ATI Radeon 9600 XT on Mandrake 10.1

Se7en.62 · 02-22-2005, 01:15 PM

I don't want to be construed as saying that ATI cards are bad. In Windows their are easily equal or better than NVidia, and generally much cheaper. In linux, however, their drivers are complete crap, usually in the order of making an ATI card about half as fast as it's NVidia counterpart. That's not acceptable. I didn't pay $500 to replace my GF4200Ti with a card that's about the same speed.

Is Gran Tourismo 4 really out this week? Oh man, there goes another 5 months of my life....

dsnuggs · 02-22-2005, 03:38 PM

This sounds a good idea then. I wonder if I can get a similar deal for my Radeon 9700PRO.

Worth a try.

But like you say. If it is as fast as an ATI.

Se7en.62 · 02-22-2005, 03:55 PM

Mine was a special deal woith a client i'm building a computer for. I do the build, install and test for free, he gets my ATI card, and I get a brand new 6600GT. I'm thinking of throwing in a couple hundred more for a 6800Ultra.

I'll have to see if they're worth the extra money. BFG has a factory overclocked job that looks mighty appealing.

d1s4st3r · 02-22-2005, 04:23 PM

Could you please explain me this?

Code:

ATI module generator V 2.0
==========================
initializing...
cleaning...
patching 'highmem.h'...
assuming new VMA API since we do have kernel 2.6.x...
doing Makefile based build for kernel 2.6.x and higher
make -C /lib/modules/2.6.10/build SUBDIRS=/lib/modules/fglrx/build_mod/2.6.x modules
make[1]: Entering directory `/usr/src/linux-2.6.10'
  CC [M]  /lib/modules/fglrx/build_mod/2.6.x/agp3.o
In file included from include/linux/pagemap.h:10,
                 from /lib/modules/fglrx/build_mod/2.6.x/agp3.c:63:
include/linux/highmem.h:87:29: macro "clear_user_page" requires 3 arguments, but only 2 given
In file included from include/linux/pagemap.h:10,
                 from /lib/modules/fglrx/build_mod/2.6.x/agp3.c:63:
include/linux/highmem.h: In function `clear_user_highpage':
include/linux/highmem.h:87: error: `clear_user_page' undeclared (first use in this function)
include/linux/highmem.h:87: error: (Each undeclared identifier is reported only once
include/linux/highmem.h:87: error: for each function it appears in.)
include/linux/highmem.h:86: warning: unused variable `addr'
include/linux/highmem.h: In function `memclear_highpage_flush':
include/linux/highmem.h:105: warning: implicit declaration of function `out_of_line_bug'
include/linux/highmem.h:108: warning: implicit declaration of function `flush_dcache_page'
include/linux/highmem.h:109: warning: implicit declaration of function `flush_page_to_ram'
include/linux/highmem.h:119:34: macro "copy_user_page" requires 4 arguments, but only 3 given
include/linux/highmem.h: In function `copy_user_highpage':
include/linux/highmem.h:119: error: `copy_user_page' undeclared (first use in this function)
make[2]: *** [/lib/modules/fglrx/build_mod/2.6.x/agp3.o] Error 1
make[1]: *** [_module_/lib/modules/fglrx/build_mod/2.6.x] Error 2
make[1]: Leaving directory `/usr/src/linux-2.6.10'
make: *** [kmod_build] Error 2
build failed with return value 2

I tried to move by hand the file that ATI's coderz call "patch" (when it's obviously a modified version of file "/usr/src/linux-2.6.x/include/linux/highmem.c" ...), then recompiled the kernel and I got this:

Code:

  CHK     include/linux/version.h
make[1]: `arch/i386/kernel/asm-offsets.s' is up to date.
  CHK     include/linux/compile.h
  CC      init/do_mounts.o
In file included from include/linux/pagemap.h:10,
                 from include/linux/nfs_fs.h:15,
                 from init/do_mounts.c:10:
include/linux/highmem.h:87:29: macro "clear_user_page" requires 3 arguments, but only 2 given
In file included from include/linux/pagemap.h:10,
                 from include/linux/nfs_fs.h:15,
                 from init/do_mounts.c:10:
include/linux/highmem.h: In function `clear_user_highpage':
include/linux/highmem.h:87: error: `clear_user_page' undeclared (first use in this function)
include/linux/highmem.h:87: error: (Each undeclared identifier is reported only once
include/linux/highmem.h:87: error: for each function it appears in.)
include/linux/highmem.h:86: warning: unused variable `addr'
include/linux/highmem.h: In function `memclear_highpage_flush':
include/linux/highmem.h:105: warning: implicit declaration of function `out_of_line_bug'
include/linux/highmem.h:108: warning: implicit declaration of function `flush_dcache_page'
include/linux/highmem.h:109: warning: implicit declaration of function `flush_page_to_ram'
include/linux/highmem.h:119:34: macro "copy_user_page" requires 4 arguments, but only 3 given
include/linux/highmem.h: In function `copy_user_highpage':
include/linux/highmem.h:119: error: `copy_user_page' undeclared (first use in this function)
make[1]: *** [init/do_mounts.o] Error 1
make: *** [init] Error 2

What the hell?!?
The first time I got this:

Code:

FATAL: Error inserting fglrx (/lib/modules/2.6.8.1-12mdk/kernel/drivers/char/drm/fglrx.ko): Invalid module format
failed.

I tried to "strace modprobe fglrx" and the last output lines told me something about invalid ELF file.
This means that compilation (compilation?!? of *CLOSED SOURCE* drivers?!?) went wrong.
Tried to "patch" (ridiculous...) "highmem.c" in kernel tree and "/usr/include/linux" and I got a waterfall of errors during "make..."
ATI SUCKS and ATI DRIVERS SUCK MORE!!!

Se7en.62 · 02-22-2005, 04:59 PM

Ok, I've heard of this patch, but I was under the assumption that is was no longer nessisary.

Was it for machines running the 2.6.10 kernels?

If so, I'd first try compiling your own version of Xorg 6.8.2, then compiling your fglrx like I did (details in one of my previous posts in this thread.)

Oh, and make sure your kernel source (is installed and) matches the version you're running.

cathectic · 02-22-2005, 05:26 PM

d1s4st3r,

You should not be applying patches to the ATi drivers (and certainly no patches to the kernel!) if you are using the latest versions of the drivers from the ATi website (8.10.19).

(If you are using any of the 2.6.11 kernel release candidates, you will need a seperate patch against the drivers *only*! However, the output you've given leads me to believe you are using a 2.6.10 kernel, in which case, you don't need *any* patches.)

For Debian, unofficial repackaged ATi drivers can apparently be found here: http://xoomer.virgilio.it/flavio.sta...installer.html

dsnuggs · 02-24-2005, 04:17 PM

Se7en.62

That's a good idea actually. I do builds and re-builds of PC's. I'll see if anyone wants a PC sorted.

d1s4st3r

As cathectic said. You should not have to do any patching if you are using the latest ATI drivers.
It is hard enough trying to get these ATI going without complicating things.
I have found that the SuSE 9.2 installs the source files for you, if selected on installation of system. That made things a lot easier for me.

BUT alas. Still not working.

I am going to try what Se7en.62 said. And I quote:

"I'd first try compiling your own version of Xorg 6.8.2, then compiling your fglrx like I did (details in one of my previous posts in this thread.)"

So that will give an up to date Xorg and fglrx.

Worth a try I think.

Se7en.62 · 02-24-2005, 04:51 PM

Let me know how it goes, and if I can offer any more advice. They were a bitch to get working, but now that they do they are pretty solid. Slow, but solid.

dsnuggs · 02-24-2005, 06:50 PM

Will do Se7en.62.

Thanks.

dsnuggs · 03-14-2005, 04:17 AM

Hi Se7en.62

How is the 6600GT going?

I think that I have given up on the ATI Radeon 9700Pro, and thought that I would go along the same lines as you have.

I have looked up the nVidia 6600GT, and the speed is said to be better than the 9800Pro. So that convinced me to purchase one. (I have not purchased one yet)!

I have since been onto the 6600GT question web site, and fond that there where more problems with that card.

Then I thought, Se7en.62 was going to install one. I'll ask him.

So, is it working 3D in Linux ok?

Did you also buy the 6800Ultra, if so, is that working ok?

Se7en.62 · 03-14-2005, 07:37 AM

I haven't switched yet, as I have a project I have to finish first before i can go back to fooling around with my box any more. I've heard about some problems with 6600GT's, but i think that Nvidia has them mostly solved. Unlike ATI, they actually pay attention to thier Linux products.

dsnuggs · 03-14-2005, 01:09 PM

That's good enough for me.

I'll buy a 6600GT.

Thanks Se7en.62

dsnuggs · 03-15-2005, 05:56 AM

You'll never guess what happened last night. I installed the new ATI driver, Version: 8.10.19, and it accepted it!!!!

I ran the fglrxconfig, and ran the fglrxinfo test. The only problem I have now is that the direct_rendering is still set to NO.

I have placed the text into the modprobe, and re-booted. But no rendering.

Any ideas?

By the way. I'm now using SuSE 9.2

Se7en.62 · 03-15-2005, 07:15 AM

OK, it's early, so this may or may not be corrent.

i think you have to edit your Xorg.conf file and select yes for direct rendering.

i'll take a closer look later today.

dsnuggs · 03-16-2005, 05:33 PM

This I shall try.

Looks about right.

Thanks Se7en.62, I'll let you know what happens.