Lightning 7970 Crossfire Crashing/Freezing

[daravelli]

Hey all, first post here. I apologize for the long post, but I want to provide as much detail as possible with the issue so that someone here might be able to help me out. Thanks for reading.

First off, here are system specs:

Asus P9X79 Motherboard
i7-3820 CPU
16gb Corsair Vengeance 1600mhz RAM
2 x MSI Lightning 7970 video cards
Rosewill Bronze 1000 watt power supply
Patriot Pyro boot drive
WD Raptor data drive (user folders located here)
3 x Asus VW246H Monitors
Windows 7 64 bit

The issue that I'm having is that my computer will freeze up during gaming only with crossfire enabled within about 15 minutes at max (usually more like 5-10). The issue occurs whether or not I have eyefinity enabled. It freezes in every game I play: Dirt 3, Dirt Showdown, Skyrim, Shogun 2, and (although to a lesser extent) Battlefield 3. With some games it will just crash to desktop with a "display driver has stopped working and has recovered" message, and sometimes it will completely freeze up, requiring a reboot. The type of crash is dependent on the game. For the few minutes that it works, crossfire performs great. What's strange is that it will run kombustor at 5760x1080 in dx11 mode (the burn in test) with no errors or other problems. It just seems to be games.

Here are the steps I've taken to resolve the problem:

Removed all overclocks in the system

Tested each lightning card individually in each of the occupied PCI-E slots - Both cards run great on their own in either slot.

Ran each lightning 7970 in crossfire with a 7970 reference card that I used for months without issue - No change, same freezes and driver crashes.

Turned off ULPS

Tested the system with a new OCZ 1250 watt power supply

Checked the integrity of game files

Tried various driver versions - 12.2, 12.3, 12.4, 12.6 beta

Fresh OS and driver install

Updated motherboard bios

Switched between gen2 and gen3 on pci-e slots through bios

Turned off eyefinity

Tried 3 different crossfire bridges, switched the bridges to the other connection, and tried two bridges simultaneously

Lowering the clocks of both cards to below reference levels

Finally, I replaced the motherboard thinking that it had to be that. The other motherboard was an Asrock Extreme6. No dice.

Nothing I've done has made any change whatsoever. At this point I'm thinking that it has to be drivers. Anyone out there have any other ideas that might get this crossfire setup working? I'd like to hear from those of you who have a 7970 crossfire setup without any problems. What drivers are you all using, etc. Any help is much appreciated, and thanks in advance.

 

[SickBeast]

It sounds like a heat problem or an unstable overclock. One of your cards is probably unstable. You need to figure out which one it is so you can RMA it in all likelihood, or else you could just tweak the clockspeeds using a software tool.

What's your case cooling like? Those cards will dump a *ton* of heat in your case.

 

[lavaheadache]

Lightnings are technically 3 slot cards (GPU reactor). How much space is in between the cards? Are you monitoring temps during gameplay? It would help to know how hot they are getting.

I suppose it could also be a psu problem. 4 8pin connectors can draw a lot of juice and if your PSU is flaky that could be a problem.

I have a Lightning too and love it =)

 

[daravelli]

It sounds like a heat problem or an unstable overclock. One of your cards is probably unstable. You need to figure out which one it is so you can RMA it in all likelihood, or else you could just tweak the clockspeeds using a software tool.

What's your case cooling like? Those cards will dump a *ton* of heat in your case.

The thing is that I don't even have anything overclocked right now in an attempt to get this thing solved. Case cooling is good; 3 230mm fans (2 in, 1 exhaust), and 3 140mm fans (1 in 2 exhaust). I've stress tested both cards and can't get either one to fail. I've tried lowering clock speeds on the cards as well.
​

 

[SolMiester]

Have you tried the 7 series release drivers?..from January?...I could be wrong but thought that XFire was broken for eyefinity unless you used the release drivers...

 

[jackstar7]

As someone using Xfire, I can only say that I'm still using the release drivers... and no issues at this point with Cross-finity.

 

[daravelli]

Lightnings are technically 3 slot cards (GPU reactor). How much space is in between the cards? Are you monitoring temps during gameplay? It would help to know how hot they are getting.

I suppose it could also be a psu problem. 4 8pin connectors can draw a lot of juice and if your PSU is flaky that could be a problem.

I have a Lightning too and love it =)

The P9X79 Deluxe leaves close to two inches between the cards, so no issue there. I'm using afterburner for monitoring and temps on the cards are generally below 65, sometimes creeping close to 70 during games. I think the max I've seen is around 73. As far as the PSU, I did test an OCZ 1250 watt on the system with the exact same results. I was actually convinced that was the issue at one point, but it doesn't seem to be.

 

[daravelli]

Have you tried the 7 series release drivers?..from January?...I could be wrong but thought that XFire was broken for eyefinity unless you used the release drivers...

As someone using Xfire, I can only say that I'm still using the release drivers... and no issues at this point with Cross-finity.

Hm... worth a try. The 12.1 drivers then? I actually haven't tried those. I'll check it out and post back. Thanks!

Edit: Well I got excited there for a second, but same results using the 11.12 RC11 driver.

 

[blackened23]

TDRs are usually a sure sign of heating issues, are you able to put more space between them? IE use a different pcie slot for gpu2?

 

[daravelli]

TDRs are usually a sure sign of heating issues, are you able to put more space between them? IE use a different pcie slot for gpu2?

Yeah and I've actually already got it in that configuration. I'd say there is a good 2 inches in between the cards.

 

[ShintaiDK]

Maybe try increase the TDR timeout value as seen here:
http://forums.anandtech.com/showthread.php?t=2238234&page=2

 

[daravelli]

Maybe try increase the TDR timeout value as seen here:
http://forums.anandtech.com/showthread.php?t=2238234&page=2

Tried it and no deal... dirt 3 still crashes within about 2 minutes. Still, that's something I didn't know about... thanks for the suggestion!

 

[ShintaiDK]

Try uninstall any GPU related tools you dont use. If there is any. Else I dont know :|

 

[daravelli]

Try uninstall any GPU related tools you dont use. If there is any. Else I dont know :|

The only thing I use is afterburner, and only for the fan control. I've tried uninstalling that and it doesn't make a difference.

It's really strange, I can run kombustor stress tests and burn-in with no issues, 3dmark has never crashed. I'm starting to wonder if it has something to do with the onboard sound. I tried disabling it, but neither dirt 3 or dirt showdown will start without it enabled. Maybe I'll test the theory with skyrim later... if it will start.

 

[TakeNoPrisoners]

That sounds very much like an unstable overclock. Those cards are overclocked from the factory, maybe the factory overclock is not stable in crossfire.

That would explain why it works with a reference 7970 as the lightning will downclock to match the slowest card.

 

[daravelli]

That sounds very much like an unstable overclock. Those cards are overclocked from the factory, maybe the factory overclock is not stable in crossfire.

That would explain why it works with a reference 7970 as the lightning will downclock to match the slowest card.

Hi... sorry if I wasn't clear on my explanation of the reference+lightning crossfire tests. That setup was NOT stable with either of the lightning cards. Same crashes. I was only saying that I knew the reference card was a stable card itself because I'd used it for a few months on its own with zero issues.


I've actually tried downclocking the cards to reference level clocks as well as lowering them well beyond that. No change whatsoever.

 

[ShintaiDK]

You could also try disable TDR completely to test with. (TdrLevel to 0)

Just to see if it HW freezes or just runs.

 

[daravelli]

You could also try disable TDR completely to test with. (TdrLevel to 0)

Just to see if it HW freezes or just runs.

Went ahead and tried that out. Nothing good... lol

 

[ShintaiDK]

So HW freezes you say

 

[DigitalWolf]

Go into your device manager and see if under sound you have listings for AMD. (like for the hdmi sound).


If you do... disable those...

 

[daravelli]

Ok, so I installed the RC11 realease drivers last night, but got the same crash initially in dirt 3. Today I tried out skyrim and got no crashes after an hour or so. So then I decided to force dx9 in dirt 3 and dx10 in dirt showdown. No crashes. Battlefield 3 isn't crashing either... it looks like that solved it! Man, and after trying so many things including changing out the MB... it was a f*****g driver, lol. Oh well, thanks to all of you who contributed to the thread. Much appreciated.

 

[BrightCandle]

I have had nothing but problems with AMD's drivers and crossfire. Most of the crashes you will see are directly related to driver bugs, not faulty hardware. They seem to still have severe problems 5 months after release. Try reporting a bug on the issue and you'll get blank looks from AMD like everything is really OK, when the reality is a lot of people are getting BSOD and severe performance problems.

I have lost confidence they will fix the issues and switched to NVidia. No problems so far.

 

[SolMiester]

Ok, so I installed the RC11 realease drivers last night, but got the same crash initially in dirt 3. Today I tried out skyrim and got no crashes after an hour or so. So then I decided to force dx9 in dirt 3 and dx10 in dirt showdown. No crashes. Battlefield 3 isn't crashing either... it looks like that solved it! Man, and after trying so many things including changing out the MB... it was a f*****g driver, lol. Oh well, thanks to all of you who contributed to the thread. Much appreciated.

When the 680 came out, HardOCP did a dual card review and stated that only Release drivers are stable with XF & Eyefinity!.......
6 mths and still no new stable driver.....LMAO!...sorry dude

 

[SolMiester]

I have had nothing but problems with AMD's drivers and crossfire. Most of the crashes you will see are directly related to driver bugs, not faulty hardware. They seem to still have severe problems 5 months after release. Try reporting a bug on the issue and you'll get blank looks from AMD like everything is really OK, when the reality is a lot of people are getting BSOD and severe performance problems.

I have lost confidence they will fix the issues and switched to NVidia. No problems so far.

Sig update required....

 

[ocre]

perhaps its the PSU.

Your PSU is rated 1000w but its easy to be fooled when you have multiple 12v rails. Your PSU most likely is struggling to keep the up with the draw when you CF lightning 7970s.

For some reason we see video cards rated by watts when its pretty useless considering that the PSUs are limited by the AMPs they can supply on their 12v rails. Most decent PSUs today have a very high AMP rating for a single rail. This is the best option especially for huge GPUs and multiple configurations. this PSU for example:

http://www.newegg.com/Product/Produc...82E16817139021

is rated for 750watts but it uses a large 12v rail that supplies 62amps

Your PSU has several 12v rails and none of them are high:

http://www.newegg.com/Product/Produc...scrollFullInfo

"+12V1@20A, +12V2@20A, +12V3@30A, +12V4@30A"

A lot of people will tell you its best to have single rails with a high amp but the multiple rails can work. They can be problematic especially if your barely cutting it in the first place.

Back to your situation. MSI recommends 40amps 12v rail for a single 7970. None of your rails can supply this by itself. Combined this can be achieved. But you cannot draw the max a rail can supply for extended periods of time. as a matter of fact, from the link above:

http://www.newegg.com/Product/Produc...scrollFullInfo

"12V Combined Output 900W (75A)"

your PSU can only really output 75A with all the 12v rails added up. this is 25% less than the rated amps for the rails (20+20+30+30=100w). This is because the rating for the rails is peak not sustainable. so if the combined output rating of your PSU is 75watts and MSI recommends a 12v rail of 40w for a single vanilla 7970 then your CF setup is definitely pushing the limits of your PSU.

This is why i vote: your PSU is your most likely issue.

For the future, i recommend you get powerful single rails with high wattage. You should also keep track of your cards power draw and make sure you have plenty of amps and then some left over. Never run close to its max rating because PSUs arent capable of 100% all the time. In this case you are running over your PSUs recommendation from the start.

its not all bad, you just need to try a more capable PSU and i bet your issue will go away!!!!!