Graphics Nvidia SLI crashing. - Finally Fixed!

Test your PSU - Antec Digital PC Power Tester ATX power supply tester

I've found a remarkable number of issues can be traced back to the salmon mousse, I mean PSU

 

If I go through and enable SLI; Driver IRQs stay different. I think I might try just formatting the SSD Clean and then reinstalling off a completely clean install, rather than coping an image back onto the SSD. It'd take a while, but then I'd be 100% sure it was or wasn't windows.

Tony; That wouldn't explain why it runs perfectly for one boot, then instantly fails the next. Bad PSU wouldn't slow down my bootup, anyway, it'd just boot or it wouldn't. Or it'd cook my hardware.

 

As logical as Computers are, there are a few (hardly any) instances where although you would expect a repeated process on exactly the same hardware with exactly the same software to produce the same results, but for some reason they don't.

Might be to do with software variables in the install code.

 

I think it's to do with something the nvidia drivers do upon bootup. Something with applying SLI then or something. The system just can't handle it and results in instability. Enabling SLI from within windows, on the other hand, doesn't have the issue. It's an annoying set of circumstances. Might image this install, format, try at stock, if it still happens; Copy the image back over so I don't loose anything. But OTT, but It means I won't loose too much if it does turn out to be a pointless endeavour.

 

tundra if your about at the weekend i could pop over with my rig and you can test the cards on that, just an option mate. its all on a test bench anyway

 

I'm pretty sure the cards work. I think it's a motherboard/windows problem. Each card will perform crazy things individually (Good crazy. Like massive OC's on already massive OC at stock crazy.), it's just in SLI that the things hate everything. Including each-other. Cheers for the offer, though.

Edit; Hokay. It is most definitely a GPU or Motherboard problem. A complete windows reinstall failed to fix the issue, so I'm running out of ideas now.

 

Have you tried underclocking the cards?

 

Why not just cut your losses and part out the bulldozer rig.

You have had nothing but trouble from it.

 

Because then I'd have nothing else. I think it's the motherboard, but I can't prove it.

The trouble only started when I tore out my watercooling system, then rebuilt it. I suspect I spilt a drop of coolant on the Mobo, that stopped it working, during transit; it got damn cold, then warmed up again, evaporated the coolant off, so it worked once it got to Scan.

This SLI issue might actually be related to that... I should check that board over once more.

Admittedly; I couldn't play some steam games for a while, but that's fixed now. Bios updates are always a good idea.

As far as Underclocking the cards goes; Not tried it. I rather assume that if they're fine for first boot, then fail on consecutive on the same settings; it's not clocks at fault.

I do wonder if there's not a stability issue somewhere else, but I don't have the tools to go over this board with that fine a toothed comb.

 

+1 Just get a SB set up. I don't wanna hear "Then it'd just be boring" either.... no one enjoys having a rig that doesn't work!

 

But equally; Noone enjoys selling a rig for half the price they bought it for because the parts just aren't worth anything more. I wouldn't be able to afford an SB rig, even if I sold all this off.

Even my GPUs are only worth a pittance second-hand now. If 460s go for sub £60 in some places; these are barely worth £50 a piece.

I'll solve this issue eventually, even if it's by going so drastic as to cover the board in rubbing alcohol to clean it off and be sure there's nothing causing issues there.

Now that I think about it; the only parts of my machine still worth anything approaching the original price are the case, which is nigh on spotless, my soundcard, and the SSD. That's a depressing thought.

 

Well I hope you sort it soon. Nothing worse than long term unfixable problems. I go mad if I have a problem I can't fix in 48 hours!

 

I've been working on this for just short of that now.

I'm cutting out more and more, though. It's coming down to the SLi-chips on the cards themselves, latent motherboard instability, or just PCI-E 16X socket 2 being a little bitch. I'm still working out which of the three.

Admittedly; I'm inclined to thing it's the PCI-E socket, because if I plug the main GPU in there and run it in just that socket; I get momentary freezes and lockups, but I haven't any way of really testing to be sure it's just that socket, not a windows thing. I'd rather not RMA again only to get slapped with a £30 charge because Scan can't replicate the problem.

Admittedly; I have also gone back to my F6f beta bios, because the F6 only removed options and tweaked a few other things that I wasn't happy with, plus the F6f worked flawlessly, so I figured I might as well carry on.

Okay, there are no "SLI chips" like what you find on the 590 and such, SLI seems to be wired almost straight into the GPU core, with only a transistor or two in the way.

 

Since you had to overvolt the cards since the beginning to make SLI stable, then you might get somewhere with an underclock. Just get the cores to 800 MHz or something similar and memory to 2000 MHz or even lower. Keep voltages at stock values or even increase them a little.

Also, take jizwizard's offer this weekend. Having someone else to fiddle computer parts with can be fun. Have a few beers and unwind. Sometimes taking a break from a problem helps.

 

I would, if I drank, and was also old enough to legally buy Alcohol. Also; I'm busy for the whole damn weekend. I'm then free Monday and tuesday, but I've no clue if he is at that point.

And it's only recently that I've had to overvolt the cards. Ironically; they're actually stable at stock undervolted when run singularly.

The issue is that these problems occur before windows. I'd have to mod the bios, then flash it to the cards to get them to underclock before Afterburner has had chance to start.

 

Oh, I had no idea you're not allowed to drink. May I ask, what is the legal age you can start drinking there?

Does enabling SLI in windows require a restart? (I really don't know as I never had a SLI rig)

 

Nope. It can be enabled/disabled just from the control panel. No restart.

I ran SLI for a year. Was fun but now back to single GPU to reduce the power draw in game.

 

I'm really suspicious of card two and PCI-E socket two. Just starting up with a card in the second 16x socket netted me three driver crashes so far, and the machine has only been up for a few minutes...
Doesn't seem to be related to load on the card, although Furmark just ran at half the FPS it usually does on a single card...

I think it's down to the PCI socket, although I'll tear the whole machine down and reseat the CPU to see if that can solve the issue. It seems to be that the second 16x socket keeps falling out of sync so it locks up for a second. In single GPU; this only causes stutter, but in SLI it crashes the driver, then cascades because the GPU stays out of sync. After about five minutes; Windows kindly informs me that the driver has failed to recover in the form of a BSOD.

Also it's got some slight graphical glitches, like the fact my cursor turns a light yellow over the text box here, so It's a beech to find it and edit what I'm aiming for.

 

Ok, then it should be possible to underclock the cards in windows, enable SLI, and test it.
The problem would only reappear after a restart.
If there is a mobo issue, it would still be unstable, even underclocked. If everything is fine, (till reboot), then there is something wrong with the cards themselves in SLI, or with the PSU.

 

The issue now occurs on first boot. Consistently so. Problem is that I'd have to send the whole PC back to scan to get them to check the damn thing, so I've asked Jizwizard to come over later in the week and test the GPUs seperately. Just to check that's not the issue.