Hetz Ben Hamo wrote:
> Maybe they have the best team - SGI knows what they're doing, VA too, and
> ofcourse - NVidia...
> But this "best OpenGL driver groups around" provides the user with what?
> 1. A NON STABLE driver! on 7 out of 8 machines here (read my previous email) the
> driver remains unstable, and causing the machine to hang. All of them using
> 1280x1024 resolution, with different processors, RAM and boards.
Try disabling AGP in your XF86Config file (add Option "NvAgp" "0" to
your screen section). I have two machines at home and 5 machines at work
(all but one are dual processor) with various combinations of LX and BX
chipsets, with TNT, TNT2 and GeForce2 cards. All of these machines are
rock solid and get heavy GL use daily. Even with AGP disabled, a
GeForce2 will outperform any of the other options for Linux at this
> 2. Outdated driver - the nvidia 0.9.5 driver doesn't support Geforce 2 Ultra -
> and NVidia points people to use XFree 4.0.2 with this card.
Although I have no direct experience with the Ultra under Linux, these
guys apparently had no trouble with it:
> So what do we have at the end: a joint effort to port their Detonator 3 driver
> to Linux, and from time to time a very minor update of their kernel module.
> Thats it. Nothing more, no more optimization, no DRI support, no hardware AA
> (Anti Aliasing) support, no tech support for this driver (other then a FAQ, an
> IRC channel which you need to pray that someone there can assist you).
* The drivers are already optimized - as ports of the Windows drivers,
they benefit from several years of optimization and maturity.
* DRI support is irrelevant, because they use their own proprietary,
high performance direct rendering architecture.
* The hardware doesn't support AA, but the drivers do support FSAA in
* There's no official tech support for the DRI drivers either, just a
bunch of helpful people on a mailing list. Just like there's a bunch of
helpful people on the NVIDIA IRC channel.
The way I see it, NVIDIA is being very fair in their treatment of
Linux. Their support for Linux is no more, and no less than the support
that they offer for Windows. I expect that their level of support will
be the same for Macs, as well.
Until there is a combination of DRI/hardware combination that exceeds
NVIDIA's offering, the NVIDIA drivers benefit the Linux community.
Professionals get complaint, high performance, mature drivers. Gamers
get good Quake frame rates. None of the DRI offerings that I have tested
provide the compliance and/or performance that I require to do my job.
If it weren't for the NVIDIA drivers and hardware, I would be forced to
Bashing NVIDIA serves no constructive purpose.
As for the open source issue, NVIDIA have stated that they cannot open
the drivers because they don't own all of the code. I have no way of
knowing if this is true or not, but I have no reason to doubt them. If I
were head of software engineering at NVIDIA, I would definitely choose
the cross-platfom approach that they have taken. They maintain one
codebase for all supported platforms, and all optimizations and bug
fixes that occur benefit all platforms. From their perspective, this is
a huge win.
Matrox, 3dfx and ATI all have well deserved reputations for crap OpenGL
drivers under Windows. These companies will benefit greatly from the
expertise and hard work of the DRI developers. But NVIDIA already has
very good drivers, and although a DRI solution could probably match
their existing drivers, it would be difficult to exceed them.
If NVIDIA were to release specs to their chips tomorrow, would that
necessarily be a good thing? The current DRI drivers are not at maturity
yet, and developer resources are thin. Would it be worth the effort to
start a new driver from scratch when there is so much work to be done on
the other ones? Personally, I think it makes a lot more sense to spend
the time making the current drivers solid, compliant, and fast than
developing a redundant codebase for the NVIDIA drivers (of course the
BSD, Alpha, PPC, etc. people will disagree with me, with good reason).
Open specs and/or open drivers are a good thing, but let's not forget
that 3dfx was binary only for a long time before they opened up (and
they never released specs for the VSA chips, anyway, did they?) If
someone can come up with a solid argument as to why opening up would be
to NVIDIA's benefit, maybe they will open up as well at some point.
And let's not forget that NVIDIA have been very solid supporters of an
important industry standard. They have exposed all of their DX8 features
through OpenGL extensions. Microsoft only wanted DX on the XBox, but
NVIDIA will be providing OpenGL for it (does anyone honestly think that
ATI would have gone to that effort if they had won the contract?)
They're not evil, they're not bastards, they're just treating everyone
as equals, and that deserves respect, IMHO.
I apologize for posting a long rant on NVIDIA on the DRI developer's
list, but I keep seeing people bashing NVIDIA, which will only end up
being counter productive. If you don't like NVIDIA's approach to Linux,
just ignore them.
If fully featured Radeon drivers were available today, I would be using
a Radeon instead of a GF2 (mostly because I really, really want 3D
textures). But they aren't available yet, and there's no target date.
The NVIDIA drivers allow me to use Linux for my daily job, and to
convince others that Linux is a viable alternative to NT and IRIX for
many vis sim and sci vis applications.
Mark B. Allan NASA Ames Research Center
QSS Group, Inc. Neuro-Engineering Lab
650 - 604 - 0537 (office) Mail Stop 269-2
650 - 604 - 0461 (lab) Moffett Field, CA 94035
650 - 604 - 3594 (fax) mallan@...