[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Nvidia drivers



On Sun, 1 Apr 2001, Juha Saarinen wrote:

>> *DO NOT BUY VIDEO HARDWARE FROM NVIDIA FOR USE IN LINUX*
>
>This is easier said than done. "Consolidation" (read: other vendors have
>gone bust) has taken place in the graphics card industry, and your choice
>of affordable video adapters is fairly limited. It's Nvidia, ATi, and
>Matrox, basically. 3Dfx is no more, e.g.

So one vendor disappeared..  There are many choices out there
still.

>Matrox and ATi don't have the 3D performance of Nvidia's cards,
>unfortunately.

I disagree personally, but I'm looking at things from a Linux
perspective so YMMV.


>> Their hardware may have all sorts of wonderful whizbang features
>> that work in Windows, but for the most part they are useless in
>> Linux.  This is sad because I've seen these cards run 3D
>> stuff and they are decent (in other non-Linux OS's).
>
>That's funny, because I've been running Linux on Nvidia cards only. They
>don't seem useless to me.

Then you don't get bug reports mailed to you all the time to
which you can do very little about.

As I said just below: "Some people are happy with the binary
driver".  I meant that.  *Some* people.  That doesn't mean
everyone.


>> Some people are happy with the binary driver, but that will
>> change the split second something in XFree86 or the kernel
>> changes (such as Render above)  and the driver breaks.
>
>Incorrect assumption... the RENDER extension works. I would like Nvidia to
>support some more OpenGL extensions though.

I didn't say it didn't work.  Read what I said again.

>What's important to remember is that drivers for video cards
>are a "competitive advantage" -- they make all the difference
>in benchmarks, and hence, games weenies' buying decisions.

And hence "Linux" game weenies frustration from ending up with
hardware they have to fight with.

>I assume that Nvidia doesn't want to give away all its secrets
>to the competition.

That is FUD.  You really ought to read Eric Raymond's book
sometime "The Cathedral and the Bazaar".

>They must get some of the coding right, because there's a huge
>performance difference for 3D with the nvidia driver compared
>to the XF86 nv driver, which is supposedly accelerated.

You are way wrong here.  THe XFree86 driver is created with
almost zero knowledge of the hardware.  If you give the XFree86
guys the specs for Nvidias hardware I'll guarantee you that they
will write top notch drivers.  They have *hired* one of the
XFree86 developers also, which possibly means that better open
source support may be forthcoming.  Comparing the binary nvidia
driver to the XFree86 driver is completely apples and oranges.


>> If *ANYONE* wants *GOOD* hardware that works incredibly well in
>> Linux, buy hardware from *ATI* and *Matrox*.  Both ATI and Matrox
>> cards are VERY WELL supported in Linux.  ATI is very open source
>> friendly.  Matrox is friendly to open source as well.
>
>I don't see anything on Matrox' site apart from note telling people to
>check out commercial X implementations as well as XF86... ATi had a devrel
>programme -- kudos to them for that.

By your logic, there is no video card vendor in existance that
cares about Linux whatsoever merely because their website isn't a
Linux portal.  I disagree completely.  Linux support comes in
many forms, and IMHO the most important form is giving developers
access to programming specifications, and perhaps hardware
samples, info, etc.  *That* is the type of support I am refering
to, not 1-800-mycarddon'twork support.


>> If high end 3D is what you want, get the ATI Radeon, it smokes,
>> although hardware 3D is not available in XFree86 4.0.3 for this
>> card, it is available in the XFree86 trunk code which will become
>> XFree86-4.1.0 in a few months.  I'll be making alpha packages
>> available before long, so it is a damn good buy.  Trond is
>> running Radeon 3D right now.
>
>Benchmarks! Compare a GeForce 2 or 3 to the Radeon. I'd be interested to
>see the results.

Benchmarks mean squat.  Real world software is what means
something.  I will reiterate again also that the fastest hardware
in the world is completely USELESS if it is not well SUPPORTED on
the platform you need/want to use.  *THAT* is my main arguement.


>> Feel free to send a copy of this email to anyone who is
>> investigating buying a new video card for use in Linux.  It could
>> very well save them money, and more importantly frustration.
>
>If you edit the email a bit to correct some facts plus back up your
>assertions with hard data, then yes.

You're entitled to that opinion.  I'm not going to waste my time
going all over the net pulling up facts just to "prove" squat.
I've stated my opinion, based on my experience, and bug reports,
and experience of others.  It is an educational exercise for the
reader to take this information, do their own research, make up
their own mind based on the collective information, and then
purchase video hardware that suits their needs.

Nvidia bugreports > /dev/null

That is unless it is something generic or that can actually be
looked into and fixed without requiring info from nvidia.

Remember - each person has their own requirements.  That may be
"stability", "performance", "support", or one of many other
things.  My recommendations are those primarily of "High quality
support for Linux", and "stability".  All else is second to that.


-- 
Mike A. Harris                  Shipping/mailing address:
OS Systems Engineer             190 Pittsburgh Ave., Sault Ste. Marie,
Red Hat Inc.                    Ontario, Canada, P6C 5B3
http://www.redhat.com           Phone: (705)949-2136






[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]