Random Crashes - Memtest 86 definitive?

kooky_guy · 8 Jun 2006 at 11:44

Hi Guys,

I'm experiencing random crashes on my machine and am trying to isolate the cause. I've run Memtest86 for 9 hours with no errors. Is this sufficient to completely discount the RAM?

I've experienced random reboots in Windows and application crashes (mainly in BF2) which Windows suggests is down to a hardware fault (very helpfully narrows it down to Memory, CPU, motherboard, PSU or graphics card). I've discounted the graphics card as I've got two and it happens with both individually.

Specs are:
Ausus A8n-SLI Deluxe
AMD Winchester 3000+
1Gb Corsair TwinX 1024-4400C25PT
Hitachi Deskstar T7k250 SATA2 hard drive
Various IDE hard drives
Pioneer DVDRW & Yamaha CDRW
Enermax Noisetaker 600W PSU
2 x XFX 6600GT currently in SLI config

Regards

Guy

Jay_t · 8 Jun 2006 at 12:02

Hmmm, not sure if you can rule the ram out just yet. Is your motherboard running the latest bios?

t31os · 8 Jun 2006 at 12:06

Have you been doing anything more stressful on the system when it crashes?... such as running games.

Might be worth unplugging non-essential devices, such as extra HDD's and all but 1 CD/DVD, and see how the system stability fairs, if alls well with less devices it'll be the PSU, unless of course 1 of those drives you unplugged is to blame, but its unlikely.

Also have you tested the system with idividual sticks of memory to see if that helps, its rare both sticks are duff if memory is the problem, possible but less likely.

kooky_guy · 8 Jun 2006 at 12:16

Jay_t said:
Hmmm, not sure if you can rule the ram out just yet. Is your motherboard running the latest bios?

Yup. I checked. it's on 1016. To be honest I've had trouble getting the RAM to work properly as the mboard seems to autodetect it incorrectly, but I got it to run at the correct timings by upping the voltage to 1.8V. Wish I had just gone for more standard RAM now, but I hoped to be able to do some overclocking and this was supposed to give me plenty of headroom.

All drivers are the latest versions incl chipset.

I've had constant problems getting SLI to work reliably and have just given up and bought a 7900GTX instead, but there doesn't seem to be any point in installing it until I've got this crashing problem sorted out. It's only started actually crashing recently - last couple of weeks. Prior to that it would just give extremely inconsistent performance with SLI enabled.

Not sure if that helps.

Guy

kooky_guy · 8 Jun 2006 at 12:25

t31os said:
Have you been doing anything more stressful on the system when it crashes?... such as running games.

Might be worth unplugging non-essential devices, such as extra HDD's and all but 1 CD/DVD, and see how the system stability fairs, if alls well with less devices it'll be the PSU, unless of course 1 of those drives you unplugged is to blame, but its unlikely.

Also have you tested the system with idividual sticks of memory to see if that helps, its rare both sticks are duff if memory is the problem, possible but less likely.

Well BF2 will always either crash out or reboot the machine (almost immediately), but I have also just been surfing the web when it's rebooted. I think I'll try running 3DMark and see if that has the same problems.

I'll try your suggestions with the single sticks, and unplugging a few things - I was hoping memtest would have discounted the memory though.

Not sure if this is relevant, but the motherboard (in Windows anyway) incorrectly reports all the voltages. It claims they are quite a bit lower than they actually are when measured with a voltmeter and I was originally getting low power warnings from one of the graphics cards. This was RMA'd though and the replacement has appeared fine. (although the performance in SLI has been extremely erratic).

Guy

kooky_guy · 9 Jun 2006 at 10:14

Only got a few minutes to tinker with it last night, but I disabled the restart on fatal error thing and the BSOD message mentioned Nv4_disp.dll, although it didn't mention the continuous loop that I've read about.

It definitely crashes when it's doing something graphics intensive. 3DMark05 bombed out to the desktop after about 10 seconds the first time and blue screened when I tried it again after about the same amount of time.

Does it still look like a hardware problem? I've done a bit of quick research and it seems that it can be caused by the memory timings being too aggressive, but I'm not quite sure what that means or what to do about it.

The timings I've put in are those that the memory should run at!

Guy

Dutch Guy · 9 Jun 2006 at 10:41

Overheating CPU, Northbridge or memory?

Is anything overclocked?

kooky_guy · 9 Jun 2006 at 11:18

Dutch Guy said:
Overheating CPU, Northbridge or memory?

Is anything overclocked?

I did wonder about the Northbridge overheating as the fan has a bad reputation on these motherboards, but it seemed to be running ok. CPU temp is fine. The case has masses of fans so I doubt the memory is overheating.

The only thing overclocked is the graphics card and that's only a mild tweak using coolbits. I've never got round to trying to overclock the CPU.

The first crash occurred within 5 minutes of turning the machine on so it would have to be a pretty fundamental overheat!

The other half's going shopping tomorrow so I should have a chance to try some more extensive research. I've borrowed some more memory to try.

Guy

pcoltrane · 9 Jun 2006 at 16:28

do you have the nvidia firewall installed? or the nvidia ide driver?

kooky_guy · 9 Jun 2006 at 16:37

marc mercer said:
do you have the nvidia firewall installed? or the nvidia ide driver?

I did have the firewall installed, but it caused problems so I uninstalled it. Likewise with the IDE driver.

So, neither of them are installed now.

Guy

t31os · 9 Jun 2006 at 17:19

Just to add to the thread, BF2 crashing happens on lots of systems, its generally related to the game and not the hardware its run on, though some cases there are exceptions. Just wanted to clarify before anyone else running battlefield assumes they have a hardware issue.

kooky_guy · 11 Jun 2006 at 19:26

t31os said:
Just to add to the thread, BF2 crashing happens on lots of systems, its generally related to the game and not the hardware its run on, though some cases there are exceptions. Just wanted to clarify before anyone else running battlefield assumes they have a hardware issue.

Yes, I know about the BF2 issues, but this is different - it was crashing almost as soon as the game started. 3DMark05 also crashed pretty much immediately.

Anyway, I think I've identified the problem. I tried the RAM sticks individually. One reliably causes the machine to crash whenever I try running 3DMark05 and the other hasn't crashed once.

So it looks like a dodgy DIMM. Any ideas why Memtest didn't pick it up? Luckily Corsair offer a lifetime warranty. Just got to work out how to get them replaced!!

Thanks to everyone for their help.

Guy

titaniumx3 · 11 Jun 2006 at 19:56

It could still be the motherboard. I had similiar problems with an Asus A8V and thought it was the RAM, untill I actually replaced the motherboard. Are you using the same memory slots when exchanging between the two sticks?

kooky_guy · 15 Jun 2006 at 12:14

titaniumx3 said:
It could still be the motherboard. I had similiar problems with an Asus A8V and thought it was the RAM, untill I actually replaced the motherboard. Are you using the same memory slots when exchanging between the two sticks?

Well OcUK have just returned it saying no fault found. Great.

I did use the same memory slot when testing the individual sticks. The good stick worked fine, and the bad stick caused crashes even more frequently when on it's own in the same slot than it did as part of the pair.

I've put my old Crucial memory in while I was waiting for a replacement and it's not crashed once.

So I'm stumped again now. Hopefully Corsair technical support will have some ideas. I'm guessing it's something weird that OcUK's testing didn't pick up. As I mentioned before, it passed Memtest86 with no errors, but is unusable under Windows.

I'm just fed up now.

Guy

mp260767 · 15 Jun 2006 at 12:23

have u tried running dxdiag ?

tried reinstalling gfx drivers or new drivers ?

Dutch Guy · 15 Jun 2006 at 12:36

kooky_guy said:
Well OcUK have just returned it saying no fault found. Great.

I did use the same memory slot when testing the individual sticks. The good stick worked fine, and the bad stick caused crashes even more frequently when on it's own in the same slot than it did as part of the pair.

I've put my old Crucial memory in while I was waiting for a replacement and it's not crashed once.

So I'm stumped again now. Hopefully Corsair technical support will have some ideas. I'm guessing it's something weird that OcUK's testing didn't pick up. As I mentioned before, it passed Memtest86 with no errors, but is unusable under Windows.

I'm just fed up now.

Guy

That leaves mobo incompatibility then :confused:

What timings/voltage are you using, the default ones or different?

kooky_guy · 15 Jun 2006 at 13:13

mp260767 said:
have u tried running dxdiag ?

tried reinstalling gfx drivers or new drivers ?

Yup. Put my new 7900GTX in, so all graphics drivers have been replaced (properly uninstalled and cleaned first).

I'm sure it's the memory. The machine has been working fine for over a year previously and has only recently started doing this.

What I don't understand is why it works ok in DOS mode (ie memtest86). The 'dodgy' stick renders the machine totally unstable in Windows.

kooky_guy · 15 Jun 2006 at 13:24

Dutch Guy said:
That leaves mobo incompatibility then

What timings/voltage are you using, the default ones or different?

Well I wondered about that, but it's been fine for over a year prior to this (on auto mem settings).

Timings are 2.5-4-4-8. The only difference on the auto settings are the mem speed (333, instead of >400), and the command rate 2T instead of 1T. I've tried em all with no difference. I've even tried tweaking the voltage, but nothing makes any difference. One stick works, the other doesn't.

Dutch Guy · 15 Jun 2006 at 13:53

kooky_guy said:
One stick works, the other doesn't.

If you find that is the case maybe contact OcUK via a webnote and ask them how they tested and with what hardware and settings as it looks like one stick is borked.

kooky_guy · 16 Jun 2006 at 11:53

Dutch Guy said:
If you find that is the case maybe contact OcUK via a webnote and ask them how they tested and with what hardware and settings as it looks like one stick is borked.

Well that's a bit mystifing. My memory must have enjoyed it's little holiday as they both seem to be working just fine now...

That's left me completely baffled. Ah well, fingers crossed they keep working.

Cheers everyone.

Guy