BFG 7900GS OC locks up my PC

G

GnatGoSplat

n00b
Joined
Sep 12, 2007
Messages
62
I just got a BFG 7900GS OC card from Frys. It is stock overclocked to 540MHz (normal is 450MHz). My machine intermittently locks up when I play videos!

My hardware:
Gigabyte GA-MA69G-S3H motherboard (AMD 690G chipset).
AMD X2 4400+ 2.3GHz (no OC).
HP 4GB DDR2-667 oc'd to 800 5-5-5-15 (Micron D9HNL).
Ultra X-Finity v2 600W PSU
Seagate 320GB SATA HD
Running Vista Ultimate x86 32-bit
Using latest nVidia drivers v162.22.

With the new 7900GS, my machine locks up when I play any videos. I tried playing 1080i HDTV (DVR-MS file) in Media Center. It locks up. I tried WMV and MPG in Media Player. Locks up. It can take anywhere from 5 to 40 minutes to lock up, but always within an hour. When it locks up, the image is frozen on screen with no visible corruption. About a half second of the sound is repeating over an over. I have to press and hold the power button to shut off the computer. The desktop and 3D games are fine!!! I only have problems with videos!

Everything was fine before when I had a GeForce 7300GS video card and also when I was using the onboard X1250 IGP.

Here is what I have tried:
1. It was using FFDShow decoder, so I uninstalled it. Then it was using Cyberlink PowerDVD decoder. Still locks up. So I uninstalled that. Now it's using Microsoft Vista MPEG2 decoders. Still locks up. Must not be decoders.
2. I ran Gigabyte EasyTune5 health monitor. All voltages looked fine. The lowest the 12V rail ever got to is 11.98V. I don't think it's the PSU. When it locks up, the screen freezes and at the moment it froze, I can see the voltages looked fine.
3. Suspecting maybe the northbridge was overheating since the larger card might block more air than the smaller 7300GS, I monitored NB temps which got up to 49C when it used to be 44C, so larger card did block a little air. I put a small fan pointing at the NB heatsink and it stayed 38C the whole time. Still locked up. Nope, guess it's overheating.
4. I know the CPU wasn't overheating, it peaked at about 47C. Not the CPU.
5. RAM has been fine for months at DDR2-800, but maybe it went bad. So I ran MEMTEST86 overnight (and still running). All passes, no fails. Not the RAM.
6. Suspecting maybe the video card is bad, I installed nTune and downclocked the card from 540MHz to 450MHz, then played a video. Still locked up!
7. I ran nTune's Stability stress test and checked everything and let it run for an hour. Everything was fine! I then unchecked everything except PCIe and GPU and ran it for another hour. They tested fine! GPU temp never exceeded 60C.
8. I have a 2nd machine (running Intel C2D) that has this exact same video card. I switch the cards, putting the newer one into the Intel box and the older one in the problem machine. The Intel box is fine, plays videos 3-hours straight with no problems. After about 40-minutes of playing a video, this machine locks up with the older, but identical 7900GS card. So that proves the video card is not defective.

I have ruled out decoder, NB, CPU, RAM, PSU, and video card. Both machines are running Vista Ultimate 32-bit and both are using the same 162.22 drivers so I don't think it's a driver issue.
One thing I forgot to do is pull an old dialup modem, which I will try when I get home from work, but I'm not expecting that to be the problem. Other than that, I don't know what else to try. Any ideas?
 
What are the temps of the video card? Have you tried removing and reinstalling ALL video drivers?
 
GnatGoSplat said:
I have ruled out RAM....[
Click to expand...

By running Memtest? Nah. I've had Memtest pass bad RAM before.

Do the process-of-elimination thing and pull half the RAM. If it still locks, pull that half out and put the other half in (I'm guessing 4 1GB sticks, right? So do 2 and 2?)

GnatGoSplat said:
I have ruled out PSU...
Click to expand...

By monitoring voltages with EasyTune5? Nah. Software is NOT the way to monitor voltages. They only monitor the voltage at one chip, on one point of the board. Resisance isn't taken into consideration, nor is the chip "fast" enough to record sudden drops. There's a reason why there are $30 Radio Shack DMM's and $400 Flukes and on board monitoring isn't even a $30 Radio Shack. Let me give you a scenario: load of the GPU causes +12V to drop to around +10V. Machine locks. At which point, voltage is going to bounce back to a no-load voltage because the load is essentially removed when the PC locks. How's software going to catch that? Heck.. how is a DMM without data logging going to catch that? It's not.

A 7900GS uses more juice than a 7300GS or X1250 IGP so a PSU problem isn't out of the question. An X-Fi 600W should be able to handle it, but a bad PSU may exhibit enough of a drop on the +12V rail to cause the machine to lock. Swap it with the PSU in the Intel rig temporarily.
 
RoBo: I did more troubleshooting yesterday and GPU temps were 55C when it locked up. I did try uninstalling and reinstalling the driver through Vista's equivalent of Add/Remove programs.

jonnyGURU: I'm disappointed that Memtest has passed bad RAM before. I got home and as suspected, total of 16hrs and Memtest found no problems. I took your suggestion and pulled half the RAM (originally 4 sticks of 1GB, now 2 sticks). Played a video. 1-hour goes by and NO LOCKUP. Thinking maybe it was bad RAM after all, I switch the RAM, but I put the suspect bad RAM back into the other 2 sockets, play the video for 1-hour. NO LOCKUP! Ok, very weird, I guess the RAM itself and the sockets are fine. I put 4 sticks back in it and try again thinking maybe I just didn't have it seated quite right from the beginning, or something... nope, locks up quick this time, probably not even 5-minutes. On a hunch, I slow it down to DDR2-667 (I had it oc'd to 800). Played video. 1-hr later, no lockup. So strange, it doesn't mind an overclock in 2GB configs, but 4GB it doesn't like. So I'm suspecting maybe VDIMM may be dropping with the additional load of extra RAM (and somehow adding the 7900GS affected this) so I tried raising VDIMM by +0.05V in BIOS. WITH the overclock to 800, it was able to play video for 2+ hours. So I thought I had it fixed, but to test, I let it play videos on repeat overnight. When I woke up this morning, it was locked up. I raised VDIMM to +0.15V and currently have it looping videos. I'll find out what that does when I get home from work tonight. If that doesn't work, I'll try the RAM non-overclocked looping vids overnight and see what happens.

You're right about software monitoring. It's common sense now that I think of it, at the moment the voltage drops, the machine hangs so the display would only show the moment before drop anyway. I wasn't thinking! Anyway, I do happen to have a Tenma DMM with monitoring. I hooked it up and let it record min/max voltages and checked it after the machine locked up. Min voltage ever recorded was 11.99V. I don't know what the sampling rate of the monitoring is, so it's possible that's not really accurate either. I also have a PC-based 100MHz dual-channel DSO I could use instead... I wonder though, it seems my trouble may be with VDIMM, does that come from the 12V rail or one of the other rails? I haven't tried swapping the PSUs yet because it's a rat's nest of cabling behind both machines and one of them I can't swap the PSU on without physically removing it from its cubby and also unplugging all the networking stuff sitting on it. Not to mention the rat's nest of PSU cabling inside (I wish they were modular). I've been saving that task as a very last resort.
 
Ok. The problem with the RAM is actually quite common when using that many sticks. The motherboard simply can't buffer to both banks fast enough. If you have to have four gigs of RAM, then a slower speed is necessary. In my own machine, I've found four sticks of Corsair Dominator is faster than four sticks of cheap Kingston, but the Corsair locks up. Two sticks of Corsair runs fine, and happens to be faster over all. So I stick with 2GB of RAM in two slots. I don't do a lot of productivity stuff that requires that much RAM (Dreamweaver and Fireworks at the most.) Hadn't noticed much different between 2GB and 4GB and, yes, I'm using XP-64 so I can actually see all of my RAM. :)
 
Ah, thanks jonnyGURU, that makes sense. I have no problem running the RAM a little slower. I did have 2GB before, but I also run VMware. With VMware running at the same time as playing games, I noticed around 80% RAM being used so I thought it would be a good idea to add the additional 2GB which did seem to boost performance a bit. An even 3GB would have been plenty (it only sees 3.3GB anyway), but I decided to add the additional RAM in a pair for dual channel and I happened to have 2 sticks sitting around anyway.

Since the additional RAM did seem to speed things up, if the machine is locked up when I get home tonight, I'll just try slowing it down and see if it's stable that way.
 
Yeah. Knowing that you're using VMWare... stick with the 4GB. ;) So do try the slower timings. I think overall you'll be happier.
 
Thanks johnnyGURU, you were right about the RAM. Slowing it down to 667 seems to have completely fixed the problem. 17-hours straight playing videos and no problems. The machine was also up all weekend long with not so much as a glitch.
 
You must log in or register to reply here.
Back
Top