[ut2003io] AMD Systems

[David Krider]

On Wed, 2002-11-20 at 20:41, Peter Kemmerer wrote:
> I've had the same problem, and have discussed it with Ryan, Daniel, and
> a very nice fellow from nVidia offlist for awhile.  I'm running 2.4.19
> on an AMD 1800XP system, motherboard uses the VIA KT333 chipset,
> VisionTek Ti4600, have tried AGPGART and nvidia drivers, various apeture
> sizes, mem=nopentium enabled, yada, yada, yada.  Here are the main
> points of our discussion and examination:

> So, running BeyondUnreal and not being able to play UT 2003 at all, I've
> decided to migrate to a P4-2.4B, an Intel Desktop motherboard and
> reinstall my OS.  I'll probably dual-boot this time, as Unreal 2 is not
> planning a Linux release.  My feeling is that UT 2003 is stressing the
> AGP bus to the point where it's exposing a weakness in how the Linux
> kernel handles either the AGP bus on a VIA KT333 chipset and/or the
> processor itself.  I believe that because I can run UT 2003 just fine if
> I install Windows on this very same box, and because many of you have
> been able to run UT 2003 just fine on some Linux boxes.

It may be purely physical. I just went through a similar exercise. I had
a dual slot 1 PIII that would occasionally lock up. I suffered with it
for over a year, until it got to the point of being weekly. I finally
deduced that it was hardware-related, because nothing I did could fix
it. To make a long story short, the best I could come up with was that
it was overheating, and causing a partial short to open up completely
when it reached a certain temperature. (Keeping an industrial-strength
floor-standing fan on it kept it running.) But it manifested itself only
when I pounded it, like when gaming or doing a dual mp3 encoding.
I kept the thing running long*er* by disabling AGP. But that just
underscored that it was happening when I had to much going on. Maybe
this is your problem as well.

Even if this is the case, you're still hosed, and need a new mobo. I
took the opportunity to buy an MSI dual K7D Master, and put 1200's in
it. I also got a couple ThermalTake fans which are too tall to be able
to slide the drives by in my Antec case, but overall, it runs between an
almost-frigid 38 at idle to an unimpressive 42 degrees C when playing
UT2003. So far in the past 3 or 4 months, I've never locked it up, and
that's been a huge relief. (But this machine gets rebooted daily so my
wife can use the Windows 2000 installation, and I still run games there
too.) I'd recommend this board to anyone. (I also bought an MSI GF4600
VIVO that's running well, though I haven't played with the video stuff
yet.)

At work I have a Tyan 2466N-based dual 1800+ that plays great as well,
but I just locked it up twice in a row yesterday doing something with
the floppy drive, of all things. I upgraded to the latest RH kernel, and
got my floppy thing done, but that box is still on my watch list. This
one, though, stays in Linux 24x7, and that's the only two lockups I can
remember in about 6 months since I got it.

My point is that you shouldn't have to shy away from AMD if you replace
the guts of your rig. On the other hand, I made it a point to avoid VIA
chipsets when I went shopping for both of my AMD systems. I have a
friend who has both Tyan 2460 and 2466-based systems, and the former has
given him a lot of grief, though I think he worked out those problems to
be specifically caused by a faulty video card. I think both the newer
Tyan board and especially the MSI board are good to go. However, the
performance of a P4... with the new hyperthreading stuff... (But I don't
know if the HT stuff needs kernel support.)

Regards,
dk