Help me troubleshoot

[themisfit610]

Hi all,

So I recently built a new system (well, did a major upgrade) for a client.

He now has:

Giant Thermaltake full tower with a plethora of fans
Antec 550W PSU from a couple years ago (previously drove a 3500+ and SLI'd 7800GTXs)
Gigabyte P43 motherboard
E5200 2.5 GHz 2MB
8GB of DDR2 800 - mixed sticks from OCZ and A-Data
GTX 260 (192sp)
X-Fi Fatal1ty
160GB HD
DVDRW
Vista x64

He uses this system for two things only: Maya 2008, and Battlefield 2.

He's been using lots of Maya recently, and apparently when he's been using it for a few hours, the system will get a BSOD, then reboot, but shut off before it can get into windows. Then it powers up, shuts off, powers up, shuts off etc... all in an endless loop.

He has to unplug it, let it sit for a minute or so, then power it back up.

So, I've got the system on my test bench, and am trying to stress test it. I've never actually seen his problem.

First, I started with a full memory test, so I ran Memtest all night, and it checked out perfectly fine.

Second, I loaded up some stress testing apps and did this:

1) Endless loop of furmark at 1024x768 windowed mode with 32xAA to heat up the GPU
2) Endless loop of 1080p x264 encoding, easily loading the CPU to 100%. Okay, not endless, but with maxed settings so it gets about .2fps, so 8 days to do a full movie
3) Coretemp running

With the GPU at 100%, after a full night of furmark, it's still running at 85c with no problems.

The CPU is still pegged at 100%, with temps between 49c and 55c on both cores. No problems!

This system seems totally, rock solid stable. I can't figure it out!!

Maybe it's a problem with Maya specifically? But I don't see it causing a BSOD!

Help!

~MiSfit

 

[Operandi]

Man.... your in the same boot as myself (see here).

Try running Orthos to stress the CPU and RAM.

Maya might also be utilizing parts of the GPU that games might not leading games to run fine but Maya to crash. I once had a video card that would crash under all D3D games but run anything OpenGL endlessly, even the same game that would crash under D3D would run fine OpenGL. I swapped out the card for an identical unit and the problem was gone.

Otherwise the power up and shutting off sounds like a PSU issue, some of the older Antecs are not particularly great.

 

[themisfit610]

Yeah, I'd considered the PSU, but it was previously driving 2x 7800GTX cards, which are real power hungry beasts! It's SLi certified and everything.

x264 should stress the hell out of the CPU, but I will definitely try Orthos.

I have established that heat on the CPU isn't a problem, but if Orthos shows errors, what can I conclude?

 

[Operandi]

Originally posted by: themisfit610I have established that heat on the CPU isn't a problem, but if Orthos shows errors, what can I conclude?

It depends, Orthos has a couple of different tests you can run. Some stress the CPU others hit the RAM harder, it will be pretty self explanatory when you fire it up.

 

[KGB]

Originally posted by: themisfit610
Yeah, I'd considered the PSU, but it was previously driving 2x 7800GTX cards, which are real power hungry beasts! It's SLi certified and everything.

x264 should stress the hell out of the CPU, but I will definitely try Orthos.

I have established that heat on the CPU isn't a problem, but if Orthos shows errors, what can I conclude?

Maya isn't just any-old app. What version is he running?

It looks like AutoDesk only supports workstation cards (Quadro or FireGL).
Do you see any incompatability here? :Q

 

[themisfit610]

Oh sure - I know maya prefers having a Quadro, but he has several other systems with consumer GPUs (both mac and PC) that run Maya without a hitch. Sometimes you run into little graphical glitches because the drivers for gaming cards can take shortcuts in the name of performance, but it should never BSOD!

He's running Maya 2008 on all systems and this is the only one that's had any issues. There's even a machine with SLI'd 6800 Ultras, a 3500+ Athlon, and Vista x64, and it runs well. A bit slow, but well.

~MiSfit

 

[KGB]

No need to get into a hissy-fit. I'm just saying once you go outside of the supported hardware, all bets are off. Believe me, I'm an old hand at Unigraphics, SDRC, Catia, Solid Works, Pro-E and AuotCAD. Just because it work on other non-supported hardware doesn't mean that it isn't the issue here. I've seen stranger shit.

Now, what is getting logged into Event Viewer?

 

[themisfit610]

LOL you're right of course.

It looks like there are some display driver errors.

Also, Orthos was failing out (instantly), so I did a few things:

1) Matched sticks of RAM with some extra RAM I had lying around. Now instead of OCZ and A-Dat memory, I have all matching DDR2-800 sticks from Avant at 1.8v
2) Updated to the latest BIOS from Gigabyte

So far Orthos is running strong.

I'm going to make sure the CPU/MB is stable, and then see if I can get maya to crash. If it does, I will try installing the maya 2008 SP1 update, and also possibly change drivers. I'll probably throw a Crysis demo loop in there as well...

Sigh...

~MiSfit

 

[KGB]

What diplay driver errors are you getting?

Have you tried rolling back to an older version?

Is there a soft-mod available for that GTX card?

 

[themisfit610]

When I get home I'll see if I can get the exact error messages. I'm not so hot with the error logs.

The soft-mod is the first thing I looked into, but unfortunately you can't do a GTX260 AFAIK. nVidia did just release 2 GT200 powered Quadros, and maybe eventually, but not for now.

I originally had the client roll back to the 178 drivers (from the 180 drivers), but he's had the same problem.

The system just completed an overnight Orthos / Crysis demo loop run with no problems. It seems like general usage is totally stable. Tonight, I'm going to see if I can get Maya to crash...

~MiSfit

 

[KGB]

Good results.

We'll talk this evening.

 

[KGB]

Was out last evening.

Hello... 1,2,3 anyone home?

 

[themisfit610]

LOL I was out last evening also.

Here's the score:

The system has been totally stable for the last several days. I ran another overnight loop of memtest, and there were no problems. It's currently chugging away doing a large batch render in maya. We'll see how that goes. I also spent a good 10 minutes playing with a scene in Maya, and didn't encounter any issues.

The error log shows (for example) during a 2 hour period 5 crashes of nvlddmkm. Specifically, it says "display driver nvlddmkm stopped responding and has successfully recovered. These errors were all several weeks ago when the client was using the system frequently.

The reliability and performance monitor doesn't reflect any of these errors, showing perfect 10 scores on these dates.

Strange.

~MiSfit

 

[themisfit610]

Any more thoughts?

I loaded up Far Cry 2 and Call of Duty, and they were both silky smooth

I also ran a loop of an Animation in Maya all night using real-time rendering, and it was fine.

Seems good to me!!

~MiSfit

 

[Operandi]

You found that Othos was failing earlier and you addressed that with different RAM correct?

If that was the problem and you fixed it I would say your done.

 

[themisfit610]

Yes, I did switch the RAM and update the BIOS. It seems to be rock-solid. I just keep having this nagging feeling that I'm going to drop the system off, and 4 hours later he's going to call and say "it's doing it again".

Sigh....

~MiSfit

 

[Raduque]

Is he running dual monitors?


Edit: Also have him disable Aero.

 

[themisfit610]

Yes he's running dual monitors. It's an interesting idea to disable Aero, I might do that.

~MiSfit