One of my SuSE 9.0 boxes has a "Powered By ATI" card (Sapphire Atlantis Radeon 9000 Pro, 64MB DDR RAM). I've been trying unsuccessfully for months to get the dual-head and 3D modes working on it. I'd like to get the TV-out functioning as well, but that's not too much of a concern. All of these features work flawlessly under "that other operating system", so the card is definitely not faulty. Kernel compatibility doesn't seem to be an issue either because I've been compiling the ATI driver module for my 2.4.x kernels myself and it always loads fine. The versions of Xfree86 I've been trying it with don't seem to be an issue either; I've used the 4.2 and 4.3 series. I got my hopes up when ATI released a new version of their Radeon driver for Linux yesterday but after wrestling with it to get dual-head for 45 minutes I still can't get any further than before, which is: machine enters run level 5, X tries to start, both screens go black for a second, then the X server dies and complains, "Fatal server error: no screens found". ATI's "documentation" is vague, but seems to allude to needing 2 instances of X running in order for dual-head mode to work. It would be nice if they gave instructions on how to actually do that. So I continue to use it in single display mode. At least that works fine. Does anyone have an 8500 or later series Radeon working in dual-head mode with 3D? If so, please let me in on your secret, send me your Xfree86Config file, or at least point me to some hard info on the subject. What little info I found on ATI's website is useless. I've also tried SuSE's support database, but the information I found there was badly outdated (i.e., dual-head mode requires 2 cards...). Thanks in advance, The deranged one
R.U. Deranged wrote:
One of my SuSE 9.0 boxes has a "Powered By ATI" card (Sapphire Atlantis Radeon 9000 Pro, 64MB DDR RAM).
I recently picked up an ATI boxed Radeon 9000/64mb, seems we have similar cards.
I got my hopes up when ATI released a new version of their Radeon driver for Linux yesterday but after wrestling with it to get dual-head for 45 minutes I still can't get any further than before, which is: machine enters run level 5, X tries to start, both screens go black for a second, then the X server dies and complains, "Fatal server error: no screens found". ATI's "documentation" is vague, but seems to allude to needing 2 instances of X running in order for dual-head mode to work.
The one thing I noticed was that when using the ATI tool to build the config file, it did not reference both bus ids. Before I purchased this card, I had a Radeon VE which worked with dual head and the config had references to two bus ids. I wonder if their config tool is broken. I attempted to modify the config file by hand to mimic the working one I had for my Radeon VE, but no luck. I've not revisited this of late, but might give it another try noting that you said AT( has released an new driver. I also tried the Gatos drivers but no luck with them either. I don't believe it's necessary to have two servers unless you're planning on running two different instances on each monitor. My attempts are to have a single screen that spans both monitors, is this what you're trying to do, or do you want two separate distinct desktops?
Does anyone have an 8500 or later series Radeon working in dual-head mode with 3D? If so, please let me in on your secret, send me your Xfree86Config file, or at least point me to some hard info on the subject. What little info I found on ATI's website is useless. I've also tried SuSE's support database, but the information I found there was badly outdated (i.e., dual-head mode requires 2 cards...).
Thanks in advance, The deranged one
If anyone does have a working solution, please post to the list. -- Until later, Geoffrey Registered Linux User #108567 Building secure systems inspite of Microsoft
At 05:03 AM 3/4/2004, Geoffrey wrote:
The one thing I noticed was that when using the ATI tool to build the config file, it did not reference both bus ids.
Seems to be the same thing happening here. It can't match a device for PCI:2:0:0.
Before I purchased this card, I had a Radeon VE which worked with dual head and the config had references to two bus ids. I wonder if their config tool is broken. I attempted to modify the config file by hand to mimic the working one I had for my Radeon VE, but no luck. I've not revisited this of late, but might give it another try noting that you said AT( has released an new driver.
Nice coincidence, I have a Radeon VE in another machine. But I wasn't ever able to get it running with dual-head under Linux either. The installation script for the latest Radeon driver is much improved and offers more options than the last version, but during the course of several reconfigurings, it did seem to be generating faulty or incomplete entries in the XF86Config file.
I don't believe it's necessary to have two servers unless you're planning on running two different instances on each monitor. My attempts are to have a single screen that spans both monitors, is this what you're trying to do, or do you want two separate distinct desktops?
When booting into "that other operating system", I usually have my Radeon in extended desktop mode (but sans Hydravision) so I can drag stuff between my two SVGA monitors as if the desktop were displayed on a very wide single monitor. I don't think I'll ever need to use two distinct desktops, and I don't care much about being able to use cloned mode either. So yeah, it appears we're after the same configuration here.
R.U. Deranged wrote:
At 05:03 AM 3/4/2004, Geoffrey wrote:
The one thing I noticed was that when using the ATI tool to build the config file, it did not reference both bus ids.
Seems to be the same thing happening here. It can't match a device for PCI:2:0:0.
What does lspci show? Here's mine: 01:00.0 VGA compatible controller: ATI Technologies Inc: Unknown device 4150 01:00.1 Display controller: ATI Technologies Inc: Unknown device 4170
Before I purchased this card, I had a Radeon VE which worked with dual head and the config had references to two bus ids. I wonder if their config tool is broken. I attempted to modify the config file by hand to mimic the working one I had for my Radeon VE, but no luck. I've not revisited this of late, but might give it another try noting that you said AT( has released an new driver.
Nice coincidence, I have a Radeon VE in another machine. But I wasn't ever able to get it running with dual-head under Linux either.
Hmm, well I can provide my previously working config, but that's not much use with moving forward.
The installation script for the latest Radeon driver is much improved and offers more options than the last version, but during the course of several reconfigurings, it did seem to be generating faulty or incomplete entries in the XF86Config file.
I'll check it out. What I'm going to try and do is use their new tool and then compare it to my previous working config for my VE. I'll let you know if I have any luck.
I don't believe it's necessary to have two servers unless you're planning on running two different instances on each monitor. My attempts are to have a single screen that spans both monitors, is this what you're trying to do, or do you want two separate distinct desktops?
When booting into "that other operating system", I usually have my Radeon in extended desktop mode (but sans Hydravision) so I can drag stuff between my two SVGA monitors as if the desktop were displayed on a very wide single monitor. I don't think I'll ever need to use two distinct desktops, and I don't care much about being able to use cloned mode either. So yeah, it appears we're after the same configuration here.
I'll likely try this today, I'll post my results. -- Until later, Geoffrey Registered Linux User #108567 Building secure systems inspite of Microsoft
At 06:14 AM 3/4/2004, Geoffrey wrote:
What does lspci show? Here's mine:
01:00.0 VGA compatible controller: ATI Technologies Inc: Unknown device 4150 01:00.1 Display controller: ATI Technologies Inc: Unknown device 4170
Here are the relevant entries: 00:00.0 Host bridge: nVidia Corporation nForce2 AGP (different version?) (rev c1) 00:08.0 PCI bridge: nVidia Corporation nForce2 External PCI Bridge (rev a3) 00:1e.0 PCI bridge: nVidia Corporation nForce2 AGP (rev c1) 02:00.0 VGA compatible controller: ATI Technologies Inc Radeon R250 If [Radeon 9000] (rev 01) 02:00.1 Display controller: ATI Technologies Inc Radeon R250 [Radeon 9000] (Secondary) (rev 01) I'm not too versed on bus piping (especially the nForce2 chipset's), but that looks a bit fishy to me.
I'll check it out. What I'm going to try and do is use their new tool and then compare it to my previous working config for my VE. I'll let you know if I have any luck.
Well bear in mind the new script does have one saving grace: it backs up your working XF86Config file before hosing your settings. ;-)
I'll likely try this today, I'll post my results.
Good luck. I'm half-expecting your next email to be posted with Pine. ;-)
At 06:36 AM 3/4/2004, R.U. Deranged wrote:
Well bear in mind the new script does have one saving grace: it backs up your working XF86Config file before hosing your settings. ;-)
I take that back. A quick glance at the headers shows the backups were generated earlier by SaX, not ATI's new script. Be warned. ;-)
On Thu, 04 Mar 2004 06:36:47 -0600
"R.U. Deranged"
Here are the relevant entries:
00:00.0 Host bridge: nVidia Corporation nForce2 AGP (different version?) (rev c1) 00:08.0 PCI bridge: nVidia Corporation nForce2 External PCI Bridge (rev a3) 00:1e.0 PCI bridge: nVidia Corporation nForce2 AGP (rev c1) 02:00.0 VGA compatible controller: ATI Technologies Inc Radeon R250 If [Radeon 9000] (rev 01) 02:00.1 Display controller: ATI Technologies Inc Radeon R250 [Radeon 9000] (Secondary) (rev 01)
I'm not too versed on bus piping (especially the nForce2 chipset's), but that looks a bit fishy to me.
AFAIK the nforce2 supports 8x AGP (AGP V3). If the Radeon supports that as well you'll have to go to kernel 2.6.x for 3D. At least that's what I read and what solved the problem for me... Cheers, Ingo -- "What're quantum mechanics?" "I don't know. People who repair quantums I suppose." --Rincewind, Terry Pratchett "Eric"
At 07:00 AM 3/4/2004, Ingo Strauch wrote:
AFAIK the nforce2 supports 8x AGP (AGP V3). If the Radeon supports that as well you'll have to go to kernel 2.6.x for 3D. At least that's what I read and what solved the problem for me...
My 9000 is indeed 8x. My biggest reservations against moving to the 2.6 kernel at this time are that the ATI drivers apparently don't support it yet, and the nForce2 driver installation documentation says that the nForce2 drivers for Linux only support the 2.4 kernel (which I find hard to believe, but it still makes me nervous). How is your 3D performance under 2.6? And have you ever been able to get dual-head mode working?
On Thu, 04 Mar 2004 07:24:58 -0600
"R.U. Deranged"
At 07:00 AM 3/4/2004, Ingo Strauch wrote:
AFAIK the nforce2 supports 8x AGP (AGP V3). If the Radeon supports that as well you'll have to go to kernel 2.6.x for 3D. At least that's what I read and what solved the problem for me...
My 9000 is indeed 8x. My biggest reservations against moving to the 2.6 kernel at this time are that the ATI drivers apparently don't support it yet,
There is a subdirectory under build_mod named 2.6.x, just type 'make' there. But you need to copy over the resulting module by hand to /lib/modules/...
and the nForce2 driver installation documentation says that the nForce2 drivers for Linux only support the 2.4 kernel (which I find hard to believe, but it still makes me nervous).
Ok, this could be a problem. But doesn't kernel 2.6 contain this forcedeth driver? But I don't have an idea if that's any good.
How is your 3D performance under 2.6?
Well, I never measured it. Currently I'm back to kernel 2.4.x since I don't have sound with the new kernel and my CMedia onboard chip.
And have you ever been able to get dual-head mode working?
Never tried. Cheers, Ingo -- "What're quantum mechanics?" "I don't know. People who repair quantums I suppose." --Rincewind, Terry Pratchett "Eric"
R.U. Deranged wrote:
At 06:14 AM 3/4/2004, Geoffrey wrote:
What does lspci show? Here's mine:
01:00.0 VGA compatible controller: ATI Technologies Inc: Unknown device 4150 01:00.1 Display controller: ATI Technologies Inc: Unknown device 4170
Here are the relevant entries:
00:00.0 Host bridge: nVidia Corporation nForce2 AGP (different version?) (rev c1) 00:08.0 PCI bridge: nVidia Corporation nForce2 External PCI Bridge (rev a3) 00:1e.0 PCI bridge: nVidia Corporation nForce2 AGP (rev c1) 02:00.0 VGA compatible controller: ATI Technologies Inc Radeon R250 If [Radeon 9000] (rev 01) 02:00.1 Display controller: ATI Technologies Inc Radeon R250 [Radeon 9000] (Secondary) (rev 01)
I'm not too versed on bus piping (especially the nForce2 chipset's), but that looks a bit fishy to me.
I don't know, at least your's identifies your card, mine was listed as Unknown Device.
I'll check it out. What I'm going to try and do is use their new tool and then compare it to my previous working config for my VE. I'll let you know if I have any luck.
Well bear in mind the new script does have one saving grace: it backs up your working XF86Config file before hosing your settings. ;-)
Hah, I learned a long time ago, backup before you screwup!
I'll likely try this today, I'll post my results.
Good luck. I'm half-expecting your next email to be posted with Pine. ;-)
mailx. ;) -- Until later, Geoffrey Registered Linux User #108567 Building secure systems inspite of Microsoft
Ok, now I really have no idea what's going on. ATI has removed from their website the new Linux Radeon drivers they released the other day, and as is typical of their lack of customer communication skills, there is no mention of it. (Why don't they offer user support forums like other major hardware vendors?) The other day I downloaded fglrx-4.3.0-3.7.1.i386.rpm from http://www.ati.com/support/drivers/linux/radeon-linux.html?type=linux&prodType=graphic&prod=productsLINUXdriver&submit.x=7&submit.y=7 Now it's nowhere to be found and 3.7.0 is back in its place. So I guess this means we shouldn't use 3.7.1. Should I delete my copy? I'd ask ATI but I wouldn't expect their response, if any, to amount to anything more than a form letter rambling about something only vaguely related to my specific query.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Thursday 04 March 2004 17:35, R.U. Deranged wrote:
Ok, now I really have no idea what's going on. ATI has removed from their website the new Linux Radeon drivers they released the other day, and as is typical of their lack of customer communication skills, there is no mention of it. (Why don't they offer user support forums like other major hardware vendors?)
The other day I downloaded fglrx-4.3.0-3.7.1.i386.rpm from http://www.ati.com/support/drivers/linux/radeon-linux.html?type=linux&prodT ype=graphic&prod=productsLINUXdriver&submit.x=7&submit.y=7
Now it's nowhere to be found and 3.7.0 is back in its place. So I guess this means we shouldn't use 3.7.1. Should I delete my copy? I'd ask ATI but I wouldn't expect their response, if any, to amount to anything more than a form letter rambling about something only vaguely related to my specific query.
I have been lurking this thread with some interest. I have a 9600 XT and am using the fglrx 3.7.0 driver and the km_fglrx 3.7.0 kernel mods. I have gotten it to run but it is flakey to say the least. It is obvious that the Catylist devs seem to have a different idea of how to create a XF86 driver. The problem as I see it is the disconnect between ATI and Catylist. They are not the same company IIRC. Catylist is contracted to build the drivers for ATI. I suspect, but could be wrong, that ATI probably prompted Catylist to build 'nix drivers and they have attempted to comply. But I also supect that they don't have anyone with a lot of prior knowledge or ability to do so as we in the Linux community have become accustomed to. The part that I find baffling is that now OS X 10.3 (aka Panther) opts for XF86 as an alternative to their Quartz engine. If Apple were to go to XF86 exclusive over Quartz then we might see a change. Another "hope" is that if Linux becomes (in their eyes) a viable market force then we again may see some improvement. Either way I would try to find a more direct point of contact to the Catylist developers rather than the default point of contact provided on the ATI website - frankly I don't think the care one way or the other. Perhaps if you or someone else finds this elusive pipeline to the driver/ kernel mod devs we might be able to get a little headway - and if one does find this please post it as well as the content of the any discussions or communications. I know I certainly will if I find myself fortunate enough to stumble upon this. Cheers, Curtis. - -- Spammers Beware: Tresspassers will be shot, survivors will be shot again! Warning: Individuals throwing objects at the crocodiles will be asked to retrieve them! If pro is the opposite of con, then the opposite of progress must be congress! -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.2 (GNU/Linux) iD8DBQFAR/7yN9r/ngHXpykRAj7MAJ9B6PDNeXyjV5Tx9lQ1AvTKMc4CNgCgnhvF 6NwSKH1AFrz6AuJzXlUHtuQ= =ksEM -----END PGP SIGNATURE-----
At 10:15 PM 3/4/2004, Curtis Rey wrote:
It is obvious that the Catylist devs seem to have a different idea of how to create a XF86 driver. The problem as I see it is the disconnect between ATI and Catylist. They are not the same company IIRC. Catylist is contracted to build the drivers for ATI. I suspect, but could be wrong, that ATI probably prompted Catylist to build 'nix drivers and they have attempted to comply. But I also supect that they don't have anyone with a lot of prior knowledge or ability to do so as we in the Linux community have become accustomed to.
This does have all the earmarks of a pass-the-buck scenario. Contractors can claim ATI isn't giving them the resources they need to create Linux drivers, while ATI can maintain plausible deniability, blaming shortcomings on the contractors or XFree86 itself. I can almost remember a time when companies actually took direct responsibility for supporting their own products. Now it's just a finger-pointing game.
The part that I find baffling is that now OS X 10.3 (aka Panther) opts for XF86 as an alternative to their Quartz engine. If Apple were to go to XF86 exclusive over Quartz then we might see a change.
I'd heard about that. Personally I'd rather have Quartz at the heart of my Linux desktop, but that's a different discussion. ;-)
Either way I would try to find a more direct point of contact to the Catylist developers rather than the default point of contact provided on the ATI website - frankly I don't think the care one way or the other.
I agree. I wonder if they realize Nvidia is way ahead of them in Linux support. When/if Linux starts to take significant chunks out of Windows' share of the desktop market over the next couple of years, ATI will likely find themselves in a distant second place if they don't make a more solid Linux commitment soon.
Perhaps if you or someone else finds this elusive pipeline to the driver/ kernel mod devs we might be able to get a little headway - and if one does find this please post it as well as the content of the any discussions or communications. I know I certainly will if I find myself fortunate enough to stumble upon this.
We'll probably need a translator to communicate with the driver devs. I have a feeling they're not Canadian. ;-)
Just want to thank everyone for the helpful suggestions and links so far. I've tried several approaches from the resources I've been given but nothing is working. This could of course be due to incompetence on my part. ;-) But it certainly doesn't help that ATI mysteriously revoked their latest driver release either. So I'm going to try doing it the hard way: recompiling my kernel, X server, the DRI module, and the ATI drivers from scratch with explicit support for each other. Hope I get all the compile options right. I might royally hose my system, but I think I've already done that anyhow. ;-) If that doesn't work, I'm permanently banishing all ATI products to my Win-only boxes and buying Nvidia for my Linux boxes from now on. In fact I recommend against buying ATI for any Linux user currently in the market for a video card; clearly the support just isn't there yet. And I sense that I'm not alone on this. ;-)
FYI, here's recent info on ATI's revoked 3.7.1 drivers: http://www.rage3d.com/board/showthread.php?s=35e30a26d4ac02f18db0e06c8818b58e&threadid=33745504 Also an article on Slashdot: http://slashdot.org/articles/04/03/06/1436223.shtml (The above article was obviously submitted before the drivers were pulled, but Slashdot took a while in posting it. Scroll down a bit for comments pointing this out.) Now I'll be waiting for ATI's "fix" for 3.7.1. But it's the last chance I'm giving them.
R.U. Deranged wrote:
Now I'll be waiting for ATI's "fix" for 3.7.1. But it's the last chance I'm giving them.
I've given up on the ati drivers. I've gone back to the stock Xfree that comes with SuSE 9.0 Pro and successfully configured my dual head, just no 3D. :( Since I couldn't get 3D working with the ATI drivers on a single head, well, you know the saying, two heads are better than one... If you get the ATI drivers working with dual head and 3D please let me know -- Until later, Geoffrey Registered Linux User #108567 Building secure systems inspite of Microsoft
participants (4)
-
Curtis Rey -
Geoffrey -
Ingo Strauch -
R.U. Deranged