Inexplicable video artifacts and crashes

[hugoc]

Hello all,

Maybe this isn't the most appropriate place to ask this question, but I've found this forum to be full of knowledgeable people so I'll go ahead anyway. Apologies for the long post, I have tried to include as much info and detail as possible to help you to help me. :)

My system (don't laugh, I can't afford to upgrade, it was once proud and mighty):

ABit BP6 with 2x Celeron 366@550 (stable, for 3 years), RU BIOS
Creative Annihilator Pro (GeForce DDR)
Single stick of 256MB PC100 RAM
CMI8738 sound
Adaptec 2904 SCSI w/ Pioneer slot-loading CD-ROM
2x Seagate EIDE drives on UDMA33 channels (Highpoint controller disabled as it's a POS)
Goldstar CD-RW
D-Link DE528 NIC
Windows 2000 SP4, DirectX 9.0b

Here's the issue: the graphics card seems to have a problem with intensive 3D games. Quake 3 will run for 10-15 minutes and then crash, sometimes bringing the whole system with it. Total Annihilation: Kingdoms will run for 20-30 minutes in hardware-3D mode and then the graphics will become corrupted. STOP errors showed a memory allocation fault in nv4.disp (i.e. the graphics drivers).

Because of this I assumed it was a graphics problem. My GPU cooling has been upgraded, I pulled off the poxy little stock fan and attached a larger heatsink designed for passive cooling an NVidia TNT2 card, and stuck a 40mm fan on top of that with as little thermal grease between the two as I could get away with (I spread a thin layer on the GPU slug, and then ground down with the heatsink until I could hear the heatsink scraping the slug). I don't have a thermal probe to check the temperature, however, I'm sure that my new HSF attached with thermal goop probably performs a lot better than the stock cooling superglued on.

I upgraded the graphics drivers to the Detonator 53.03 with no improvement. I suspected an overheating problem, however, 3dMark 2001SE can run in a loop for hours without any artifacts or crashes, however, if I wait for TA: Kingdoms to corrupt the graphics and then quit it, 3DMark will have similarly corrupted graphics afterward.

I have experimented with overclocking the graphics card in the past. It ran just fine at higher clock speeds (at least, with the software that ran without problems before), however, I didn't notice any real improvement in speed so I reset the clock speeds and left them there.

I double-checked that my CPUs weren't overheating by running stress tests, found no problem. My CPU temps are 18-19C idle and about 28C under load. Case temp is around 3 degrees higher than that.

I used to be running Windows XP (I downgraded again since 2K is a little more responsive on this system). I never tested Kingdoms running XP, however, Quake 3 showed similar problems with crashing in XP as it does in 2K. However, it would run Serious Sam 2 for hours with absolutely no problems.

Another odd occurrence is that occasionally the screen in Win2k will go black, as though the monitor had gone into power-saving mode, for a split second and then come back on. Sometimes when it does this, there will be some slight artifacts with the mouse pointer, a scrollbar or other single GUI item. XP never did this.

I used to run this system without the case on, but I recently redesigned my airflow/fan arrangement and put the case back on. The CPU and case temperatures are no higher with the case on than when it was off, so I believe my new case cooling scheme is working well and is not the cause of any instabilities.

It could be:

a) problems with the games (I have patched them, no improvement)
b) problems with the drivers (I have upgraded them, same)
c) overheating of the GPU (unlikely as it can run 3DMark or Serious Sam 2 forever)
d) overheating of the CPU (unlikely as it can run burn-ins forever)
e) problems with Windows (I keep it as up-to-date as possible, and Quake3 exhibited the same problems in XP and 2K).

So, I can't really explain this problem. Does anyone have any thoughts or directions I could investigate further in?

[RCtruckguy]

i hate to tell you this but maby your system is too old/slow. earlier versions of the games stressed it less so they could run fine but now they dont run as well. i used to be able to play CS 1.2 on my dialup, ragefury pro and p3 system and now it has new parts (see sig) and cable and it works much better. when version 1.5 came out i started to lag badly video wise.

well that was the case with me and CS so i upgraded, too bad you cant do the same

~Alex

[hugoc]

Alex,

Thanks for the reply, but that explanation doesn't seem to fit. Here's why I think that:

1) The problems occurred on freshly installed versions and patched versions of the games

2) The patched update of TAK has lower system requirements than the off-the-CD install

3) Both Q3 and TAK were released when my spec was top-of-the-line

4) It doesn't fit that the graphics corruption in TAK would still be there in 3DMark if 3DMark is run after TAK, but if TAK isn't run first, 3DMark works fine

5) If it were just the case that my system wasn't up to scratch, you'd expect overall poor video performance, as you saw in CS, but what I'm seeing is perfectly smooth video until a sudden crash e.g. Q3A runs like silk in 1024x768x32 for 10 minutes and then the system hangs. That's not really a performance issue, it's a stability problem, but I can't track down what's causing it.

[RCtruckguy]

1) reason is why i thought that, some time new versions no matter what the box says push your system more.

another reason i though of is maby your video card is getting hot after a while. maby TAK heats up the vid card alot more so it gets hot and artifacts, i personally have a 120mm fan aligned with my video card and cpu, so i know it gets good cooling.

your cpu temps are great but maby your video card is in a hot spot. i sudgest cutting a cardboard copy of your case side, then cut out the normal holes that you use now, then put a 92 or 120mm fan aligned with the video card. test that and see if it helps if it dosnt your only out a few $ for a 120mm fan and you can put your old case side back on.

GL~Alex

[hugoc]

That was what I initially thought too. However, I don't think it's the problem either. I can run 3DMark or Serious Sam II for hours without any glitches or hiccups.

I know SSII used to overheat my card, in the big outdoor areas the gfx would slow and sometimes crash, but after I upgraded the GPU cooler this problem went away. Now the game runs without any problems.

One would think that if a seriously taxing game like SSII could run for hours, a game that doesn't push the hardware very much like TAK should be fine too.

At the moment, I'm inclined to write this off as bugs in the games as I can't track any kind of problem in my system. My last idea is to put Win98 back on this system as a dual-boot and see if it's a Win2k incompatibility.

[hibner]

Maybe your vid card is just going bad, thats how mine went. Try to borrow someones video card to see if thats the problem.

One other thing, If you have anything overclocked, Set everything to default and try again.

[Kenshiro]

try running memtest. The problem could be your memory.

[hugoc]

I ran memtest, it found no problems.

Hibner,

I thought that if the videocard was on the way out, a few hours running 3DMark and Serious Sam II would have shown it - unless you can think of any better GPU stress-testers (seriously)?

Running at stock speeds doesn't make it stable, just slow. :) Moreover, the system can run all the other stress tests I mentioned (SSII, 3DMark, SETI@Home) for a long, long time while overclocked without any problems at all.

[hugoc]

I just noticed something else: when I run dxdiag and test Direct3D, it shows a spinning cube. When it runs the tests for DirectX 7 and 8 modes, there are artefacts, when it runs the test for DirectX 9, it's perfect. Could this be a DirectX problem?