[ut2003io] radeon

Andrew Pilley ashridah at cyber.com.au
Sat Sep 28 07:12:14 EDT 2002

hrm. just noticed a few misconceptions here. going to clean them up a bit.

On Fri, Sep 27, 2002 at 10:21:39PM -0700, Peter Jay Salzman wrote:
> i guess i wasn't being fair.  i kind of wanted to hear something along
> the lines of:
>    "oh, yeah.  rob is working on it right now.  he just started new
>    classes this september so it's it's slow going, but he said he
>    should get it done by the end of the semester".
> or something like that.  :)

things cost money. fact of life. the weather channel is supporting
development because it's in their best interests. question: what are
weather channel's best interests? good question. tell me they're YOUR 
games, and i'll laugh at you. (no offence :) )

as for ATI, their binary drivers are coming along, or so people say.
you can't just release a driver after a few hours of hacking tho. most of
the time time, QA has to play with it, and marketing has to approve it.
give it a few weeks at least.

> > 	I'm guessing that you really like Nvidia cards?
> well... yes and no.  binary only drivers suck, but it certainly wouldn't
> stop me from getting a kick ass video card.  i'm a ... an open source quasi
> fanatic.  a zealot who can still drink a big fat cup of realism.

see below.

> what i object to most is using hardware which can be pulled from
> underneath me.  suppose X 5.0 is released tomorrow and debian puts it in
> right away, and nvidia decides not to release new drivers.  or for some
> reason the drivers get delayed.

a) X 5 isn't going to stop you from using X 4. xlib DIDN'T CHANGE AT ALL
between 3.3.6 and 4 (so far as i know). they just changed the backend.
XFree86 5 is miles away anyway. beyond miles at that.

> or on the kernel side.  suppose modversioning or the device driver
> interface changes in some fundamental way that renders the kernel driver
> inoperable.  and nvidia doesn't want to release a new driver until the
> module interface has been settled.

this is an incorrect assumption. back when 2.4-test's were going around,
i wrote patches for the wrapper source code for the nvidia drivers a few
times myself. it's REALLY not that hard, and only takes a few minutes of
searching. usually, if you can't do it, someone else probably has.
just don't ask for support for 2.5.x, i'll question you about how much
you like your data if you ask me. :)

> anything can happen.  at least if the stuff was open sourced, i'd have a
> ghost of a chance to fix the code myself.  i've done some kernel
> hacking.  not the world's best, but i've done a little.

the wrapper code is open enough. nvidia don't care if we provide mini-patches
and XFree86 4 aren't changing their binary module system any time soon.
they wrote it to support binary-only drivers because they realised they'd
get more drivers from manufacturers that way. it's only when people start
doing funky things like compiling X with gcc 3.2 that you start to run
into minor issues (which are fixable with a little hacking around)

> case in point:
> way back when, when 3dfx ruled the world, a new kernel version came out
> that had changed the layout of the file operations table.  entries were
> moved around for some reason.  this happened in a stable kernel.  a few
> drivers were affected.  the 3dfx kernel module broke.  i was able to fix
> it.  in fact, it was easy.

see above. i've done that myself with nvidia's wrappers.

> prime example of why i'd rather have open source drivers.  not because
> they're better (in fact, everyone agrees the radeon drivers are pretty
> below standard compared to the windows radeon drivers), but because i
> like being in control.  i don't like wondering if my video card will
> still work if i need to upgrade to kernel 2.5.whateverthey'reupto.

if you need 2.5.x, you need a decent backup plan. and see comments about
wrappers. that problem is fixable (there are patches around for 2.5.x)
and what's wrong with the console? :)

> my main workstation has a voodoo 5 in it.  talk about hardware getting
> pulled from underneath.  i wonder how many other linux voodoo 5 users
> are still waiting for the 2nd GPU to be used...  ;)

i wonder how many of them are expecting the voodoo 5 to play doom 3 at
all? increasingly, they're going to be forgotten by game developers.

> > 	Yes, beware: the XiG drivers do cost actual money.  Be very
> > afraid.  Run! Run away! Buy Nvidia! ;-)
> i think i will.  i built a machine specifically for gaming.  i'll stick
> an nvidia in it.  guess i should've done it by now.

heh. yeah. for the cost of that Xserver, you'd probably want a good reason
to spend it. i doubt gaming is it, when you can buy a fast card with the
same money.

> i just *HATE* the research that comes along with buying new hardware.
> video cards and motherboards are the hardest things to compare... :)
> maybe i'll get one tomorrow.  i've set a few nvidia cards up at our
> installfests.   what are they up to?  the geforce4?

hardware hunting is an art. knowing who to believe and who to ask is
fun, but not something i want to do regularly. thank god i have friends
who like playing with hardware. :)


> pete
> -- 
> Fingerprint: B9F1 6CF3 47C4 7CD8 D33E 70A9 A3B9 1945 67EA 951D

Andrew Pilley  -   Cybersource Pty Ltd, Lvl 9, 140 Queen St. Melbourne
ashridah at cyber.com.au        Ph. 9642-5997 www.cyber.com.au

More information about the ut2003 mailing list