SLI crashing randomly - desktop, in-game, during HW,...

evan1715

Distinguished
May 26, 2011
188
0
18,710
Heya!

Idk what the hells going on honestly -_- lol.
I recently added on another 660 Ti to my setup to enable SLI. It was a much cheaper option off eBay to get a used card and SLI than to modernize the card. Both cards work just fine when in single use.

When I run my SLI, it will randomly just freeze my computer. It can be when I turn on the computer and the desktop loads, it can be while browsing the internet, while doing HW on PP/Word, while in-game, or during skype. Just as random as can be. Sometimes it'll be just the driver crashing and resetting without full system freezing, but most of the time it's just the entire thing freezing, not even shutting down or auto-restarting.

Yes, yes, I have, of course, reinstalled the driver, done a driver clean sweep, checked connections, overclocked the gfx cards, underclocked the gfx cards, used the default clocks on the cards, blah, blah. I even reformatted for a fresh, clean Windows install. The driver is up to date. The driver's power management mode is set to "prefer maximum performance." I have tested it with single card and it does not crash/freeze with single card use.

Here's the setup:
Windows 10 Pro 64-bit
EVGA 2x660 Ti SLI -- Temperatures stay below 72ºC & I'm using the SLI bridge that came with MoBo
Intel i7-3770K @ 4.4GHz -- Temperatures stay below 80ºC
Corsair 4x4GB @ 1600MHz -- At stock clock & speeds
ASRock Z77 Extreme4 -- At stock clock
Corsair 120GB SSD
Seagate 500GB HDD
Corsair 750W PSU

So, can anyone assist or give some guidance of what I should try?
 
Solution
(late reply because i'm trying to confirm results after days of crashing or non-crashing rather than just one day.)

the error wasn't the timeout thing, the error had multiple ones like nvlddkms.sys or AV_dxgmms2 occuring.

i don't think it'd be the PSU because it happens even when at desktop or idle.

so, with the windows 10 anniversary update, i think it solved the problem. ever since the update there has not been a crash. i am still unsure what the problem was in the first place, but the win10 update seems to have helped so far.
Which Corsair 750 watt PSU ?

Any hints in Event Viewer

What happens at stock settings for CPU / GPus

Are cards exact same model ? Are they "synchronized" (MSI Afterburner) to run at same speed ?

Win 10 is famous for borking SLI and other driver installs. When re-installing Win 10

1. Do a fresh install not an upgrade

2. Connect boot drive data cable to lowest numbered SATA port .... but don't connect:
-Ethernet cable
-SATA Data Cable for any other drives

3. Install Windows to SSD

4. Turn off Windows ability to install Hardware drivers
http://www.tenforums.com/tutorials/8013-windows-update-...

5. Install all hardware drivers from original media in boxes

6. Connect ethernet cable and run Windows Update till it stops doing anything

7. Install latest drivers for ALL hardware from manufacturers web sites. The reason we didn't do this in step 5 is 3 fold:
-Sometimes MoBo CD comes with licensed utilities with product key embedded and these are not on web site versions
-Its advisable to establish a working out of box conditions
-Sometimes newer drivers require latest Windows Updates

If you have a 2nd drive ....

8. shut down and connect HD data cable to 2nd lowest numbered SATA port.

9. Boot to BIOS and make sure that SSD is the 1st boot device, if not fix it

10. Boot to Windows
 
the cards to work individually. i have not tried another SLI bridge. i would have to order one.

the model of the PSU is Corsair TX750 v2.31.
i haven't downclocked the CPU. curious, how would that effect it?
yes, the cards are the same model at the same clocks. EVGA 660Ti
if i am to reinstall Windows again, it'll be some time away. idk if that'll work, just by changing the method. last time, i think i did install stuff by use of motherboard CD's and then downloaded the latest driver from nvidia's website.

yes, the two cards have the same amount of vram. they're both 2gb.
 
i tried stock CPU. i thought it was working for a day, but it crashed again.

in event viewer, there seems to be something - "nvlddmkm.sys"
this could also just be the driver crashing because of the crashed computer. other drivers say they had an error too. however, sometimes when I get the blue screen, it'll say there was a problem with nvlddmkm.sys
 
Got a feeling it's the PSU, how old is it, do you have access to another you could try. May be getting old/weak, you mentioned dropping the CPU OC down and it sort of worked that would have allowed a little more juice to the GPUs, so it may just be at the edge power wise
 
(late reply because i'm trying to confirm results after days of crashing or non-crashing rather than just one day.)

the error wasn't the timeout thing, the error had multiple ones like nvlddkms.sys or AV_dxgmms2 occuring.

i don't think it'd be the PSU because it happens even when at desktop or idle.

so, with the windows 10 anniversary update, i think it solved the problem. ever since the update there has not been a crash. i am still unsure what the problem was in the first place, but the win10 update seems to have helped so far.
 
Solution