Click to See Complete Forum and Search --> : No Hardware Acceleration with R9700


DemosthenesB
06-08-2003, 03:38 PM
A while ago I got my Radeon 9700 drivers to work properly, however, now when I go to play a game (Return to Castle Wolfenstein: Enemy Territory) I find that I have no Hardware acceleration.

When I try to run Wolfenstein it gives me a message:
************************************************** *********
You are using software Mesa (no hardware acceleration)!
Driver DLL used: libGL.so.1
If this is intentional, add
"+set r_allowSoftwareGL 1"
to the command line when starting the game.
************************************************** *********
...WARNING: could not set the given mode (4)
Initializing OpenGL display
...setting mode 3: 640 480
Using XFree86-VidModeExtension Version 2.1
XFree86-VidModeExtension Activated at 1280x1024
Received signal 11, exiting...

I am using Slackware 8.1 and kernel 2.4.20. I got my drivers from the official ati site. Thank you in advance.

DemosthenesB
06-08-2003, 04:25 PM
I have tried to install the ati drivers using the rpm file and I need a lot of dependencies. I search for them and found that a lot of them are made specifically for redhat, and cannot find any source tarballs, is it necessary to intall the rpm?

I had "successfully" installed it before by converting the rpm into tar(rpm2tgz) then moving the appropriate files into /lib/modules, then running the "make.sh" in the build_mod directory, and then running the "make_install.sh." Everything seemed to install fine...

Thanks again.

bwkaz
06-08-2003, 04:27 PM
Well, the RPMs are for RedHat systems. ATI is pretty crappy that way -- they seem to have created a driver that only works on one system. :(

Anyway, your rpm2tgz thing should have worked... is there a libGL.so file in the resulting tarball? If so, put it in /usr/lib (overwrite the one that's there), and rerun /sbin/ldconfig as root.

There might also be a different libGL.so file in the ET directory somewhere. Look around, and if you find it, replace it with the one in /usr/lib.

DemosthenesB
06-08-2003, 05:10 PM
OK, i downloaded the drivers again and converted it to tar.gz. the tarball consists of lib, opt, and usr which I moved to th main directory, overwriting anything that I already had and everything that was included in the drivers.

That should have covered me for the libGL file right?

So then I go to do the installation again and find that it does not work... I can run the "make.sh" file fine in the build_mod dir, but when I run "make_install.sh" it says:
- creating symlink
- recreating module dependency list
depmod: *** Unresolved symbols in /lib/modules/2.4.20/kernel/drivers/char/drm/fglrx.o
failed.

It didn't do this when I first tried it a while ago. And thanks for your help bwkaz.

bwkaz
06-08-2003, 06:18 PM
Yeah, that should cover libGL (it should have been in the usr/lib directory in the tarball, if anywhere).

When you ran the first script, did it use the right kernel source directory?

What happens if you do a depmod -ae -- what symbols are unresolved?

DemosthenesB
06-08-2003, 06:30 PM
What did you mean by the right kernel source directory? All the instructions say is to go into the directory and(/lib/modules/fglrx/buil_mod) and type ./make.sh(or sh make.sh), and it works fine. The output to it is as follows:
ATI module generator V 2.0
==========================
probing for VMA API version...
cleaning...
patching 'highmem.h'...
patching 'drmP.h'...
patching file drmP.h
Hunk #1 succeeded at 283 (offset 28 lines).
compiling 'agpgart_be.c'...
compiling 'agp3.c'...
compiling 'i7505-agp.c'...
compiling 'firegl_public.c'...
linking of fglrx kernel module...
duplicating results into driver repository...
done.
==============================
You must change your working directory to /lib/modules/fglrx
and then call ./make_install.sh in order to install the built module.
==============================
Hope that helps.

Doing a depmod -ae outputs:
depmod: *** Unresolved symbols in /lib/modules/2.4.20/kernel/drivers/char/drm/fglrx.o
depmod: _alloc_pages_R324fe15c
depmod: init_mm_R98a17957
depmod: __free_pages_R1287f72e
depmod: remove_proc_entry_R0f0893c0
depmod: vsprintf_R954cbb26
depmod: __wake_up_Rb76c5f1e
depmod: remove_wait_queue_R833d735e
depmod: register_chrdev_R917f516b
depmod: mem_map_Re76ef071
depmod: zone_table_R7b4be21b
depmod: add_wait_queue_Rc3391385
depmod: unlock_page_R34a48471
depmod: create_proc_entry_R5327b2b9

Thank you very much for your help bwkaz.

DemosthenesB
06-08-2003, 07:38 PM
ok, I got the second command(./make_install.sh) in /lib/modules/fglrx dir to work. I recompiled my kernel, changing my processor type from Pentium-MMX to Pentium-III, why I had it at MMX?... I have no idea.

So I go back and try to reinstall Enemy Territory and when I run it I get the same output(software Mesa and no hardware acceleration).

Thanks again.

DemosthenesB
06-08-2003, 09:16 PM
I read something about having to enable ISSE support on the kernel, do you know exactly where it is using xconfig? Or is it already enabled? Or do I have to apply a patch? thanks again.

bwkaz
06-08-2003, 09:45 PM
SSE support in the kernel depends on which processor you pick. The Pentium-MMX didn't have SSE support, so it wouldn't have that enabled. But the P3 did, so that would.

Is there a libGL in the tarball?

Does anyone else here know much about ATI's drivers?

DemosthenesB
06-08-2003, 10:23 PM
When I go to configure the kernel I did put in the P3 processor family, but there are no exra options to enable or disable SSE support, does it automatically become supported once the P3 is selected? Should I try to download the same kernel(2.4.20) from a different location, move it to the same location as my current kernel(overwriting anything) and try to configure it again.

I found a patch(from http://people.redhat.com/mingo/mmx-patches/), but it is 2.4.14(mmx-2.2.14-A0), and although I have never patched a kernel it looks like it wont work as it is 6 versions behind. There are also some other ones that are 2.3.33 and above, but I read that odd numbers were developement versions and buggy.

There is a libGL.so.1.2 under "usr/X11R6/lib" within the tarball, and I already copied the entire directory to the real /usr. So that should have taken care of it, right? I also copied that file into /usr/lib and then ran the ldconfig.

I found no libGL in the ET directory so there is nothing to replace there.

bwkaz, can you tell me what you *think* the problem is, maybe I configured something wrong in the kernel, something I am overlooking, or something I installed a while ago that I may not remember...

Thank you SO much for your help bwkaz. And anyone else who even thinks they could possibly know something is very welcome to share their knowledge. Thanks again.

DemosthenesB
06-08-2003, 11:28 PM
Just thought of something else. Maybe the problem is in my XF86Config file? Maybe I'm not specifying something correctly? I tried searching for and example file(one with the sti drivers I have) but to no avail...

If you think this could be the problem then just tell me the section of the file I shoul post.

Thanks again.

DemosthenesB
06-09-2003, 03:04 PM
OK., I was browsing around and came upon a command called "glxinfo," which displays the OpenGL renderer. the output is a follows:
name of display: :0.0
display: :0 screen: 0
direct rendering: No
server glx vendor string: SGI
server glx version string: 1.2
server glx extensions:
GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_EXT_import_context
client glx vendor string: SGI
client glx version string: 1.2
client glx extensions:
GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_EXT_import_context,
GLX_ATI_pixel_format_float
GLX extensions:
GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_EXT_import_context
OpenGL vendor string: VA Linux Systems, Inc.
OpenGL renderer string: Mesa GLX Indirect
OpenGL version string: 1.2 Mesa 3.4.2
OpenGL extensions:
GL_ARB_multitexture, GL_EXT_abgr, GL_EXT_blend_color,
GL_EXT_blend_minmax, GL_EXT_blend_subtract
glu version: 1.3
glu extensions:
GLU_EXT_nurbs_tessellator, GLU_EXT_object_space_tess

visual x bf lv rg d st colorbuffer ax dp st accumbuffer ms cav
id dep cl sp sz l ci b ro r g b a bf th cl r g b a ns b eat
----------------------------------------------------------------------
0x23 24 tc 0 24 0 r y . 8 8 8 0 0 16 0 0 0 0 0 0 0 None
0x24 24 tc 0 24 0 r y . 8 8 8 0 0 16 8 16 16 16 0 0 0 None
0x25 24 dc 0 24 0 r y . 8 8 8 0 0 16 0 0 0 0 0 0 0 None
0x26 24 dc 0 24 0 r y . 8 8 8 0 0 16 8 16 16 16 0 0 0 None

On the OpenGL renderer string it says "Mesa GLX Indirect." This is also what it says when I start up Enemy Territory(" You are using software Mesa (no hardware acceleration)! "). Could it have something to do with that?

Another thing, while I was searching for anything that said "glx" on my computer I came across a directory "Mesa 5.0.1"(Which I remember deleting a while ago) in my home directory, and when I "ls" to look it is not there. There is no leading period or anything. Just thought that was kind of wierd...

So thats the latest on my problem. Thanks.

bwkaz
06-09-2003, 07:12 PM
Originally posted by DemosthenesB
On the OpenGL renderer string it says "Mesa GLX Indirect." This is also what it says when I start up Enemy Territory(" You are using software Mesa (no hardware acceleration)! "). Could it have something to do with that? Yes, it's that exactly. ET sees that data, and says "you're using software rendering!" You might try the force option that the message is talking about, there's a chance it might help.

Another thing, while I was searching for anything that said "glx" on my computer I came across a directory "Mesa 5.0.1"(Which I remember deleting a while ago) in my home directory, and when I "ls" to look it is not there. Makes perfect sense if you were using locate to do the searching. locate doesn't look on your HD, it looks in its database. Use updatedb to update said database.

When I go to configure the kernel I did put in the P3 processor family, but there are no exra options to enable or disable SSE support, does it automatically become supported once the P3 is selected? Yes, it does get automatically enabled.

I found a patch(from http://people.redhat.com/mingo/mmx-patches/), but it is 2.4.14(mmx-2.2.14-A0), and although I have never patched a kernel it looks like it wont work as it is 6 versions behind. There are also some other ones that are 2.3.33 and above, but I read that odd numbers were developement versions and buggy. That patch is for kernels that didn't support SSE by themselves (basically, a lot of 2.2 kernels and some 2.3 ones). Don't bother with it -- you already have the functionality.

There is a libGL.so.1.2 under "usr/X11R6/lib" within the tarball, and I already copied the entire directory to the real /usr. So that should have taken care of it, right? I also copied that file into /usr/lib and then ran the ldconfig.

I found no libGL in the ET directory so there is nothing to replace there. If it still didn't work after copying that to /usr/lib and running ldconfig, then try the force workaround that the message is saying. Hopefully, the renderer string doesn't matter.

bwkaz, can you tell me what you *think* the problem is, Unfortunately, I really have no idea. I mean, I do -- ET sees a software renderer string in the GL library, and says "I'll be too slow", but it might be wrong, I don't know...

DemosthenesB
06-09-2003, 08:45 PM
Sorry I hadn't posted this earlier but I had tried to force the software render. It was disgusting, even in the title screen it would go at around 1fps(literally).

So basically I have to find a way to let ET know that I have opengl drivers. But maybe I really don't, I could have installed them wrong or something... Are there any standalone opengl drivers thast could work? Or do you know how to uninstall the "Mesa" files so it doesn't see them? Opengl.org pointed me to "Lokigames" which is supposed to have a link that explains how to use opengl for gaming, but the link is broken(http://faqs.lokigames.com/setupandtroubleshootingonx4faq.html).

Thanks again for taking the time to help me.

bwkaz
06-09-2003, 10:24 PM
Have you run fglrxconfig (or whatever it's called)? Don't know if it'll help or not, but it's worth a shot...

I just pulled down the RPM myself, and converted it to a cpio file, then extracted it and moved the libGL into my /usr/lib (rather than copying all the files to /usr, then doing whatever else). Running glxgears after that produced the same FPS as normal (I use nVidia's drivers normally, and no, nVidia's libGL was nowhere to be found), which was VERY strange. glxinfo also listed my normal renderer string (nVidia Corporation something or other).

So I wonder if glxinfo gets its information from the X server rather than the libGL file like I thought? Might want to double-check that fglrxconfig stuff...

DemosthenesB
06-09-2003, 11:42 PM
I have run the fglrxconfig, many times.

When you exracted the cpio file, you found that the libGL was in the fglrx/usr/X11R6/lib with the filename "libGL.so.1.2" right? and all you did was
cp libGL.so.1.2 /usr/lib
right? That was what I did(then running ldconfig) and it still says the same thing.

I found that running the glxgears gave me 170 to 200 fps, is that what I should get with my graphics card? If so, then the libGL should be working... Maybe the "Mesa" files are "in the way" do you think it would be safe to remove them, if so then how? Should I simpy delete everything that contains the word "mesa"?

One more thing, I also have another file that is called fgl_glxgears, but whenever I run it the program outputs"Error: couldn't get fbconfig."

Wow, thanks again for all your patience, I don't think I've ever had this difficult a problem.

bwkaz
06-10-2003, 09:59 PM
"couldn't get fbconfig"?

Hmm... no idea on that one.

And yeah, glxgears should be going quite a bit faster than 200fps. 200fps is in the software-rendering category.

Yes, that stuff was exactly what I did (well, I moved /usr/lib/libGL.* and /usr/lib/libGLcore.* out of the way first). Hmm...

You might try moving everything that has a Mesa in it to somewhere else (like /root) temporarily. Rerun ldconfig.

What happens if you ldd /usr/X11R6/bin/glxgears? Which libGL does it link to? Try this after moving the Mesa stuff, and after moving the GL in /usr/lib out of the way then retrying the copy.

DemosthenesB
06-10-2003, 10:49 PM
I'll keep looking for fbconfig, but I doubt it has much to do with my problem.

I tried to move the libGL files out of the way before copying the ones from "fglrx", but I don't seem to have a "libGLcore" file, just libGL.so, libGL.so.1, and libGL.so.1.2, should I have libGLcore?

When I try lld /usr/X11R6/bin/glxgears the output is as follows
libGL.so.1 => /usr/lib/libGL.so.1 (0x40025000)
libXext.so.6 => /usr/X11R6/lib/libXext.so.6 (0x4008e000)
libX11.so.6 => /usr/X11R6/lib/libX11.so.6 (0x4009b000)
libpthread.so.0 => /lib/libpthread.so.0 (0x40154000)
libm.so.6 => /lib/libm.so.6 (0x40169000)
libc.so.6 => /lib/libc.so.6 (0x4018a000)
libdl.so.2 => /lib/libdl.so.2 (0x402ad000)
/lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)

doesn't seem to see anything that has to do with mesa, and it sees my libGL.so.1 in /usr/lib, and I tried it after moving anyting that said mesa into my root directory. I also tried to run glxgears again and still with the same framerate...

Thanks again, and I'm sorry I couldn't respond earlier, I was at a friends house trying to convert him to opensource with linux. I'm sure he will be posting here soon to ask about his internet connection, that topic seems to spring up a lot.

bwkaz
06-11-2003, 10:42 PM
Originally posted by DemosthenesB
I tried to move the libGL files out of the way before copying the ones from "fglrx", but I don't seem to have a "libGLcore" file, just libGL.so, libGL.so.1, and libGL.so.1.2, should I have libGLcore? No, sorry, I should have been more specific. I have libGLcore, because my previous drivers were from nVidia. Only the nVidia drivers use libGLcore.

libGL.so.1 => /usr/lib/libGL.so.1 (0x40025000)
libXext.so.6 => /usr/X11R6/lib/libXext.so.6 (0x4008e000)
libX11.so.6 => /usr/X11R6/lib/libX11.so.6 (0x4009b000)
libpthread.so.0 => /lib/libpthread.so.0 (0x40154000)
libm.so.6 => /lib/libm.so.6 (0x40169000)
libc.so.6 => /lib/libc.so.6 (0x4018a000)
libdl.so.2 => /lib/libdl.so.2 (0x402ad000)
/lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000) it sees my libGL.so.1 in /usr/lib, and I tried it after moving anyting that said mesa into my root directory. I also tried to run glxgears again and still with the same framerate... Rats. Maybe try an md5sum on /usr/lib/libGL.so.1 and the libGL.so.1.2 that's in the fglrx package? If they're exactly the same, then try doing the same thing with the old libGL.so.1 and the fglrx package version. If they're the same, then that's the problem -- but I haven't the slightest clue how to fix it. :(

Does glxinfo say "direct rendering: yes", or "no"?

Thanks again, and I'm sorry I couldn't respond earlier, I was at a friends house trying to convert him to opensource with linux. Hey, don't apologize, no big deal. I work during the day anyway, so I didn't even notice anything. :)

Maybe Google has some idea on the fbconfig message?

Rüpel
06-12-2003, 03:08 AM
hi there,

i'm going to buy a radeon 9700 soon, so i'm pretty interested in this thread. :rolleyes:

since i'm going to use gentoo i also searched for that on the gentoo forums and i've found a pretty long, pretty interesting and pretty sticky thread, about getting ati radeon cards to work with gentoo:

http://forums.gentoo.org/viewtopic.php?t=54793

DemosthenesB, what linux-distribution do you use? did you compile your kernel yourself? check this posting in the aforementioned thread about setting up your kernel-config right:

http://forums.gentoo.org/viewtopic.php?p=359289#359289

i'm far from being a linux-guru, but is that ati-module actually active (lsmod)? what about your X-configuration (XF86Config file)?

DemosthenesB
06-12-2003, 05:01 PM
I tried the md5sum and they are all the same, so I guess that is my problem.

I also went and searched for any libGL file, and I removed the libGL.so.1 from the /usr/X11R6/lib directory(since they have the same "fingerprint" the libGL.so.1.2 should act the same as libGL.so.1, only I thought it was maybe a newer version) but ET looks for libGL.so.1 and exits saying that it couldn't find it.

when I try glxinfo it also looks for the libGL.so.1, but I also tried it before moving the file and it said "direct rendering: no."

I had seached google for fbconfig, and I found a file called "FrontBase-1.2-w.rpm" which does not sound like it has anything to do with openGL, radeon drivers, or graphics to me so I am pretty reluctant to install it...

Rupel, thanks a lot for those links, I will read them and try all the stuff the other people did. I have found threads that were Slackware specific, but they only describe how to install them(changing the file from rpm to tgz, and so on) and people just give up when they find out they have no hardware acceleration. This one looks promising as it is 7 pages long.

I am using Slackware 8.1, with a customized 2.4.20 kernel that I had confiured correctly like the post says.

The ati-modules are active, but not on startup, I have to call them with "insmod fglrx"

As for the XF86Config file(with ati i is called XF86Config-4) and one thing I'd point out is # === disable/enable XAA/DRI ===
Option "no_accel" "no"
Option "no_dri" "no"
which is basically telling it to use hardware acceleration, maybe the game is looking in the default XF86Config file, should I copy the XF86Config-4 file into the original XF86Config file(of course not until after saving the file into a different directory)?

Well thank you guys for your help, I will post back as soon as I have read the entire thread at the gentoo forums and hopefully I will be posting that it is working perfectly. Hey and when I get this thing to work maybe we can all play ET and own everyone.

Rüpel
06-13-2003, 03:19 AM
i don't know if you've already read all of that lengthy gentoo-thread, but on page 5 there's someone posting a working XF86Config file. here is the link to his posting:
http://forums.gentoo.org/viewtopic.php?p=346629#346629

maybe you verify these lines in your config-file:

Section "dri"
Mode 0666
EndSection

and in the modules section

Load "glx" # libglx.a
Load "dri" # libdri.a


maybe you also have a look in your XFree86-logfile and watch for errors/warnings - it usually can be found in /var/log. just type the following at your console:
grep '^(WW)' /var/log/XF*; grep '^(EE)' /var/log/XF*

DemosthenesB
06-21-2003, 05:15 PM
Ok, I read the thread a while ago and have been fighting with my kernel ever since. I finally got my kernel going correctly. I enabled MTRR support and loaded agp as a module. Now whenever I "startx" my screen flashes differently, so I know something is working, I'm sure my only problem now is the libGL.

The thread also says something about "emerge ati" is that "emerge" a gentoo-only command? And what exactly are the "use flags," please excuse my noob-ness.

Whenever I start up my computer I get a message: File /usr/X11R6/lib/libGL.so is too small, not checked.
This also happens when I try to ldconfig. I still get the same error when starting up ET and at glxinfo, and the same fps with glxgears.

If anyone knows what that message means please post, or if you have gotten it and how you fixed it.

Thanks again.

Rüpel
06-22-2003, 05:18 AM
Originally posted by DemosthenesB
The thread also says something about "emerge ati" is that "emerge" a gentoo-only command? And what exactly are the "use flags," please excuse my noob-ness.


yes, emerge is gentoo-only. basically that's the thing gentoo is all about :rolleyes:

emerge xyz means install xyz (in a fancier way, i.e. automatically resolve all dependencies, get the source from the web, build xyz, copy xyz's files into their specific locations, and even more)

since gentoo builds all programs from source on your local computer, you are able to control the build-process of each application. that's what the use flags are for. if you are not going to use gnome for instance, and you emerge an application with optional gnome-support (which you actually don't need), emerge looks at your use-flags (where you have defined 'i don't use gnome') and the optional gnome-support is not built in - resulting in the tighter, faster system.

well, somehow. in short... :rolleyes: ;)

Rüpel
06-22-2003, 05:22 AM
btw: you didn't respond to my last posting. are your XF86Config file and the X logfile ok?

DemosthenesB
06-22-2003, 12:41 PM
Rupel, I do have all those verified in my XF86 file, sorry I overlooked your post. doing the "grep '^(WW)' /var/log/XF*; grep '^(EE)' /var/log/XF*" gives me:
(WW) Open APM failed (/dev/apm_bios) (No such device)
(WW) fglrx: No matching Device section for instance (BusID PCI:1:0:1) found
(WW) fglrx(0): DRI initialization failed!
(EE) module ABI minor version (2) is newer than the server's version (1)
(EE) Failed to load module "dri" (module requirement mismatch, 0)

I'm going to keep looking for this topic under Slackware. Thanks a lot for you help.

Rüpel
06-22-2003, 01:12 PM
have you tried changing BusID "PCI:1:0:1" to BusID "PCI:1:0:0" in your XF86Config? i'm not sure, what that exactly does, but on someone else's system, that did the trick.

as long as you get these errorlines in your xfree-logfile (the ones starting with "(EE)" mentioning DRI) you won't have any hardware 3D.

DemosthenesB
06-22-2003, 05:49 PM
Sorry I hadn't added that in my last post but I had checked my XF86 file and it was already set at 1:0:0. I had tried to set it to 1:0:1 to what would happen and X didn't start.

Do you have any idea what that line(File /usr/X11R6/lib/libGL.so is too small, not checked.) means? I don't think I have ever gotten it before.

mdwatts
06-22-2003, 05:54 PM
Use

lspci

to verify the BusID is correct for your video card.

i.e.

01:05.0 VGA compatible controller: nVidia Corporation NV20 [GeForce3 Ti200] (rev a3)

Of which I have

BusID "1:5:0"

in /etc/X11/XF86Config-4.

DemosthenesB
06-22-2003, 06:10 PM
mdwatts, I tried your command and it gives me two ID that have the "ATI" description.
01:00.0 VGA compatible controller: ATI Technologies Inc: Unknown device 4e44
01:00.1 Display controller: ATI Technologies Inc: Unknown device 4e64

Currently I have it set at 1:0:0, the VGA compatible one, what is the difference? I had already tried the 1:0:1 ID but X did not work at all. Another odd thing is that they both have the "Unknown decive" label at the end.

Thank you for your help.

mdwatts
06-22-2003, 06:14 PM
"01:00.0 VGA compatible controller" should be the correct device.

Do you have another display controller or does the 9700 have dual display?

bluesky
06-22-2003, 06:18 PM
The radeon driver is available in tar.gz format at gentoo downlaoad sites:

http://gentoo.oregonstate.edu/distfiles/

It is for XFree86 4.2 but I believe it would work for 4.3 as well since gentoo i suing 4.3 any way. :p

DemosthenesB
06-22-2003, 07:51 PM
mdwatts:
The 9700 does have dual display, but I did not set it as an option in the config(fglrxconfig).

bluesky:
Thanks for the link but doing the "rpm2tgz" in slackware has the same files and directories, thanks anyways.

Do any of you think getting the DRI drivers from dri.sourceforge.net when(if) they make drivers for the 9700 make any difference? These drivers are supposed to enable 3d support as well but no luck so far(at least for me).

bluesky
06-22-2003, 10:24 PM
Well. If you tried the fglrx drivers w/o success then probably you may want to try the drivers from driverguide.com.

http://www.driverguide.com/

DemosthenesB
06-23-2003, 12:42 AM
Ok, I think I got something out of my drivers. First I tried installing straight out of the rpm package, but no luck. So I searched around and noticed a "patch" directory. I moved some files from /lib/modules/fglrx/build_mod/patch/drivers/char/drm(the files were drmP.h, drmP.h.rej, drmP.h~, and drm_proc.h) to /usr/src/linux-2.4.20/drivers/char/drm.

So I build the modules and then loaded them. After that I tried the fglrxconfig and I got some extra options(like what type of FireGL I would like) so I know I did something by moving those files. I also get less options for the horizontal and vertical sync.

However, now when I try to "startx" it just takes me to a blank screen(after some flickering). it gives me the same errors as when I type grep "'^(WW)' /var/log/XF*; grep '^(EE)' /var/log/XF*" at the console.

bluesky, I'd rather leave driverguide.com the last possible option. The fglrx drivers are *supposed* to work right? Thanks anyways.

Thanks again for your help.

bluesky
06-23-2003, 06:43 AM
Yeah. The fglrx driver that redhat and gentoo used comes from the drivers posted by ATI and which says only for XFr3386 4.1 and 4.2. There is another link which has driver (obviously comes from the same ATI source) that has been designed for XFree86 4.3:

http://www.schneider-digital.de/html/body_download_ati.html

I only used "radeon" driver( my card is radeon 7000 ) so I can't make any assessment. But just in case you want to try. I know many people have been using drivers from the link. :p

BTW, if you do succeed making it work, you may want to post at least a brief account of what you did. since many people are interested to know, including myself. :p

Rüpel
06-23-2003, 07:25 AM
there were people that made that xfree4.3 driver work - but all of them ecountered instabilities when actually playing a game for some minutes (or even less, i.e. just moving in an fps-game).

at the moment it is better to stick with xfree 4.2 using the 2.51 ati drivers.

at least, thats what i remember from reading "ati 9700pro linux" threads for several weeks now. my own card is ordered but not shipped yet...

DemosthenesB
06-23-2003, 01:35 PM
bluesky:
Trust me, if I ever get these drivers to work I will gladly post exactly what I did, I don't want anyone else to go through this mess.

Rupel:
What is your current video card, and what linux distrobution do you have?

DemosthenesB
06-23-2003, 02:01 PM
I forgot to add something to my last post.

The error I now get from starting x with the XF86Config-4 file I get from fglrxconfig is as follows:
(==) Log file: "/var/log/XFree86.0.log", Time: Mon Jun 23 10:22:06 2003
(==) Using config file: "/etc/X11/XF86Config-4"
(EE) module ABI minor version (2) is newer than the server's version (1)
(EE) Failed to load module "dri" (module requirement mismatch, 0)
(WW) fglrx: No matching Device section for instance (BusID PCI:1:0:1) found
Module called exit() function with value=1
XIO: fatal IO error 104 (Connection reset by peer) on X server ":0.0"
after 0 requests (0 known processed) with 0 events remaining.

If anyone knows what the problem is or what the problem could be please post.

Thank you.

Rüpel
06-23-2003, 02:09 PM
do you know, which xfree86 version you have? i can't find it in this thread...

my current video card is a voodoo banshee based 16mb card (with a busted fan) in a more than 4 years old computer running mandrake 8.2 (and win2k).
erm, yes. i've ordered more than just a new graphics card ;) :D

so yes: all the things i know about radeon drivers under linux (which is very little), i know from reading forum-threads here and there... :rolleyes:

btw: we've made it to the top-ten "hot threads" list on the starting page of justlinux.com

http://jjbgf.de/wbboard/images/smilies/OLA.gif
http://jjbgf.de/wbboard/images/smilies/party2.gif

Rüpel
06-24-2003, 09:07 AM
just to let you know: my dealer contacted me about my new graphics-card and he's unable to ship the radeon9700pro model. so now i switched to a radeon 9800 pro card. let's see, if i can get that one to work later this week... :)

DemosthenesB
06-24-2003, 11:25 AM
Wow, thats awsome. Are they going to charge you extra or anything?Were they like "Aw, I'm sorry Rupel, but I can't send you this slightly out-of-date card, I guess this shiny new superior graphics card will just have to do..."

And btw, good luck on getting that to work, not that I've every read anything on problems with it or anything, just that Linux and new hardware usually don't mix well.

Rüpel
06-24-2003, 12:42 PM
nah, of course i have to pay the difference, but it's only 35 bucks (dollars and euros are somehow 1:1, aren't they?) - and no: when i ordered the card they already told me: "we can't promise, that the radeon9700pro will actually be available". these people are ok. :)

and i've checked the ati website and they link to the 2.51 drivers if you select a radeon 9700 pro. on the other hand they link to a 'search for a driver in the community'-page when you want to have a linux-driver for a 9800 pro board.

so i fear i have to stick with win2k and the linux command-prompt for a while (until ati releases new drivers, which will hopefully support xfree 4.3 and my radeon 9800 pro)

blackhawk714
06-24-2003, 09:24 PM
I don't mean to be rude here, but I get PCWORLD magazine every month (gift) and i read the R9800Pro review and the R9700Pro beats the new 9800 Pro by 6 frames per second running quake 3 arena. just thought I'd let you know that in case you might want to get the card from a different dealer! :D

Rüpel
06-25-2003, 02:00 AM
i've checked quite a bunch of reviews (both, off- and online) before i switched. besides the fact that i'm not going to play quake (whatever version), it's a matter of the drivers (which were not optimized for the 9800 chip), as far as i understood. i'm pretty sure that if your magazine repeats its benchmarks with the new catalyst drivers released earlier this week, the 9800 pro will beat the 9700 pro.

DemosthenesB
06-28-2003, 04:56 PM
Well everybody I finally have hardware acceleration, but I had to reformat my linux drive to get it. I know it isn't exactly the right thing to do, more of a quick way to get things done. I am obviously a noob and you can tell from the thread that there wasn't much else to do. I am eternally grateful to all those who helped me.

I now get over 4,000 fps from running glxgears and there is no mention of Mesa Indirect in glxinfo.

I really wish I could say my problems are over but now it comes back down to the begining of my reason for starting this thread, getting Wolfenstien to work. I will search for ET Linux specific threads but I just wanted to see what you guys thought. When I go to install the game ("sh et-linu...") I get the following output:
Verifying archive integrity...Error in MD5 checksums: b232c368617d9dcfe1b70cc0a568c70a is different from f76b8599439dd882a52ba038d4700edc

Thank you guys again, especially Rupel and bwkaz.

bwkaz
06-28-2003, 05:38 PM
Which distro did you install, for curiosity's sake?

Tha bad MD5 checksum almost assuredly means that the et-linux-whatever.bin file has been corrupted. See if you can download it again, or something.

DemosthenesB
06-28-2003, 05:53 PM
I am using Slackware 9(I've tried other distro's but I still think Slackware is the best). I know that people said the new version of XFree doesn't mix well with the drivers for the 9700 but I believe my fps in glxgears speaks for itself. What distro do you have bwkaz?

Thanks for the input, I'll get right on that.

bwkaz
06-28-2003, 08:28 PM
Linux From Scratch, loosely based on v3.3.

Although I'm building a CVS May 15 2003 version on another partition (have to boot over to that and pull down/compile/install Mozilla sometime... ;)).

So even with the same GL library you're not getting the "mesa indirect" stuff... strange...

Well, I suppose if it's working, no big deal, right? ;)

DemosthenesB
06-28-2003, 08:49 PM
Well, I believe it had to do with the fact that I had installed Mesa prior to trying to get the drivers to work last time. Maybe there was some interferace, although I had removed everything that said "Mesa" but thats already behind us.

Well, I did as you said and downloaded the game from a different location, and everything installed without a problem. However, when I go to play a game(after picking the server and all that stuff) when the "loading" bar is almost all the way done the game crashes, throwing me back into X with the 800*600 setting I had in ET. The message it gives me is as follows:
Initializing Sound Scripts
...loading 'sound/scripts/vo_allies.sounds'
...loading 'sound/scripts/vo_axis.sounds'
...loading 'sound/scripts/oasis.sounds'
done.
LOADING... graphics
LOADING... maps/oasis.bsp
stitched 0 LoD cracks
...loaded 19509 faces, 178 meshes, 1005 trisurfs, 0 flares 0 foliage
LOADING... entities
LOADING... game media
LOADING... - textures
LOADING... - models
LOADING... - weapons
LOADING... - items
LOADING... - inline models
LOADING... - server models
/usr/local/bin/et: line 5: 15242 Killed ./et.x86 $*

I'm sorry to bother you with yet another of my problems, thanks again.

bluesky
06-29-2003, 06:24 AM
So you solved the installation problem. Great!!! Are you sure that your card work now? Have you tested on a simpler 3D game?

DemosthenesB
06-29-2003, 02:30 PM
I have tried to install Armagetron from linuxgames.com but no such luck, I'll try to figure that one out later. I re-installed ET and put the settings to "Fastest" so that the graphics are all dumbed down and it works fine, I'm guessing my drives don't know how to handle some of the more complex graphical features of ET. So they do work, within bounds. I can't wait until ATI releases their own drivers for the latest version of XFree, or DRI that uses the hardware to the fullest. The only problem I get now is that at random times the game just freezes and it plays the last second of audio in a constant loop, I then have to stop the process by going back into the terminal I stated x with and pressing ctrl+c. Then I have to restart because I cannot get back into X.

Rüpel
06-29-2003, 03:06 PM
Originally posted by DemosthenesB
The only problem I get now is that at random times the game just freezes and it plays the last second of audio in a constant loop, I then have to stop the process by going back into the terminal I stated x with and pressing ctrl+c. Then I have to restart because I cannot get back into X.

see my post on page 3 of this thread:
http://justlinux.com/forum/showthread.php?s=&threadid=103370&perpage=15&pagenumber=3#post586789

that's exactly the same problem dozens of radeon 9700/9500 users report, when using xfree 3.0 and the fireGL drivers. that's why i told you to stick with xfree 4.2 and the 2.51 drivers. but maybe it isn't that easy to downgrade slackware to xfree 4.2...

well, and maybe ati finally releases new drivers next week ;)

btw: my radeon 9800 pro is here - but i still have to wait for the case of my new computer. so all i can tell you is, that a sapphire radeon 9800 pro looks really red and cool :rolleyes: :D

DemosthenesB
06-29-2003, 03:24 PM
I had tried to downgrade my xfree by removing the package through the package tool(pkgtool) and then downloading XFree 2.0(I had forgotten to get it before removing xfree and had to use lynx, quite an experience). I followed the instructions it gave me and then my ati drivers said it couldn't find some header files or something and then re-re-formatted and just used 4.3.

I got armagetron installed but it dies even before the game begins. So I got glTron(I'm not really that big a fan of tron(never even seen the movie) or tron games, those are just the ones that caught my attention) and it works perfectly.

DemosthenesB
06-29-2003, 03:29 PM
oh, and btw: Congratulations on getting your 9800 pro, I was SO happy when I got my 9700 last November. I hope you get it to work without any problems, and if you do have problems that you learn a lot from them.

bwkaz
06-29-2003, 09:49 PM
Originally posted by DemosthenesB
Initializing Sound Scripts
...loading 'sound/scripts/vo_allies.sounds'
...loading 'sound/scripts/vo_axis.sounds'
...loading 'sound/scripts/oasis.sounds'
done.
LOADING... graphics
LOADING... maps/oasis.bsp
stitched 0 LoD cracks
...loaded 19509 faces, 178 meshes, 1005 trisurfs, 0 flares 0 foliage
LOADING... entities
LOADING... game media
LOADING... - textures
LOADING... - models
LOADING... - weapons
LOADING... - items
LOADING... - inline models
LOADING... - server models
/usr/local/bin/et: line 5: 15242 Killed ./et.x86 $* Hmm... is there anything in your kernel logs about "Out of memory: killing process XXXX"? If so, try either adding more memory or making your swap partition bigger (you do have a swap partition, right?).

Even if not, make sure that you have either a bunch of memory, or a bunch of memory plus a bunch of swap. And don't run anything else that uses a lot of physical RAM or swap, if you can.

DemosthenesB
06-29-2003, 10:16 PM
Oh, so thats basically telling me that I am running out of RAM, and to save my computer it is exiting. I do have 256MB RAM, it was enought to run the "normal" setting in windows, and no, I never installed swap partitions, I had tried but I just couldn't get the computer to "see" it durin the install. Should I have enable RAM support in the kernel config under MTD(Memory Technology Devices)? I'll try that and see what I get.
Is there any way of adding a swap partition now? Do I even really need a swap partition? I never run anything in the background while trying to play and I do have 256MB...

bwkaz
06-30-2003, 10:13 PM
Originally posted by DemosthenesB
Oh, so thats basically telling me that I am running out of RAM, and to save my computer it is exiting. Well... maybe. It depends on what your kernel logs say. If they mention what I was saying ("Out of memory: killing process XXX"), then yes, you ran out of memory, and something needs to get killed. Either the whole system, or your game... :D

and no, I never installed swap partitions, I had tried but I just couldn't get the computer to "see" it durin the install. Well, that's the problem -- either get more RAM, or turn on swap (how will be explained, hang on a minute ;)).

Should I have enable RAM support in the kernel config under MTD(Memory Technology Devices)? I'll try that and see what I get. It won't help. MTD's are stuff like Flash ROMs and such -- memory on a stick or memory on a PCI card or something like that. It won't affect how real RAM gets used.

Is there any way of adding a swap partition now? Do I even really need a swap partition? Yes, and yes. Even with 768MB, I have a 384MB swap partition (and it got used VERY often, whenever I played Rune, UT, or UT2k3, back when I only had 256MB of RAM).

To create one, resize your partitions and create another one. Then, reboot, and mkswap /dev/hdXX (specify the new partition for /dev/hdXX). Then do a swapon /dev/hdXX (all as root).

If that succeeds, then add a line to your /etc/fstab looking like:

/dev/hdXX swap swap defaults 0 0 and reboot. Your swap should be up after a reboot (your distro's init scripts should do the swapon for you).

DemosthenesB
07-01-2003, 05:06 PM
I added a swap partition of 500 MB and it seems to run just fine, thank you very much bwkaz. Oh, and about the problem with the game just exiting unexpectedly, for those who are having this problem just add "+set r_ext_compiled_vertex_array 0" and it should run perfectly. I got it from http://zerowing.idsoftware.com/linux/et/. The game still doesn't run at my cards potential but I blame that on the drivers.
Where exaclty should my kernel logs be? I check my syslog in /var but nothing about the kernel killing the game in there. Thanks again bwkaz.

AlexPlank
07-01-2003, 05:15 PM
Post a how i did it with all the steps you did because this thread is compicated with many suggestions; I hope this will work for 9500 Pro as it is just a slowed down 9700

bwkaz
07-01-2003, 06:44 PM
Originally posted by DemosthenesB
Where exaclty should my kernel logs be? I check my syslog in /var but nothing about the kernel killing the game in there. Thanks again bwkaz. Either near the end of the output of dmesg (if you haven't rebooted since the game got killed), or near the tail end of /var/log/messages (on RH/Mandrake/probably others) or /var/log/sys.log (on LFS).

Other distros call this file something different. If you know how to read /etc/syslog.conf (or the syslog.conf(5) manpage ;)), then figure out where the kern.* messages are going, from /etc/syslog.conf, and check the end of that file.

Rüpel
08-22-2003, 09:37 AM
i just remembered this thread. my 9800pro is working with gentoo. :D

have to use a 2.6 kernel (because of the kt400 chipset), but it works.

Rüpel
09-05-2003, 09:13 AM
just yesterday ati released new drivers: 3.2.5

why not give it a try DemosthenesB?
http://www.ati.com/support/driver.html

Icarus
09-05-2003, 10:21 AM
Yup, new ATi drivers but I've found a few problems...

1) The download is RPM only, so anyone using a non-RPM based distro will have to use Alien to convert it

2) It removes the /etc/X11/XF86Config and wants to you reconfigure everything again from scratch...and my old config has mouse problems for some reason, need to fix that still...

But I do have to say when I finally managed to open a terminal and run glxgears, the new drivers are faster then the firegl :)
now to fix my mouse...

DemosthenesB
09-10-2003, 07:30 PM
Sorry for the late reply, but I was working on my Linux From Scratch system and then school started and Chemisty AP kinda took over my weekdays. I'll try the new drivers on my slack partition as soon as I can, but right now I tried to use them on LFS(which has XFree 3) and I get this error :You arent running in a 'login shell'.
Please login directly from a console
or use 'su -l' to get the same result.
I've already started a thread on that here (http://justlinux.com/forum/showthread.php?s=&postid=621825#post621825)
(when I was trying to get the old XFree drivers to work on LFS wich gave me the same problem) and I guess its really an odd problem...

Congratulations on your new card Rupel, I hope you know Savage (http://pc.ign.com/articles/434/434307p1.html) is coming out for linux soon, a game I've been anticipating for quite a while.

bwkaz
09-10-2003, 09:56 PM
Originally posted by DemosthenesB
but right now I tried to use them on LFS (which has XFree 3) ???

No, not by default it doesn't...

The current BLFS book (which requires probably LFS 4.1 or LFS CVS, which is soon to be 5.0) installs XFree86 4.3.0, not 3.x.x. Plus, the new ATI drivers are going to require XFree86 4.x.x; they won't work with 3.x.x.

As for the error -- do what it says. You may have to shut down X also, but this depends on the driver itself.

Hit ctrl-alt-f2 and log in. Then run it from there.

DemosthenesB
09-11-2003, 12:08 AM
bwkaz, I'm real sorry, I did mean 4.3.0, really sorry, I should have double checked or something. So just to clarify, my LFS system has 4.3.0, not 3. And I have done what it said, "su -l", I also tried "su --login", and I have tried running the script right after I login right from the normal console out of X.

Thank you for the help bwkaz, and again, sorry for the mistake, I always get confused with the two and I don't know why as I have never used XFree 3

bwkaz
09-11-2003, 07:07 PM
Originally posted by DemosthenesB
bwkaz, I'm real sorry, I did mean 4.3.0 Oh, OK, no big deal. ;) I was just wondering why you'd be using something about 5 versions old... :p

Anyway, AAAAUUUGHHH! It's a problem with the drivers, sort of. I just pulled them down myself, ran the RPM through rpm2cpio and extracted the cpio file. As it turns out, the make_install.sh script that you're running is doing some PATH magic with the "which" utility (without checking for it first... of course) to try to see whether or not you're in a login shell.

The tests run like this:

(1) Run "which depmod". If this succeeds, then assume you're in a login shell and continue. (So this is failing.)

(2) Add /usr/local/sbin:/usr/sbin:/sbin to the front of your PATH, and rerun "which depmod". If it succeeds this time, then print the message you're getting.

So the problem is that depmod is not in your PATH. After you exit X, and log in to a terminal (as root directly; don't mess with su at all), do an echo $PATH and which depmod. If your PATH has /sbin and /usr/sbin in it (/usr/local/sbin shouldn't be needed), then the which command should have printed "/sbin/depmod" or something like that.

If your PATH doesn't have /sbin in it, then add it manually (make sure you export it too) before you rerun the script.

DemosthenesB
09-15-2003, 09:58 PM
bwkaz, I tried what you said and it worked just fine, thank you so much. One thing I found odd was that my LFS partition did not even have the "which" utility, I had to grab it off my slack partition.
Where did you find how to do that? I had been searching for a while before posting and couldn't find anything useful. Thanks again, it works just fine now.

bwkaz
09-16-2003, 08:26 PM
You can get which from

http://www.xs4all.nl/~carlo17/which/

It's not installed by default in LFS, basically because it's not needed to recompile LFS from LFS.

It is mentioned in the BLFS book, though. But it's pretty simple to install.

Where did I find out how to do what? The tests? That was from reading the script, and from experience with bash. :)

BKewl
09-22-2003, 01:29 AM
Dagnabit Demosthenes tell us how you finally got the drivers to run! :) I've been to 1 other board and a mailing list and people have had the same problems; it's a plague. From the time you reformatted your drive, what did you do that was meant to facilitate installing the drivers? Give us the lowdown in great detail!

DemosthenesB
09-22-2003, 09:17 PM
Well, it has been a while since I installed the drivers on my slackware partition, but I'll give it a shot anyway.

OK, so I reformatted in order to make sure there were no dependencies getting in the way of the install.
After this I got the drivers in .rpm(which is the only format provided) and ran "rpm2tgz fglrx-ver.rpm" which converted the rpm into tgz.
I then made a new directory in my home directory(called it fglrx) and untarred the files within the package into the folder.
Then I copied all the files in my "fglrx" folder into their respective locations("lib" into /lib and "usr" into /usr).
After this go to "/lib/modules/fglrx/build_mod" and run the "make.sh" script(either with "./" or "sh").
If all goes well move one level up(to be in "/lib/modules/fglrx" and run the "make_install.sh" script.
If your path is in order you can run Xconfig with "fglrxconfig," and if not then just search for the fglrxconfig file and run it.

That should have done it, as it worked for me. I think that the reason why the drivers did not work was because I had downloaded a bunch of 3D-graphics programs in order to run 3Dwm(which I never did use anyway). I hope that does it, if I think of anything new I will be sure to add it to the thread and if you see any problems with my post or need some clarification just ask.

BKewl
09-22-2003, 11:25 PM
I think there's a problem with running the rpm version of the FireGL drivers in Slack. I did that, and it turns out my libGL.so.1.2 didn't get overwritten, so it was the old version. Just beware out there anyone. I'd recommend following DemosthenesB's directions for installation. I myself ended up rpm2tgz'ing the rpm, then running installpkg on the result, then untarring the tgz and copying over the /untarred_dir/usr/X11R6/lib/libGL.so.1.2 to /usr/X11R6/lib/, overwriting the old one there. You may want to backup the existing one. Also, you may need to make a symbolic link in /usr/lib so that certain programs can find it (i.e. RTCW). To do that, cd to /usr/lib and type "ln -s /usr/X11R6/lib/libGL.so.1 libGL.so". That'll create a symbolic link pointing to the libGL.so.1, which for some reason seems a little more transparent.