I'm currently putting together (ordering parts for) another machine. It
will be a AMD64 X2. Now I'm looking into a video card for this. Up till
now, I've always used NVidia. But I also want to test 3D acceleration
under Ingo's -rt patch. So now I need something that does not have a
priority module.
I'm not much of a gamer, although I do play every so often. So I don't
need the highest quality card, but I also want something that is still
pretty good. For example, I currently have a NVidia GeForce 6800 GT
card. So I'm hoping to get something equivalent.
I'm looking at the ATI Radeons.
Any recommendations? (links to info would also be nice ;-)
Thanks,
-- Steve
On 11/4/05, Steven Rostedt <[email protected]> wrote:
> I'm currently putting together (ordering parts for) another machine. It
> will be a AMD64 X2. Now I'm looking into a video card for this. Up till
> now, I've always used NVidia. But I also want to test 3D acceleration
> under Ingo's -rt patch. So now I need something that does not have a
> priority module.
>
> I'm not much of a gamer, although I do play every so often. So I don't
> need the highest quality card, but I also want something that is still
> pretty good. For example, I currently have a NVidia GeForce 6800 GT
> card. So I'm hoping to get something equivalent.
>
> I'm looking at the ATI Radeons.
>
> Any recommendations? (links to info would also be nice ;-)
>
> Thanks,
>
> -- Steve
Not a recommendation. Just a point to be aware of. The ATI Radeons, to
get the best acceleration, seem to require that you use the ATI closed
source drivers. Currently I haven't found an ATI closed source driver
that supports 2.6.14. so I'm forced to use the Xorg radeon driver. I
have no idea if this is very good. I don't think so as my glxgear
numbers are pretty low. Much lower than the ATI driver running on
2.6.13-X.
NOTE: I'm horrible at setting up all the 3D stuff as I don't
understand what to chose and what to avoid. This could be far better
than the results I get!
- Mark
Dnia piątek, 4 listopada 2005 16:10, Mark Knecht napisał:
> (...)
> Currently I haven't found an ATI closed source driver
> that supports 2.6.14. so I'm forced to use the Xorg radeon driver.
This patch works fine on my amd64+radeon.
http://ati.cchtml.com/show_bug.cgi?id=211
--
The only thing necessary for the triumph of evil
is for good men to do nothing.
- Edmund Burke
On Fri, 2005-11-04 at 07:10 -0800, Mark Knecht wrote:
> On 11/4/05, Steven Rostedt <[email protected]> wrote:
> > I'm looking at the ATI Radeons.
> >
> > Any recommendations? (links to info would also be nice ;-)
> >
> > Thanks,
> >
> > -- Steve
>
> Not a recommendation. Just a point to be aware of. The ATI Radeons, to
> get the best acceleration, seem to require that you use the ATI closed
> source drivers. Currently I haven't found an ATI closed source driver
> that supports 2.6.14. so I'm forced to use the Xorg radeon driver. I
> have no idea if this is very good. I don't think so as my glxgear
> numbers are pretty low. Much lower than the ATI driver running on
> 2.6.13-X.
I'm sure ATI will come out with a proprietary driver for 2.6.14. I just
want some 3D accel for working with -rt. If I want to play a game, I'll
just back out to 2.6.13 and use the ATI binary driver. But right now
with NVidia, it is either -rt with no accel, or accel with no -rt.
Thanks,
-- Steve
On 4 Nov 2005, Mark Knecht announced authoritatively:
> On 11/4/05, Steven Rostedt <[email protected]> wrote:
>> I'm currently putting together (ordering parts for) another machine. It
>> will be a AMD64 X2. Now I'm looking into a video card for this. Up till
>> now, I've always used NVidia. But I also want to test 3D acceleration
>> under Ingo's -rt patch. So now I need something that does not have a
>> priority module.
>>
>> I'm not much of a gamer, although I do play every so often. So I don't
>> need the highest quality card, but I also want something that is still
>> pretty good. For example, I currently have a NVidia GeForce 6800 GT
>> card. So I'm hoping to get something equivalent.
>>
>> I'm looking at the ATI Radeons.
>>
>> Any recommendations? (links to info would also be nice ;-)
>
> Not a recommendation. Just a point to be aware of. The ATI Radeons, to
> get the best acceleration, seem to require that you use the ATI closed
> source drivers. Currently I haven't found an ATI closed source driver
> that supports 2.6.14. so I'm forced to use the Xorg radeon driver. I
> have no idea if this is very good. I don't think so as my glxgear
> numbers are pretty low. Much lower than the ATI driver running on
> 2.6.13-X.
Surely it depends on the Radeon. I get perfectly respectable numbers on
my Radeon 9250 (purchased because Dave Airlie said it was well-supported
by X.org :) ). It's not exactly Doom-playing quality, but it's good
enough for everything I'm doing with it, and it only cost thirty quid:
for that price you can afford to try it and see :)
(Later Radeons' 3D aren't fully supported yet in free drivers, as far as
I know.)
--
`Heinlein is quite competent at putting together sentences, but usually
he also puts together a plot to go with them.' --- Russ Allbery
On Fri, 2005-11-04 at 08:56 -0500, Steven Rostedt wrote:
> I'm currently putting together (ordering parts for) another machine. It
> will be a AMD64 X2. Now I'm looking into a video card for this. Up till
> now, I've always used NVidia. But I also want to test 3D acceleration
> under Ingo's -rt patch. So now I need something that does not have a
> priority module.
>
> I'm not much of a gamer, although I do play every so often. So I don't
> need the highest quality card, but I also want something that is still
> pretty good. For example, I currently have a NVidia GeForce 6800 GT
> card. So I'm hoping to get something equivalent.
people who buy a 3D card for linux that depends on a closed source
module take a few risks, and they should be aware of them (I suspect
they are) so let me make some of them explicit:
By buying a piece of hardware that requires a closed module you take the
risk that one of the following can happen at any time
1) The vendor in the future stops considering linux important and you're
stuck with old kernels; for example as a side-effect of getting a good
deal to supply graphics chips to a certain game console maker
2) The vendor in the future stops considering the hardware you bought
important enough to spend time on; after all they got their cash and the
product cycles for consumer hardware are often in the 3 to 6 month
timeframe. Result: you're stuck with old kernels.
3) The vendor gets sued and convicted for GPL violations and stops doing
linux as a result. (not saying it will happen, but it sure is a risk you
are taking)
4) The linux kernel developers change the kernel in a way that the
module in question no longer is possible and the vendor stops updating
the driver
5) The vendor goes out of business and thus stops updating the driver
6) The vendor doesn't release an x86-64 binary (or other architecture)
and your next PC can't use the module anymore
7) The vendor starts charging money for the driver or updates thereof.
Open source is not just something for developers, but also for users. It
means that you or anyone else can keep the open driver going even when
the manufacturer stops doing so. By using a closed driver you get all
the disadvantages of the open source model (yes there are some just that
normally the benefits outweigh them by far) without getting the gains.
Be very sure you want to do this before spending your hard earned money
on hardware that doesn't work without closed drivers.
On Mon, 2005-11-07 at 08:42 +0100, Arjan van de Ven wrote:
> people who buy a 3D card for linux that depends on a closed source
> module take a few risks, and they should be aware of them (I suspect
> they are) so let me make some of them explicit:
Are there good 3D cards that don't depend on a proprietary module, that
can run on a AMD64 board? That was pretty much my questing to begin
with :)
>
> By buying a piece of hardware that requires a closed module you take the
> risk that one of the following can happen at any time
Yep, I know all these, since I've been a NVidia user for some time. But
NVidia was good enough for my needs since the only times I needed 3D was
when I wasn't playing with experimental kernels.
>
> 1) The vendor in the future stops considering linux important and you're
> stuck with old kernels; for example as a side-effect of getting a good
> deal to supply graphics chips to a certain game console maker
I was able to get some hacks out for NVidia on some new kernels before
they were official released. But they were not great, just worked.
> 2) The vendor in the future stops considering the hardware you bought
> important enough to spend time on; after all they got their cash and the
> product cycles for consumer hardware are often in the 3 to 6 month
> timeframe. Result: you're stuck with old kernels.
So far NVidia is good at having one driver to do most of their boards.
It would take a major design change of a model to stop this, and by
then, I would probably have a new video card anyway.
> 3) The vendor gets sued and convicted for GPL violations and stops doing
> linux as a result. (not saying it will happen, but it sure is a risk you
> are taking)
Could happen, but I doubt it. This might happen if one of the above do
first :)
> 4) The linux kernel developers change the kernel in a way that the
> module in question no longer is possible and the vendor stops updating
> the driver
I've also hacked my kernel to get NVidia working. (Changing
EXPORT_SYMBOL_GPL back to EXPORT_SYMBOL) It's ok as long as I'm using
this just for myself. Which currently I am.
> 5) The vendor goes out of business and thus stops updating the driver
MS folks would have the same problem.
> 6) The vendor doesn't release an x86-64 binary (or other architecture)
> and your next PC can't use the module anymore
Hmm, x86-64 _is_ what I'll be using this on :-/
> 7) The vendor starts charging money for the driver or updates thereof.
Good way to lose customers.
>
> Open source is not just something for developers, but also for users. It
> means that you or anyone else can keep the open driver going even when
> the manufacturer stops doing so. By using a closed driver you get all
> the disadvantages of the open source model (yes there are some just that
> normally the benefits outweigh them by far) without getting the gains.
> Be very sure you want to do this before spending your hard earned money
> on hardware that doesn't work without closed drivers.
>
I totally agree with you on this, that's why my question was about a
good "Open Source" 3D card in the first place. I want to try out 3D on
Ingo's RT patch set and NVidia (because of the above that you mentioned)
doesn't cut it anymore. I've heard that the Radeon open source drive
isn't too bad so I went with them. I don't need the best 3D, but I do
need something.
So you are right. I've been a loyal NVidia customer for several years
now, but since there is no alternative of a reliable 3D driver for them,
I had to leave them to do what I needed. Now they risk me never going
back if I find out that I like ATI better.
-- Steve
On Mon, Nov 07, 2005 at 07:42:51AM -0500, Steven Rostedt wrote:
> On Mon, 2005-11-07 at 08:42 +0100, Arjan van de Ven wrote:
>
> > people who buy a 3D card for linux that depends on a closed source
> > module take a few risks, and they should be aware of them (I suspect
> > they are) so let me make some of them explicit:
>
> Are there good 3D cards that don't depend on a proprietary module, that
> can run on a AMD64 board? That was pretty much my questing to begin
> with :)
http://www.xgitech.com/
Not the fastest pieces of hardware out there by some way, but they
_do_ have open-source drivers. Don't know if they work with AMD64,
though.
There's also the OpenGraphics development work. See the summary at
http://kerneltrap.org/node/5743
Hugo.
--
=== Hugo Mills: hugo@... carfax.org.uk | darksatanic.net | lug.org.uk ===
PGP key: 1C335860 from wwwkeys.eu.pgp.net or http://www.carfax.org.uk
--- Anyone using a computer to generate random numbers is, of ---
course, in a state of sin.
On Llu, 2005-11-07 at 07:42 -0500, Steven Rostedt wrote:
> Are there good 3D cards that don't depend on a proprietary module, that
> can run on a AMD64 board? That was pretty much my questing to begin
> with :)
Some of the radeons - R3xx is pretty close to usable R2xx works well.
Support for running 32bit hardware accelerated apps on 64bit kernel
recently went in.
On Mon, 2005-11-07 at 15:24, Alan Cox wrote:
> On Llu, 2005-11-07 at 07:42 -0500, Steven Rostedt wrote:
> > Are there good 3D cards that don't depend on a proprietary module, that
> > can run on a AMD64 board? That was pretty much my questing to begin
> > with :)
>
> Some of the radeons - R3xx is pretty close to usable R2xx works well.
> Support for running 32bit hardware accelerated apps on 64bit kernel
> recently went in.
I'm using a Radeon 9600 (R350) with debian packages (32bits userspace,
64bits kernel) since a few weeks, and it works very well, apart from a
few visual glitches here and there (the only GL apps I use are some free
games, blender and some screensavers). The only thing I had to recompile
was the drm moduleset.
xav
Dnia poniedziałek, 7 listopada 2005 13:55, Hugo Mills napisał:
> http://www.xgitech.com/
>
> Not the fastest pieces of hardware out there by some way, but they
> _do_ have open-source drivers. Don't know if they work with AMD64,
> though.
[ http://www.bit-tech.net/hardware/2005/05/03/club3d_volari_v3xt/2.html ]
System Setup:
AMD Athlon 64 FX-55 (operating at 1600MHz - 8x200); (...)
Volari V3XT seems to work on amd64.
--
The only thing necessary for the triumph of evil
is for good men to do nothing.
- Edmund Burke
Dnia poniedziałek, 7 listopada 2005 15:22, Paweł Sikora napisał:
> Dnia poniedziałek, 7 listopada 2005 13:55, Hugo Mills napisał:
> > http://www.xgitech.com/
> >
> > Not the fastest pieces of hardware out there by some way, but they
> > _do_ have open-source drivers. Don't know if they work with AMD64,
> > though.
>
> [ http://www.bit-tech.net/hardware/2005/05/03/club3d_volari_v3xt/2.html ]
>
> System Setup:
> AMD Athlon 64 FX-55 (operating at 1600MHz - 8x200); (...)
>
> Volari V3XT seems to work on amd64.
ps).
"XGI had released its source-code to the Linux community, but what they
failed to mention in their press release was that the 3D acceleration portion
of their drivers would remain closed-source."
(...)
Although it will be very unlikely for XGI to open up the 3D portion of their
drivers anytime in the near future, due to fierce competition from NVIDIA
and ATI drivers."
--
The only thing necessary for the triumph of evil
is for good men to do nothing.
- Edmund Burke
On Mon, Nov 07, 2005 at 07:42:51AM -0500, Steven Rostedt wrote:
> On Mon, 2005-11-07 at 08:42 +0100, Arjan van de Ven wrote:
>
> > 5) The vendor goes out of business and thus stops updating the driver
>
> MS folks would have the same problem.
...which proves the point Arjan is making.
For one, I have an ISDN-adapter which doesn't work with any version of
MS-Windows from this millennium (no drivers available), while it's still
working great on current Linux kernels.
Regards,
Toon.
On Mon, 2005-11-07 at 16:20 +0100, Toon van der Pas wrote:
> On Mon, Nov 07, 2005 at 07:42:51AM -0500, Steven Rostedt wrote:
> > On Mon, 2005-11-07 at 08:42 +0100, Arjan van de Ven wrote:
> >
> > > 5) The vendor goes out of business and thus stops updating the driver
> >
> > MS folks would have the same problem.
>
> ...which proves the point Arjan is making.
>
> For one, I have an ISDN-adapter which doesn't work with any version of
> MS-Windows from this millennium (no drivers available), while it's still
> working great on current Linux kernels.
well despite your post; the Windows people are a lot better at keeping
old drivers working (win 9x to a NT based kernel was obviously a huge
change though). In linux you can use an old driver maybe for 6 months if
you're lucky.. in windows 6 years is no exception. So the problem is a
lot bigger in linux for the owner of such a card than it is in windows.
On Mon, 7 Nov 2005, Alan Cox wrote:
> On Llu, 2005-11-07 at 07:42 -0500, Steven Rostedt wrote:
> > Are there good 3D cards that don't depend on a proprietary module, that
> > can run on a AMD64 board? That was pretty much my questing to begin
> > with :)
>
> Some of the radeons - R3xx is pretty close to usable R2xx works well.
> Support for running 32bit hardware accelerated apps on 64bit kernel
> recently went in.
>
I would also suggest avoiding PCI-E cards for the time being as
there is no DRI support for them. (although if I'm not mistaken that will
improve in 2.6.15).
Gerhard
--
Gerhard Mack
[email protected]
<>< As a computer I find your faith in technology amusing.
El Mon, 7 Nov 2005 16:20:09 +0100,
Toon van der Pas <[email protected]> escribi?:
> > MS folks would have the same problem.
>
> ...which proves the point Arjan is making.
>
> For one, I have an ISDN-adapter which doesn't work with any version of
> MS-Windows from this millennium (no drivers available), while it's still
> working great on current Linux kernels.
agreed, I have the same problems with a creative (a well know vendor)
webcam. Stopped working on SP2.
The problem in windows (and linux closed-sourced drivers) is _much_ bigger
than it seems. Soon all CPUs will be 64-bit capable and dual core, and that
brings two problems: 64-bit-compatible drivers and SMP-safe drivers. I used
to have a cheap winmodem which would deadlock my smp machine. There're lots
of crappy windows drivers which haven't been even tested in a smp machine
it seems. If you buy a dual-core machine and your vendor has gone out of
bussines or the "support period" is expired you migh need to buy new
hardware :(
The 64-bit driver issue is bigger: Even HP has already a (somewhat long)
list of printers which HP is definitively _never_ going to support in 64 bit
platforms (and that's HP, you can imagine what small vendors are doing).
It seems that most of desktop hardware makers are supporting the 64 bit
windows platform just for new devices - unless you had 64 bit in mind when
you created it (which is not possible, because nobody knew what was going to
happen in the 64 bit desktop land until intel supported x86-64) supporting
64 bit may be very well impossible, since they need to port the drivers
of all the thousands of devices they've made through years, and the windows
platform doesn't make it easier (pointers are obviously 64-bit long in the
windows 64 bit platform but longs are still 32 bit for compatibility and
driver programmers don't find it helpful I've heard)
I wish there was opensource drivers for windows. Maybe the fact that some
companies are opening them in linux will make that easier in the future :/
Hi,
I'd like to thank all those that responded. Unfortunately, I was not
very patient and already bought something before most of you responded.
That's my fault.
So, this is what I'm getting:
ATI (Connect3D) Radeon X800 GTO 256MB GDDR3/PCI-E/VIVO/DVI (Retail Box)
For $179 at Monarch computer.
Gerhard Mack recommended avoiding the PCI-E since there is no support
yet for DRI. Right now I don't care about that (thinking it will be
supported in the future) and all that I read was to go with the PCI-E
instead of the AGP. I could be wrong.
I'll keep the list posted of problems/solutions that I come across with
this card.
Thanks,
-- Steve
On Mon, 2005-11-07 at 16:31 +0100, Arjan van de Ven wrote:
> On Mon, 2005-11-07 at 16:20 +0100, Toon van der Pas wrote:
> > On Mon, Nov 07, 2005 at 07:42:51AM -0500, Steven Rostedt wrote:
> > > On Mon, 2005-11-07 at 08:42 +0100, Arjan van de Ven wrote:
> > >
> > > > 5) The vendor goes out of business and thus stops updating the driver
> > >
> > > MS folks would have the same problem.
> >
> > ...which proves the point Arjan is making.
> >
> > For one, I have an ISDN-adapter which doesn't work with any version of
> > MS-Windows from this millennium (no drivers available), while it's still
> > working great on current Linux kernels.
>
>
> well despite your post; the Windows people are a lot better at keeping
> old drivers working (win 9x to a NT based kernel was obviously a huge
> change though). In linux you can use an old driver maybe for 6 months if
> you're lucky.. in windows 6 years is no exception. So the problem is a
> lot bigger in linux for the owner of such a card than it is in windows.
>
Only if the Linux driver is closed source. Otherwise, the driver should
be upgraded with the kernel. Most all open source hardware drivers are
already included in the kernel, and maintained as long as there's
someone that has the device that can maintain it.
-- Steve
On Mon, 7 Nov 2005, Steven Rostedt wrote:
> Date: Mon, 07 Nov 2005 12:30:24 -0500
> From: Steven Rostedt <[email protected]>
> To: LKML <[email protected]>
> Cc: Gerhard Mack <[email protected]>, Xavier Bestel <[email protected]>,
> Alan Cox <[email protected]>, Hugo Mills <[email protected]>,
> Nix <[email protected]>, Anshuman Gholap <[email protected]>,
> Mark Knecht <[email protected]>, Diego Calleja <[email protected]>,
> Toon van der Pas <[email protected]>, [email protected]
> Subject: Re: 3D video card recommendations
>
> Hi,
>
> I'd like to thank all those that responded. Unfortunately, I was not
> very patient and already bought something before most of you responded.
> That's my fault.
>
> So, this is what I'm getting:
>
> ATI (Connect3D) Radeon X800 GTO 256MB GDDR3/PCI-E/VIVO/DVI (Retail Box)
> For $179 at Monarch computer.
>
> Gerhard Mack recommended avoiding the PCI-E since there is no support
> yet for DRI. Right now I don't care about that (thinking it will be
> supported in the future) and all that I read was to go with the PCI-E
> instead of the AGP. I could be wrong.
>
> I'll keep the list posted of problems/solutions that I come across with
> this card.
>
> Thanks,
>
> -- Steve
Way ahead of you .. I have such a list since I own the PCI-E X300 ;)
Your better off than I am since I bought mine two months ago.
Gerhard
--
Gerhard Mack
[email protected]
<>< As a computer I find your faith in technology amusing.
On Mon, 2005-11-07 at 12:44 -0500, Gerhard Mack wrote:
>
> Way ahead of you .. I have such a list since I own the PCI-E X300 ;)
>
> Your better off than I am since I bought mine two months ago.
>
> Gerhard
Are you running this on a x86_64 machine?
-- Steve
On Mon, 7 Nov 2005, Steven Rostedt wrote:
> On Mon, 2005-11-07 at 12:44 -0500, Gerhard Mack wrote:
>
> >
> > Way ahead of you .. I have such a list since I own the PCI-E X300 ;)
> >
> > Your better off than I am since I bought mine two months ago.
> >
> > Gerhard
>
> Are you running this on a x86_64 machine?
Yes.. it works I just don't get direct rendering.
Gerhard
--
Gerhard Mack
[email protected]
<>< As a computer I find your faith in technology amusing.
On 11/7/05, Steven Rostedt <[email protected]> wrote:
> On Mon, 2005-11-07 at 12:44 -0500, Gerhard Mack wrote:
>
> >
> > Way ahead of you .. I have such a list since I own the PCI-E X300 ;)
> >
> > Your better off than I am since I bought mine two months ago.
> >
> > Gerhard
>
> Are you running this on a x86_64 machine?
>
> -- Steve
Steven,
Hi . I run my ATI PCI_Express card on a 64-bit kernel. (2.6.14-rt6)
It works fine for my needs, although as I said earlier my glxgears
numbers are nothing to shout about:
0000:01:00.0 VGA compatible controller: ATI Technologies Inc RV370
5B60 [Radeon X300 (PCIE)]
0000:01:00.1 Display controller: ATI Technologies Inc RV370 [Radeon X300SE]
mark@lightning ~ $ glxgears
Xlib: extension "XFree86-DRI" missing on display ":0.0".
3170 frames in 5.0 seconds = 634.000 FPS
3416 frames in 5.0 seconds = 683.200 FPS
3294 frames in 5.0 seconds = 658.800 FPS
mark@lightning ~ $
I'm using the radeon driver from the Xorg-X11 package. The only
problem I've run into which remains unsolved is that when I run either
Quicken or IE6 under Crossover Office 5.0 all of the icons in those
windows programs show up in black and white, not color, so they are
somewhat unreadable. Other than that no real problems.
- Mark
On Mon, 2005-11-07 at 10:01 -0800, Mark Knecht wrote:
Mark thanks for the update.
>
> Steven,
> Hi . I run my ATI PCI_Express card on a 64-bit kernel. (2.6.14-rt6)
> It works fine for my needs, although as I said earlier my glxgears
> numbers are nothing to shout about:
>
> 0000:01:00.0 VGA compatible controller: ATI Technologies Inc RV370
> 5B60 [Radeon X300 (PCIE)]
> 0000:01:00.1 Display controller: ATI Technologies Inc RV370 [Radeon X300SE]
>
> mark@lightning ~ $ glxgears
> Xlib: extension "XFree86-DRI" missing on display ":0.0".
> 3170 frames in 5.0 seconds = 634.000 FPS
> 3416 frames in 5.0 seconds = 683.200 FPS
> 3294 frames in 5.0 seconds = 658.800 FPS
These aren't too shabby, but then again compared to my NVidia (non-rt
obviously) with their binary driver):
0000:01:05.0 VGA compatible controller: nVidia Corporation NV40 [GeForce
6800 GT] (rev a1)
$ glxgears
49961 frames in 5.0 seconds = 9992.200 FPS
48599 frames in 5.0 seconds = 9719.800 FPS
55592 frames in 5.0 seconds = 11118.400 FPS
47395 frames in 5.0 seconds = 9479.000 FPS
What do you get with the ATI binary driver?
>
> mark@lightning ~ $
>
> I'm using the radeon driver from the Xorg-X11 package. The only
> problem I've run into which remains unsolved is that when I run either
> Quicken or IE6 under Crossover Office 5.0 all of the icons in those
> windows programs show up in black and white, not color, so they are
> somewhat unreadable. Other than that no real problems.
>
Sorry, I can't help you there. I don't have access to IE or Quicken
(well, it's on my wife's computer, but she won't let me put Linux on
it! :-(
-- Steve
On Mon, 7 Nov 2005, Steven Rostedt wrote:
> On Mon, 2005-11-07 at 08:42 +0100, Arjan van de Ven wrote:
>
>> people who buy a 3D card for linux that depends on a closed source
>> module take a few risks, and they should be aware of them (I suspect
>> they are) so let me make some of them explicit:
>
> Are there good 3D cards that don't depend on a proprietary module, that
> can run on a AMD64 board? That was pretty much my questing to begin
> with :)
radeon 9200/9250 is decent though not exceptional (quality is fine,
opengl performance is about 2 years behind state of the art). and uses the
opensource x.org radeon opengl module.
>>
>> By buying a piece of hardware that requires a closed module you take the
>> risk that one of the following can happen at any time
>
> Yep, I know all these, since I've been a NVidia user for some time. But
> NVidia was good enough for my needs since the only times I needed 3D was
> when I wasn't playing with experimental kernels.
>
>>
>> 1) The vendor in the future stops considering linux important and you're
>> stuck with old kernels; for example as a side-effect of getting a good
>> deal to supply graphics chips to a certain game console maker
>
> I was able to get some hacks out for NVidia on some new kernels before
> they were official released. But they were not great, just worked.
>
>> 2) The vendor in the future stops considering the hardware you bought
>> important enough to spend time on; after all they got their cash and the
>> product cycles for consumer hardware are often in the 3 to 6 month
>> timeframe. Result: you're stuck with old kernels.
>
> So far NVidia is good at having one driver to do most of their boards.
> It would take a major design change of a model to stop this, and by
> then, I would probably have a new video card anyway.
>
>> 3) The vendor gets sued and convicted for GPL violations and stops doing
>> linux as a result. (not saying it will happen, but it sure is a risk you
>> are taking)
>
> Could happen, but I doubt it. This might happen if one of the above do
> first :)
>
>> 4) The linux kernel developers change the kernel in a way that the
>> module in question no longer is possible and the vendor stops updating
>> the driver
>
> I've also hacked my kernel to get NVidia working. (Changing
> EXPORT_SYMBOL_GPL back to EXPORT_SYMBOL) It's ok as long as I'm using
> this just for myself. Which currently I am.
>
>> 5) The vendor goes out of business and thus stops updating the driver
>
> MS folks would have the same problem.
>
>> 6) The vendor doesn't release an x86-64 binary (or other architecture)
>> and your next PC can't use the module anymore
>
> Hmm, x86-64 _is_ what I'll be using this on :-/
>
>> 7) The vendor starts charging money for the driver or updates thereof.
>
> Good way to lose customers.
>
>>
>> Open source is not just something for developers, but also for users. It
>> means that you or anyone else can keep the open driver going even when
>> the manufacturer stops doing so. By using a closed driver you get all
>> the disadvantages of the open source model (yes there are some just that
>> normally the benefits outweigh them by far) without getting the gains.
>> Be very sure you want to do this before spending your hard earned money
>> on hardware that doesn't work without closed drivers.
>>
>
> I totally agree with you on this, that's why my question was about a
> good "Open Source" 3D card in the first place. I want to try out 3D on
> Ingo's RT patch set and NVidia (because of the above that you mentioned)
> doesn't cut it anymore. I've heard that the Radeon open source drive
> isn't too bad so I went with them. I don't need the best 3D, but I do
> need something.
>
> So you are right. I've been a loyal NVidia customer for several years
> now, but since there is no alternative of a reliable 3D driver for them,
> I had to leave them to do what I needed. Now they risk me never going
> back if I find out that I like ATI better.
>
> -- Steve
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
--
--------------------------------------------------------------------------
Joel Jaeggli Unix Consulting [email protected]
GPG Key Fingerprint: 5C6E 0104 BAF0 40B0 5BD3 C38B F000 35AB B67F 56B2
On 11/7/05, Steven Rostedt <[email protected]> wrote:
> On Mon, 2005-11-07 at 10:01 -0800, Mark Knecht wrote:
>
> Mark thanks for the update.
>
> >
> > Steven,
> > Hi . I run my ATI PCI_Express card on a 64-bit kernel. (2.6.14-rt6)
> > It works fine for my needs, although as I said earlier my glxgears
> > numbers are nothing to shout about:
> >
> > 0000:01:00.0 VGA compatible controller: ATI Technologies Inc RV370
> > 5B60 [Radeon X300 (PCIE)]
> > 0000:01:00.1 Display controller: ATI Technologies Inc RV370 [Radeon X300SE]
> >
> > mark@lightning ~ $ glxgears
> > Xlib: extension "XFree86-DRI" missing on display ":0.0".
> > 3170 frames in 5.0 seconds = 634.000 FPS
> > 3416 frames in 5.0 seconds = 683.200 FPS
> > 3294 frames in 5.0 seconds = 658.800 FPS
>
> These aren't too shabby, but then again compared to my NVidia (non-rt
> obviously) with their binary driver):
>
> 0000:01:05.0 VGA compatible controller: nVidia Corporation NV40 [GeForce
> 6800 GT] (rev a1)
>
> $ glxgears
> 49961 frames in 5.0 seconds = 9992.200 FPS
> 48599 frames in 5.0 seconds = 9719.800 FPS
> 55592 frames in 5.0 seconds = 11118.400 FPS
> 47395 frames in 5.0 seconds = 9479.000 FPS
>
> What do you get with the ATI binary driver?
Last time I tried (circa 2.6.14-rc3-rtX) it didn't load correctly.
Note that this was the ati-drivers package from portage and not an ATI
driver package from ATI so possibly that was part of the problem.
I should probably try again one of these days but high speed graphics
are not an issue for me on this machine. It's used for audio work
mostly.
Cheers,
Mark
On Mon, 7 Nov 2005, Mark Knecht wrote:
> Date: Mon, 7 Nov 2005 10:31:35 -0800
> From: Mark Knecht <[email protected]>
> To: Steven Rostedt <[email protected]>
> Cc: Gerhard Mack <[email protected]>, LKML <[email protected]>,
> Xavier Bestel <[email protected]>,
> Alan Cox <[email protected]>, Hugo Mills <[email protected]>,
> Nix <[email protected]>, Anshuman Gholap <[email protected]>,
> Diego Calleja <[email protected]>,
> Toon van der Pas <[email protected]>, [email protected]
> Subject: Re: 3D video card recommendations
>
> On 11/7/05, Steven Rostedt <[email protected]> wrote:
> > On Mon, 2005-11-07 at 10:01 -0800, Mark Knecht wrote:
> >
> > Mark thanks for the update.
> >
> > >
> > > Steven,
> > > Hi . I run my ATI PCI_Express card on a 64-bit kernel. (2.6.14-rt6)
> > > It works fine for my needs, although as I said earlier my glxgears
> > > numbers are nothing to shout about:
> > >
> > > 0000:01:00.0 VGA compatible controller: ATI Technologies Inc RV370
> > > 5B60 [Radeon X300 (PCIE)]
> > > 0000:01:00.1 Display controller: ATI Technologies Inc RV370 [Radeon X300SE]
> > >
> > > mark@lightning ~ $ glxgears
> > > Xlib: extension "XFree86-DRI" missing on display ":0.0".
> > > 3170 frames in 5.0 seconds = 634.000 FPS
> > > 3416 frames in 5.0 seconds = 683.200 FPS
> > > 3294 frames in 5.0 seconds = 658.800 FPS
> >
> > These aren't too shabby, but then again compared to my NVidia (non-rt
> > obviously) with their binary driver):
> >
> > 0000:01:05.0 VGA compatible controller: nVidia Corporation NV40 [GeForce
> > 6800 GT] (rev a1)
> >
> > $ glxgears
> > 49961 frames in 5.0 seconds = 9992.200 FPS
> > 48599 frames in 5.0 seconds = 9719.800 FPS
> > 55592 frames in 5.0 seconds = 11118.400 FPS
> > 47395 frames in 5.0 seconds = 9479.000 FPS
> >
> > What do you get with the ATI binary driver?
>
> Last time I tried (circa 2.6.14-rc3-rtX) it didn't load correctly.
> Note that this was the ati-drivers package from portage and not an ATI
> driver package from ATI so possibly that was part of the problem.
>
> I should probably try again one of these days but high speed graphics
> are not an issue for me on this machine. It's used for audio work
> mostly.
>
> Cheers,
> Mark
Not worth trying. I spent a couple weeks with ATI tech support trying to
get it working with no luck. It can work with a 32 bit kernel with some
patching but won't work at all with a 64 bit kernel.
Gerhard
--
Gerhard Mack
[email protected]
<>< As a computer I find your faith in technology amusing.
On 11/7/05, Gerhard Mack <[email protected]> wrote:
> On Mon, 7 Nov 2005, Mark Knecht wrote:
>
> > Date: Mon, 7 Nov 2005 10:31:35 -0800
> > From: Mark Knecht <[email protected]>
> > To: Steven Rostedt <[email protected]>
> > Cc: Gerhard Mack <[email protected]>, LKML <[email protected]>,
> > Xavier Bestel <[email protected]>,
> > Alan Cox <[email protected]>, Hugo Mills <[email protected]>,
> > Nix <[email protected]>, Anshuman Gholap <[email protected]>,
> > Diego Calleja <[email protected]>,
> > Toon van der Pas <[email protected]>, [email protected]
> > Subject: Re: 3D video card recommendations
> >
> > On 11/7/05, Steven Rostedt <[email protected]> wrote:
> > > On Mon, 2005-11-07 at 10:01 -0800, Mark Knecht wrote:
> > >
> > > Mark thanks for the update.
> > >
> > > >
> > > > Steven,
> > > > Hi . I run my ATI PCI_Express card on a 64-bit kernel. (2.6.14-rt6)
> > > > It works fine for my needs, although as I said earlier my glxgears
> > > > numbers are nothing to shout about:
> > > >
> > > > 0000:01:00.0 VGA compatible controller: ATI Technologies Inc RV370
> > > > 5B60 [Radeon X300 (PCIE)]
> > > > 0000:01:00.1 Display controller: ATI Technologies Inc RV370 [Radeon X300SE]
> > > >
> > > > mark@lightning ~ $ glxgears
> > > > Xlib: extension "XFree86-DRI" missing on display ":0.0".
> > > > 3170 frames in 5.0 seconds = 634.000 FPS
> > > > 3416 frames in 5.0 seconds = 683.200 FPS
> > > > 3294 frames in 5.0 seconds = 658.800 FPS
> > >
> > > These aren't too shabby, but then again compared to my NVidia (non-rt
> > > obviously) with their binary driver):
> > >
> > > 0000:01:05.0 VGA compatible controller: nVidia Corporation NV40 [GeForce
> > > 6800 GT] (rev a1)
> > >
> > > $ glxgears
> > > 49961 frames in 5.0 seconds = 9992.200 FPS
> > > 48599 frames in 5.0 seconds = 9719.800 FPS
> > > 55592 frames in 5.0 seconds = 11118.400 FPS
> > > 47395 frames in 5.0 seconds = 9479.000 FPS
> > >
> > > What do you get with the ATI binary driver?
> >
> > Last time I tried (circa 2.6.14-rc3-rtX) it didn't load correctly.
> > Note that this was the ati-drivers package from portage and not an ATI
> > driver package from ATI so possibly that was part of the problem.
> >
> > I should probably try again one of these days but high speed graphics
> > are not an issue for me on this machine. It's used for audio work
> > mostly.
> >
> > Cheers,
> > Mark
>
> Not worth trying. I spent a couple weeks with ATI tech support trying to
> get it working with no luck. It can work with a 32 bit kernel with some
> patching but won't work at all with a 64 bit kernel.
>
> Gerhard
Thanks. I was just looking at emerge output and wondering if I wanted
to mess with this. I didn't, but now I have a good reason!
Thanks,
Mark
On Mon, 2005-11-07 at 10:01 -0800, Mark Knecht wrote:
> I'm using the radeon driver from the Xorg-X11 package. The only
> problem I've run into which remains unsolved is that when I run either
> Quicken or IE6 under Crossover Office 5.0 all of the icons in those
> windows programs show up in black and white, not color, so they are
> somewhat unreadable. Other than that no real problems.
>
Um, didn't you say you were still getting audio underruns correlated
with display activity? I still think it's a bug in the Xorg radeon
driver.
Lee
On 11/7/05, Lee Revell <[email protected]> wrote:
> On Mon, 2005-11-07 at 10:01 -0800, Mark Knecht wrote:
> > I'm using the radeon driver from the Xorg-X11 package. The only
> > problem I've run into which remains unsolved is that when I run either
> > Quicken or IE6 under Crossover Office 5.0 all of the icons in those
> > windows programs show up in black and white, not color, so they are
> > somewhat unreadable. Other than that no real problems.
> >
>
> Um, didn't you say you were still getting audio underruns correlated
> with display activity? I still think it's a bug in the Xorg radeon
> driver.
>
> Lee
Hi Lee,
It very well could be a video issue causing my xruns. That's been
the biggest, but not olny, factor, but I don't know any more. I've
gone down so many paths that haven't yielded results.
I've tried the NoAccel and RenderAccel experiments you suggested.
You also mentioned not loading DRI which I guess isn't possible anyway
with a PCI-Express card. Anyway, I don't have it loaded.
While I do get xruns, they are happening lately only a couple of
times a day, and usually very late in the day after the machine has
been running for 12+ hours. If I pull Jack back to 256/2 then I don't
think I get them at all so I'm living with it for now and hoping for a
solution one of these days. Low latency really only matters to me when
recording which is <5% of the time.
I'm most interested in a kernel developer finding the time to fix
the IRQoff latency testing on 64-bit machines so I can look at that.
In case the question comes to mind for others, this HDSP 9652 sound
card runs for months at a time at 64/2 in my older Via 32-bit machine.
The problem only surfaced when I moved it to my new AMD64 machine.
Anyway, that's status. It's a frustration, not a killer issue.
Cheers,
Mark
On Mon, 2005-11-07 at 11:23 -0800, Mark Knecht wrote:
>
> I'm most interested in a kernel developer finding the time to fix
> the IRQoff latency testing on 64-bit machines so I can look at that.
Well, when I get all my pieces and then the box up and running, I might
just be that developer ;-)
-- Steve
Steven Rostedt wrote:
> On Mon, 2005-11-07 at 08:42 +0100, Arjan van de Ven wrote:
>> 2) The vendor in the future stops considering the hardware you bought
>> important enough to spend time on; after all they got their cash and the
>> product cycles for consumer hardware are often in the 3 to 6 month
>> timeframe. Result: you're stuck with old kernels.
>
> So far NVidia is good at having one driver to do most of their boards.
> It would take a major design change of a model to stop this, and by
> then, I would probably have a new video card anyway.
they just dropped support for the TNT2 (and old Geforce IIRC) boards
some months ago, so my son needed a new video card although the old one
was perfectly good.
--
seife
Never trust a computer you can't lift.
On Tue, Nov 08, 2005 at 07:59:58PM +0100, Stefan Seyfried wrote:
> they just dropped support for the TNT2 (and old Geforce IIRC) boards
> some months ago, so my son needed a new video card although the old one
> was perfectly good.
Not that the 71.x drivers don't still work for old cards. My TNT2 is
doing just fine at the moment.
They are also planing (they claim) to offer packages for the older cards
(at least I read someone was planning to continue offering packages for
the older driver series.)
Of course I might have problems when I try using 2.6.14, although 2.6.12
has no problems.
Len Sorensen
On Mon, 07 Nov 2005, Steven Rostedt wrote:
> On Mon, 2005-11-07 at 16:31 +0100, Arjan van de Ven wrote:
> > well despite your post; the Windows people are a lot better at keeping
> > old drivers working (win 9x to a NT based kernel was obviously a huge
> > change though). In linux you can use an old driver maybe for 6 months if
> > you're lucky.. in windows 6 years is no exception. So the problem is a
> > lot bigger in linux for the owner of such a card than it is in windows.
>
> Only if the Linux driver is closed source. Otherwise, the driver should
> be upgraded with the kernel. Most all open source hardware drivers are
> already included in the kernel, and maintained as long as there's
> someone that has the device that can maintain it.
I'd rather not count the drivers that have dropped out of open source
operating systems due to bit rot. If there is no maintainer, the
hardware will become useless sooner or later. With Linux's rapidly
changing "moving target" 2.6.X I'd call it sooner rather than later.
OSS drivers are good iff there is a maintainer - IOW: to the user, the
maintainer makes the difference, not the driver being open source.
--
Matthias Andree
>>>>> "Matthias" == Matthias Andree <[email protected]> writes:
Matthias> I'd rather not count the drivers that have dropped out of
Matthias> open source operating systems due to bit rot. If there is no
Matthias> maintainer, the hardware will become useless sooner or
Matthias> later. With Linux's rapidly changing "moving target" 2.6.X
Matthias> I'd call it sooner rather than later.
Matthias> OSS drivers are good iff there is a maintainer - IOW: to the
Matthias> user, the maintainer makes the difference, not the driver
Matthias> being open source.
No, a publically available spec for the hardware is what makes the
difference. Those drivers which are reverse engineered, or which have
only partially open specs are the ones which bit-rot the fastest.
For example, I've got a webcam which I've never been able to make work
properly due to the PWC driver and the lack of documentation from
Philips on the chipset. Even with heroic efforts by others, it's just
hard to do.
But if we have docs, we could support this camera until the end of
time, as long as people were interested in supporting it.
John
El Tue, 8 Nov 2005 23:08:53 +0100,
Matthias Andree <[email protected]> escribi?:
> I'd rather not count the drivers that have dropped out of open source
> operating systems due to bit rot. If there is no maintainer, the
> hardware will become useless sooner or later. With Linux's rapidly
> changing "moving target" 2.6.X I'd call it sooner rather than later.
>
> OSS drivers are good iff there is a maintainer - IOW: to the user, the
> maintainer makes the difference, not the driver being open source.
You're right - but the existence of a unmaintained open source
driver makes possible and easier to write a maintained driver. Which
is important, too
On Maw, 2005-11-08 at 23:08 +0100, Matthias Andree wrote:
> I'd rather not count the drivers that have dropped out of open source
> operating systems due to bit rot. If there is no maintainer, the
> hardware will become useless sooner or later. With Linux's rapidly
> changing "moving target" 2.6.X I'd call it sooner rather than later
If there is no maintainer and there is demand for the driver then I
guess the end users need to remember its a development model and not a
charity. The problem should then be self correcting if the docs/driver
are open
> OSS drivers are good iff there is a maintainer - IOW: to the user, the
> maintainer makes the difference, not the driver being open source.
No. If the code is open source the user can get it fixed, if it is
closed source then they are screwed because support is by vendor dictat.
If Nvidia drop TNT2 you can't pay a local consultant to fix it.
Alan
On Mon, 2005-11-07 at 11:23 -0800, Mark Knecht wrote:
> On 11/7/05, Lee Revell <[email protected]> wrote:
> >
> > Um, didn't you say you were still getting audio underruns correlated
> > with display activity? I still think it's a bug in the Xorg radeon
> > driver.
> >
> > Lee
>
> Hi Lee,
> It very well could be a video issue causing my xruns. That's been
> the biggest, but not olny, factor, but I don't know any more. I've
> gone down so many paths that haven't yielded results.
>
> I've tried the NoAccel and RenderAccel experiments you suggested.
> You also mentioned not loading DRI which I guess isn't possible anyway
> with a PCI-Express card. Anyway, I don't have it loaded.
>
> While I do get xruns, they are happening lately only a couple of
> times a day, and usually very late in the day after the machine has
> been running for 12+ hours. If I pull Jack back to 256/2 then I don't
> think I get them at all so I'm living with it for now and hoping for a
> solution one of these days. Low latency really only matters to me when
> recording which is <5% of the time.
After looking at the Radeon driver I'm increasingly convinced that it's
causing your problems.
Please try disabling Xscreensaver and see if the xruns go away - looking
at radeonfb it seems like the xruns could be triggered by video mode
switching, there are lots of places where it loops waiting for the
hardware to be ready. If that eliminates them I have a patch you can
try.
Lee