UT2004 crashes Gigabyte X800XT with with dynamic lights

[Mika Hirvonen]

My fully patched UT2004 crashes whenever there's too many
dynamic-light-generating objects around. Whenever it crashes, the screen
cycles a while between grey and blue, then turns to black and returns to
BIOS POST. There's no BSOD, just plain blue. Whatever sounds were
playing before the crash continue looping until the the computer is back
on BIOS POST.

The crash occurs whenever there's a few dozen light generating objects
around. The sure way to reproduce the problem is to go to Instant
Action, select any map and add the super berserk mutator. Once in game,
a single volley of Link Gun primary fire (or Hellbender's Shock Core
primary fire, or any other light-generating projectile) is enough to
cause the crash. Normal play is possible with dynamic lights, as long as
I don't wander into highly active areas.

If I turn dynamic lights off, the game works fine. Dynamic lights work
fine in software mode, too. The problem does not occur with other games
(like Doom 3 or World of Warcraft), nor is the computer unstable while
not gaming. Without dynamic lights, even UT2004 is very stable for long
periods of time (10+ hours on CTF-FaceClassic with 32 bots).

Prime95 on stress test mode didn't reveal any errors, nor did AtiTool's
artifact scanning. I've tried underclocking the X800 to 300MHz (both GPU
and memory), but that didn't help.

Due to the water cooling, I doubt that the CPU or the GPU is
overheating. If I leave the computer off overnight, start it up and
directly go to UT2004 while temperatures are still at ~25C, the crash
happens anyway. Normal CPU temperature is at 40C (I run distributed.net
all the time), and GPU is roughly the same (because the same water cools
both the CPU and the GPU). Case temperature is 20C-30C, depending on
ambient temperature. Idle CPU temperature is usually 1-5C above ambient
temperature (which is normally 23C). CPU and GPU temperature on full
load (for example, 3DMark 05 or Prime95 & AtiTool's burn-in test) can
rise to 55C, but drops quickly to normal when the computer becomes idle.
The hard drives are usually at 40C.

The only thing worrying me about the cooling is the ambient GPU
temperature. Zalman's GPU waterblock doesn't cool the memory chips on
X800 at all. Instead, Zalman ships 8 heatsinks for the video ram. Thus,
the GPU environment temperature is 5-9 degrees higher than GPU core
temperature on full load. Still, I don't think that the video RAM is
overheating.

UT2004 settings: 1600x1200 32bit, all details at max. Lowering detail
settings while keeping dynamic lights on didn't help at all.

Rig:
Win2000 SP4 with latest patches
Athlon64 3500+ with Cool n' Quiet on automatic (makes no difference if
switched to high performance mode)
2x 512MB DDR2 (identical chips, using DualChannel)
Abit A8V Deluxe
AOpen H600 case
Gigabyte x800 XT AGP
Enermax EG475AX-VE(G)-SFMA 470W (is this enough?)
Western Digital Caviar SE 200GB SATA150
Maxtor 160GB IDE
Maxtor 80GB IDE
Samsung DVD-R
Soundblaster Audigy 2
bulk 100mbit NIC (the 1000mbit port on A8V is disabled)
A8V's 802.11g PCI card (not in use)

Cooling:
Zalman Reserator 1 (using distilled water, bought it 7 months ago,
haven't touched it since, the reservoir is 4/5 filled with water)
Zalman ZM-GWB1 (ditto)

 

[Michael W. Ryder]

My fully patched UT2004 crashes whenever there's too many
dynamic-light-generating objects around. Whenever it crashes, the screen
cycles a while between grey and blue, then turns to black and returns to
BIOS POST. There's no BSOD, just plain blue. Whatever sounds were
playing before the crash continue looping until the the computer is back
on BIOS POST.

The crash occurs whenever there's a few dozen light generating objects
around. The sure way to reproduce the problem is to go to Instant
Action, select any map and add the super berserk mutator. Once in game,
a single volley of Link Gun primary fire (or Hellbender's Shock Core
primary fire, or any other light-generating projectile) is enough to
cause the crash. Normal play is possible with dynamic lights, as long as
I don't wander into highly active areas.

If I turn dynamic lights off, the game works fine. Dynamic lights work
fine in software mode, too. The problem does not occur with other games
(like Doom 3 or World of Warcraft), nor is the computer unstable while
not gaming. Without dynamic lights, even UT2004 is very stable for long
periods of time (10+ hours on CTF-FaceClassic with 32 bots).

Have you tried logging the power supply voltages while reproducing the
error? Maybe the power supply itself gets too hot and can not provide
stable voltages under the high load.

 

[Tony DiMarzio]

If you've already ruled out other hardware components and the driver version
as potential culprits, then I'd assume that there is some defective silicon
on the VPU, that is only being stressed in such a way as to make the system
unstable, during lighting intensive operations - like what you've described
in UT2004.

If the card's still under warranty then get it RMA'd. If it still doesn't
work then at least you've ruled out the x800.

 

[Mika Hirvonen]

Michael W. Ryder kirjoitti:

Have you tried logging the power supply voltages while reproducing the
error? Maybe the power supply itself gets too hot and can not provide
stable voltages under the high load.

Not according to Motherboard Monitor. Of course, because MBM logs
temperatures and voltages once per second, it might not have any time to
log the voltage if it goes down.

 

[kony]

My fully patched UT2004 crashes whenever there's too many
dynamic-light-generating objects around. Whenever it crashes, the screen
cycles a while between grey and blue, then turns to black and returns to
BIOS POST. There's no BSOD, just plain blue. Whatever sounds were
playing before the crash continue looping until the the computer is back
on BIOS POST.

What operating system?
If Win2k or XP, disable the reboot-on-error setting (Startup
& Recovery, "automatically reboot"), if you hadn't already.

Seems potentially a driver problem... and ATI, well, my
primary gripe with them has always been the drivers/software
side of the products.

The crash occurs whenever there's a few dozen light generating objects
around. The sure way to reproduce the problem is to go to Instant
Action, select any map and add the super berserk mutator. Once in game,
a single volley of Link Gun primary fire (or Hellbender's Shock Core
primary fire, or any other light-generating projectile) is enough to
cause the crash. Normal play is possible with dynamic lights, as long as
I don't wander into highly active areas.

If I turn dynamic lights off, the game works fine. Dynamic lights work
fine in software mode, too. The problem does not occur with other games
(like Doom 3 or World of Warcraft), nor is the computer unstable while
not gaming. Without dynamic lights, even UT2004 is very stable for long
periods of time (10+ hours on CTF-FaceClassic with 32 bots).

Prime95 on stress test mode didn't reveal any errors, nor did AtiTool's
artifact scanning. I've tried underclocking the X800 to 300MHz (both GPU
and memory), but that didn't help.

You might try monitoring power supply voltages, take some
baseline readings then compare to fully loaded gaming with
the problematic scenario recreated.

Due to the water cooling, I doubt that the CPU or the GPU is
overheating.

Well yes and no, it's more likely to have heat problems when
water cooling because most people start thinking they can do
away with fans... which they can't necessarily, there is
more than the CPU or GPU that produce heat and while the
other heat producing parts don't create nearly as much heat,
it's heat meant to be removed by the passive flow past these
parts by an active cooling (fans) nearby. It could be that
your card is ruined already due to capacitors running too
hot. Then again it might not be, isn't certain to happen in
all nor even many if not most cases, this is only a possible
scenario and only you can monitor your part's temps.

If I leave the computer off overnight, start it up and
directly go to UT2004 while temperatures are still at ~25C, the crash
happens anyway. Normal CPU temperature is at 40C (I run distributed.net
all the time), and GPU is roughly the same (because the same water cools
both the CPU and the GPU). Case temperature is 20C-30C, depending on
ambient temperature.

Then you may indeed have cool enough parts, unless this case
temp is inaccurate.

Idle CPU temperature is usually 1-5C above ambient
temperature (which is normally 23C). CPU and GPU temperature on full
load (for example, 3DMark 05 or Prime95 & AtiTool's burn-in test) can
rise to 55C, but drops quickly to normal when the computer becomes idle.
The hard drives are usually at 40C.

The only thing worrying me about the cooling is the ambient GPU
temperature. Zalman's GPU waterblock doesn't cool the memory chips on
X800 at all. Instead, Zalman ships 8 heatsinks for the video ram. Thus,
the GPU environment temperature is 5-9 degrees higher than GPU core
temperature on full load. Still, I don't think that the video RAM is
overheating.

If the heatsinks are making good contact with the chips, and
you touch-test them and they dont' feel overly hot, they're
not (overly hot). Touch-tests are not as accurate as
thermal sensors but on the other hand the accuracy is good
enought to determine ranges that are ok or bear further
scrutiny.

UT2004 settings: 1600x1200 32bit, all details at max. Lowering detail
settings while keeping dynamic lights on didn't help at all.

Rig:
Win2000 SP4 with latest patches
Athlon64 3500+ with Cool n' Quiet on automatic (makes no difference if
switched to high performance mode)
2x 512MB DDR2 (identical chips, using DualChannel)
Abit A8V Deluxe
AOpen H600 case
Gigabyte x800 XT AGP
Enermax EG475AX-VE(G)-SFMA 470W (is this enough?)

If it were some other name-brand I'd say yes... as Enermax
rates for peak output rather than sustained, I'd question it
even though it's probably sufficient.

Western Digital Caviar SE 200GB SATA150
Maxtor 160GB IDE
Maxtor 80GB IDE
Samsung DVD-R
Soundblaster Audigy 2
bulk 100mbit NIC (the 1000mbit port on A8V is disabled)
A8V's 802.11g PCI card (not in use)

Cooling:
Zalman Reserator 1 (using distilled water, bought it 7 months ago,
haven't touched it since, the reservoir is 4/5 filled with water)
Zalman ZM-GWB1 (ditto)

I think your temp readings are off, most of the online
reviews show a Reserator doing little if any better than a
regular fan & 'sink, if even that good when it comes to
higher-performance (heat producing) parts. Regardless, that
doesn't mean they're necessarily overheating. I'd try the
different driver, try the card in another system, then
proceed from there.

 

[Morituri-|-Max]

Seems potentially a driver problem... and ATI, well, my
primary gripe with them has always been the drivers/software
side of the products.

Most people never have problems with their ATI cards drivers or
software........

D'Oh! Couldn't resist!
Kidding... the rest of the post was excellent advice as usual and something
I'll keep in mind if any similar problems crop up.. just out of curiosity,
do you do hardware troubleshooting for a living?

 

[Mika Hirvonen]

kony kirjoitti:

What operating system?
If Win2k or XP, disable the reboot-on-error setting (Startup
& Recovery, "automatically reboot"), if you hadn't already.

I did disable reboot-on-error, but I still don't see a BSOD. The screen
just stays blue. Event viewer does report a crash, but does not specify
any human-readable reason for it. I'll post the full error message when
I get home.

Seems potentially a driver problem... and ATI, well, my
primary gripe with them has always been the drivers/software
side of the products.

I do have both Fedora Core 3 and WinXP 64-bit RC1 on the other drives,
I'll try UT2004 with them today. Also, I've downloaded different
versions of Catalyst drivers, but haven't had any time to test them yet.

You might try monitoring power supply voltages, take some
baseline readings then compare to fully loaded gaming with
the problematic scenario recreated.

I already did, but due to the fact that MBM can log voltages at most
once per second, it probably can't catch a sudden voltage drop.

Well yes and no, it's more likely to have heat problems when
water cooling because most people start thinking they can do
away with fans... which they can't necessarily, there is
more than the CPU or GPU that produce heat and while the
other heat producing parts don't create nearly as much heat,
it's heat meant to be removed by the passive flow past these
parts by an active cooling (fans) nearby. It could be that

The case is a bit crowded near the outer edge of the X800 (water tubes,
IDE cables, power cables), but none of the parts feel hot.

your card is ruined already due to capacitors running too

Is there any way to see whether the capacitors are damaged?

hot. Then again it might not be, isn't certain to happen in
all nor even many if not most cases, this is only a possible
scenario and only you can monitor your part's temps.

Then you may indeed have cool enough parts, unless this case
temp is inaccurate.

It's from memory, I'm not on my computer right now.

If the heatsinks are making good contact with the chips, and
you touch-test them and they dont' feel overly hot, they're
not (overly hot). Touch-tests are not as accurate as

They are at a bit higher temperature than normal body temp, but they
don't feel hot. The heatsinks don't seem loose from the chips, and don't
move when pushed towards the them.

If it were some other name-brand I'd say yes... as Enermax
rates for peak output rather than sustained, I'd question it
even though it's probably sufficient.


I think your temp readings are off, most of the online
reviews show a Reserator doing little if any better than a
regular fan & 'sink, if even that good when it comes to

Yes, if the computer is working at full burn and I keep the windows
closed, the temperatures don't differ that much from the temperatures I
got with stock cooling. The Reserator does keep the case temperature low
and cools the CPU and the GPU quicker than air cooling, though. I keep
my Reserator near an open window, so it probably gets rid of excess heat
quicker than usual.

higher-performance (heat producing) parts. Regardless, that
doesn't mean they're necessarily overheating. I'd try the
different driver, try the card in another system, then
proceed from there.

I'd rather not remove the X800 unless I'm sure that it's faulty, because
draining the system and removing the water cooling takes a while.

 

[Mika Hirvonen]

Tony DiMarzio kirjoitti:

If you've already ruled out other hardware components and the driver version
as potential culprits, then I'd assume that there is some defective silicon

I have done some tests, but I haven't tried the X800 on other computers.
The water cooling is cumbersome to remove, so I'd rather not do that
unless I have to.

I have downloaded different drivers, but I haven't had time to test them
yet. Also, I have Fedora Core 3 and WinXP 64-bit RC1 lying around, so
I'll test UT2004 with them, too.

on the VPU, that is only being stressed in such a way as to make the system
unstable, during lighting intensive operations - like what you've described
in UT2004.

That does sound plausible, but I've yet to find a way to reproduce the
problem in other games than UT2004, so I can't be sure.

If the card's still under warranty then get it RMA'd. If it still doesn't
work then at least you've ruled out the x800.

I think that the warranty expired the moment I removed the stock cooling
and replaced it with Zalman's GPU waterblock.

 

[kony]

I already did, but due to the fact that MBM can log voltages at most
once per second, it probably can't catch a sudden voltage drop.

No, I mean, YOU take readings.
Digital multimeter
I never suggest software readings except for a curious
novelty. You don't have to catch a sudden drop, rather
noting the difference in load causing the voltage to drop.
Before it gets below that threshold, if there is one it's
reaching, it's be the same trend up until that moment.

The case is a bit crowded near the outer edge of the X800 (water tubes,
IDE cables, power cables), but none of the parts feel hot.


Is there any way to see whether the capacitors are damaged?

Usually they'd look it, domed vented tops, residue on top or
bottom. Discoloration around the voltage regulation
circuits is also a strong sign of it.

I'd rather not remove the X800 unless I'm sure that it's faulty, because
draining the system and removing the water cooling takes a while.

Well... it is the surest way to isolate the card from rest
of system...

 

[kony]

Most people never have problems with their ATI cards drivers or
software........

D'Oh! Couldn't resist!
Kidding... the rest of the post was excellent advice as usual and something
I'll keep in mind if any similar problems crop up.. just out of curiosity,
do you do hardware troubleshooting for a living?

I don't even have a computer!
.... but I did stay in a Holiday Inn Express once.



Nevermind me, I always go wacko in the Spring.

 

[Mika Hirvonen]

Upgrading to Catalyst 5.3 solved the problem, thanks all.

 

[Tony DiMarzio]

you serious? lol.

you really should have tried the most recent driver in the first place.

 

[NuQ]

Upgrading to Catalyst 5.3 solved the problem, thanks all.

With most gaming video card problems, you should try the latest or different
drivers as the very first step to solving the problem. It's almost always a
driver issue.