'll see. But as-is, GMA500 just doesn't fly very well.
>
>>>Or you can do the crazy ION thing, which means that more
>>>of the money goes to nVidia than Intel, and considering
>>>the track record is pretty risky anyway. Right now, any
>>>sane person would likely avoid nVidia chips for anything.
>>
>>why? ION is not a bad platform. won't be shocked if
>>it gains market share this Chistmas season.
>
>Well, obviously it will gain market share, since it's just
>starting out. And it certainly performs better on a
>graphics side than GMA950. But Ion+Atom loses out to any
>standard CULV+GMA950 setup (it's not like people are buying
>those things for games), so in the end the upside to it is
>exactly what again?
>
>And the kinds of things nvidia has been doing the last
>few years, I'd suggest that vendors and consumers be a
>bit worried. I certainly wouldn't buy their stuff. Not
>on the low end (where non-GMA500 intel integrated graphics
>is fine) or at the high end (where AMD is spanking them),
>or in laptops (where they now have a history of trouble,
>and then lying about it to consumers and vendors
>alike).
>
>And of course, in the Linux space, nvidia really isn't
>much better than GMA500.
>
>Linus

The closed source nvidia drivers actually work on a wide range of distros with disparate kernel versions and build options, which puts nvidia ahead of GMA500 on the closed source driver front. And the nouveau driver (reverse engineered, open source) exists, partially works, and is progressing slowly, which puts nvidia ahead of GMA500 on the open source driver front too.

So nvidia support is less bad than GMA500 support, which is more a comment on the sorry state of GMA500 support than it is an edorsement of nvidia. On the other hand, Linux drivers for AMD/ATI graphics leave alot to be desired at the moment too - though the open source drivers are improving rapidly.