DescentBB

I'm getting random crashes and reboots while playing games. The reboots occur while playing Doom 3 and I'm also getting crashes from Jedi Knight II. Both games have the latest patches. Here are the specs:

AMD Athlon 64 3000 (@2400mhz)
Chaintech VNF3-250 Mainboard
ATI Radeon X800 Pro (at stock; latest Omega Drivers; fastwrites on)
Windows XP Pro SP2

Help?

does it reboot when not playing games? If not, I'd suspect it's either your video card is overheating (fan might be dead) or your PS is dying or might be your RAM. If you have 2 sticks of RAM, remove one and see how it runs.

First thing i'd try is dropping the CPU speed back to stock, or at least reducing the amount of the overclock.

The source of the problem may be the fact that Fast Writes are enabled. The performance gain is miniscule when enabled and it only contributes to instability. Try turning that off first and see if that helped.

No, it does not reboot when not playing games. So it is probably something with the video card. The fan works fine (that was the first thing I checked). I did try turning fast writes off, but the problem still occurred (although that was several driver versions ago). When I installed the Omega drivers, it actually recommended that I turn them on because I had an A64. Dunno what that is all about, but anyway... I'll run memtest tonight to make sure that the memory isn't faulty.

Probably cockroaches in the PSU.

One thing to consider is you are running an ATI video card on an nvidia platform. While usually this isn't a problem, the software for the two sometimes don't get along very well, and it is also known that nvidia video cards work better on the nforce platform than do ati.

Make sure your nforce drivers are up to date also.

The PSU is an Antec 380W. Should I pull that one and try testing it with another one?

[edit] nforce drivers are up to date as well.

also he needs to disable the Auto Reboot feature of XP

then he can find out what is causing the issue from the BSOD screen

I forgot to mention the RAM in the specs. Its 2x512MB PC4000 from Adata (500mhz). I had it running at 480mhz. I dropped it down to 470mhz (CPU is now 2350, and I dropped the CPU voltage down a notch to 1.525 volts) and it seems to be stable now. It seems kinda funny to be having problems like that after it was stable so long at the other speed. Especially since the RAM isn't even running at its stock speed.

If you OC anything, over time it will fail.

Over time == over 10 years... Stable is stable, CPU cores eventually wear out over time even if you underclock it, deep down they are still a collection of switches, switches wear out. Overclocking while still stable has virtually zero impact on lifespan of these parts.

You have to keep in mind that games are far more stressful on the CPU then just sitting out in windows waiting for the user. Stable in windows and stable while running a game are two completely different things.

My guess is you can get it stable at 2.4 GHz just by getting better cooling, a couple C cooler and it should run. So far I have not had any chip fail that I have overclocked to a constant stable level during its entire life. The oldest I have that is still running is a P3/700 that I run a 933 MHz. That chip can run at 1001 MHz perfectly fine, but some of the RAM that we got for it can't handle the 143 MHz bus that chip requires for 1001 MHz. That CPU is from mid 2000, been overclocked since day one, still running error free over 4 years later. Older CPUs that I overclocked I quit using, but they all likely still work if I found them and powered them up.

The machine has been stable at that speed for 6 months now. Its been through hours and hours of gaming and a couple of LANs. The reboots are a recent problem. Cooling shouldn't be an issue. I've got a Thermalright SLK948U model cooler on it, complete with Artic Silver 5 and a nice fan. The CPU Temp reading hasn't risen beyond 52° C.

I'm going to ask the obvious, just because it hasn't been asked:

Virus/Spyware Scans clean?

Cooling shouldn't be an issue

Personally, I would open the case and blow a big fan into it to make sure cooling isn't the issue, then move on to other suspects if it still crashes despite overkill cooling methods. Motherboard Monitor is a good util for checking how the board and PSU are handling the load as well. My system crashes if the PSU overheats long before the CPU or board indicates it's getting into danger zone. If you've added components the PSU will be under additional load from the days when everything was fine. Just watching the CPU temp is only part of the cooling story.

Check for dust buildup also.

I have a filter in the front the collects the dust, and I also blow out any that does get in quite regularly. And its completely clean of any spyware or viruses. I'll run MBM for a bit and monitor the PSU.

[Edit] OK, I'm lost as to how to set up a sensor to read the PSU temp in MBM. Help?

My guess is it's a different sensor for different motherboards so you'll just have to trial and error it. IIRC you also have to wait a moment when choosing the various sensors until it reads the first time so you may think you've selected the wrong sensor at first but give it a minute. On my system (a Shuttle box) the right sensor for the PSU is ITE8712F-3 in the pull-dwon menu in MBM setup but I have no idea if that is a universal sort of thing or even if your PSU can be monitored. If it can't try the overkill boxfan method of cooling to see if it stops the problem. No sense getting tangled up in figuring out sensors if overheating isn't the issue, although your description sounds like one to me.

OK, that looks like the right one. Says 127° C. How hot are these things supposed to get? That seems too hot to me.

Hmm, that seems waaay too high. Mine crashes at around 75C. I think different sensors have different measurement methods (or something like that). Might want to try another sensor because my guess is there's no way the PSU could function at that temp, keeping in mind my little box has a relatively small PSU so maybe it is prone to lower temp crashes than the one you've got. I think mine is like 200 watts so that is my frame of reference. Others know better than I.

Funny thing is, SiSoft Sandra reported the same temperature just after I got it. I found it high, but it hadn't given me any problems, so I left it alone. Maybe its the wrong sensor... /me goes off to get box fan

[Edit]
OK, of the options that are listed for the sensor, I get the following temps:
The case temp sensor is "LM84-9 (Local)"

My guess is your PSU is not outputting temp info and you're just getting top of the range reading. That aside, could you still intentionally crash your machine while the box fan was blowing in it? If so, unlikely a heat issue and you need to look elsewhere. If not, something (my guess is either the PSU or the video card) needs more cooling help.

FWIW, my sensors settings are:

MB: ITE8712F-1
CPU: ITE8712F-2 Diode
POWER: ITE8712F-3

This was based on booting into the BIOS to read what that monitor said and booting back into Windows and matching the temps from the BIOS to what each MBM sensor was reading.

hmm there is a easy burn in test for the cpu you could try. start your temp-monitor, then start the windows-build in calculator. switch the calculator to scientific-mode.

now calculated a factorial of a high number, lets say 800000. (its the n!) button. this will keep you cpu busy at 100% for hours. the caculator will ask you something like "this will take a long time, continue?" just press nothing, it will keep calculating the number in backround at 100%. now let it do the job for 3 or 4 minutes and watch how the cpu-temp rises. if it goes over 75c you have a cpu-fan prob. if its stays in medium range you know at least that cpu /fan is not the problem.

edit: i forgot 1, download this atitool:

http://www.techpowerup.com/atitool/

it was designed by ati for overclockers, and can find artifacts at the rendering-process, strong sign for memory or gpu/probs. more important, on newer ati-cards it can read and display the gpu-core temperatures.

DigiJo wrote:
edit: i forgot 1, download this atitool:

http://www.techpowerup.com/atitool/

it was designed by ati for overclockers, and can find artifacts at the rendering-process, strong sign for memory or gpu/probs. more important, on newer ati-cards it can read and display the gpu-core temperatures.

do they have something like this for NVidia cards?

Google "Artifact Tester". Run both the software and the DirectX tests, since I seem to get artifacts on software first.

I checked the video core temperature right after I crashed out of JK2. 60° C. That doesn't seem overly hot for a video card. Suggestions?

Did you add anything to the system recently (hardware wise)? If not, you might want to try unplugging something non-critical and see if it stabilizes. You might be pulling more current than your power supply can handle. (or really close to it)

google for memtest86 and run a full diag twice

I have added a TV tuner card, but that was after the problem began. Also, Memtest ran fine.

You never said whether the boxfan test cured the problem.

The box fan did seem to solve the problem. However, I'm running with some different drivers now, and haven't had an issue yet.