Turned On Crossfire... GPU / PCI-E Slot Dead?

[clownshoes]

I really have no idea what is going on and am in desperate need of help.

I have 4 Sapphire 290x cards on an ASUS RIVF motherboard. I finished building this system about a week ago and it has been running flawlessly until today.

Today I decided to crossfire 2 of the 4 cards together in catalyst control center. The second I applied the change to the cards, my display went blank and one of the lights next to the PCI-E lane switches on the RIVF board went off.

I'm not sure if it's the card or the PCI-E slot but now the card in slot-1 isn't working. The light next to the slot-1 lane switch isn't lit up and neither of the display ports are working.

The biggest problem is that the computer is water cooled so I can't simply remove one of the video cards and test the slot. All 4 cards are connected together in the loop so I would essentially need to tare the entire system down, drain the loop, put it back together and fill it up just to test the slot.

I really don't see how simply enabling crossfire on 2 of the cards could kill the gpu or the PCI-E slot. Anybody have any suggestions? I haven't messed around with overclocking or anything like that. I've just been running the system at stock for the last weak and testing everything out (running flawlessly until this).

Thanks, any help is greatly appreciated.

 

[Rizen]

Are there any BIOS options to enable/disable specific PCI-E slots for testing purposes? I thought I have seen some motherboards with those features. Sorry, I don't know much about your product specifically, but that is where I would start. If you aren't sure, try first powering off the machine and resetting the CMOS/BIOS and going from there.

I'm doubtful that just enabling CrossFireX would kill a card. There's probably another issue.

 

[clownshoes]

Thanks for the response.

Yeah there is an option to turn on/off the PCI-E slots. The Rampage IV boards have PCI-E lane switches on the board itself. This is what I mentioned in my first post as well. The light next to lane switch 1 is off indicating that there is no card in that slot (or the card / slot itself is busted). All the other 3 lights are on.

 

[Rizen]

Did you check to see if there are BIOS options to switch on/off slots on the motherboard? Or jumpers or dip switches next to the slots?

 

[HOODedDutchman]

First step reset cmos. If you can boot into windows with 1 of the other cards running your monitor and switch crossfire back off to see if the primary comes back on it'll help diag.

 

[clownshoes]

Did you check to see if there are BIOS options to switch on/off slots on the motherboard? Or jumpers or dip switches next to the slots?

Yeah, the switches I mentioned above do this. The problem appears to be with the first slot. I just don't know if it's the motherboard or the video card. Is there some other way you recommend I use these switches to test? The cards in slots 2,3 and 4 all display output when connected to a monitor and all show up in hardware monitor. The first card does neither.

First step reset cmos.

Thanks, this was the first thing I did (clear CMOS button on back of RIVF). It doesn't seem to make a difference.

All the cards are acting funky now though. After a clean Windows install and multiple CMOS resets, they are still using engine clocks that I set in cgminer before all these problems began (I turned the engine clock down to 935 from 1000). This is with a clean cgminer config and no specification for engine clock.

Additionally, cgminer profiles that worked flawlessly before now cause cgminer to throw up a bunch of error text and ultimately stop responding. When I am able to get cgminer to run, I get 100% HW errors now (never got a single HW error prior to this whole crossfire mess).

I'm starting to get desperate. So many things are acting up that I don't even know where to begin.

Does crossfire change anything in the bios of the card itself? Is it possible that some crossfire settings got carried over to the new installation in the card bios and this is causing the problems? If so, is there a way to reset the 290x bios to factory default?

I know this is a lot of information and I greatly appreciate people taking the time to help me.

 

[Rizen]

Can you boot into Windows and disable Crossfire? Does that help?

If not, can you boot into Windows Safe Mode, and completely uninstall the AMD drivers? Then install them fresh again?

Edit: I just saw this:

After a clean Windows install

So you did a complete reinstall of Windows after this started happening and you are still experiencing the same problems?

 

[clownshoes]

First step reset cmos. If you can boot into windows with 1 of the other cards running your monitor and switch crossfire back off to see if the primary comes back on it'll help diag.

Can you boot into Windows and disable Crossfire? Does that help?

If not, can you boot into Windows Safe Mode, and completely uninstall the AMD drivers? Then install them fresh again?

Edit: I just saw this:

So you did a complete reinstall of Windows after this started happening and you are still experiencing the same problems?

Since the display went out right when I applied the crossfire setting, I wasn't able to see exactly what was happening or if there was anything else I needed to do. I had to do a hard shutdown because I couldn't get the display working.

After messing around with things for awhile, I was able to get the display working on the old installation but windows was only detecting 3 of the 4 cards.

My first thought was to revert the crossfire setting in the catalyst control center but it actually showed crossfire as disabled so there was nothing for me to revert (like I said, the display went blank so I wasn't able to see what was going on).

I messed around with a bunch of different things in windows and the BIOS. I tried to uninstall the graphics driver but right when I uninstalled it, the display once again went blank and didn't return. This forced me to hard reboot / mess around with clear CMOS and swapping the monitor to different cards to finally get a signal again.

When I finally got the signal back I decided to do a clean install of Windows. I used all the same drivers for this install as I did the last so everything should have worked like it did before (minus the video card in slot 1). This is why I'm wondering about the video card bios.

 

[HOODedDutchman]

Not to be a dick but this is why hardware should be tested prior to throwing blocks on. Your board does have the ability to switch off individual pcie slots tho does it not ? Not sure if this will help in diagnosing tho as you have no crossfire bridges so turning off a slot may stop communication between other cards ? Not sure as I have no experience with 290/X crossfire.

 

[clownshoes]

Not to be a dick but this is why hardware should be tested prior to throwing blocks on. Your board does have the ability to switch off individual pcie slots tho does it not ? Not sure if this will help in diagnosing tho as you have no crossfire bridges so turning off a slot may stop communication between other cards ? Not sure as I have no experience with 290/X crossfire.

Yeah, I most definitely tested everything out before I threw it in the water loop.

Get this though. I finally decided to take a break from this computer so I powered it down and unplugged it (something I did many times in the troubleshooting process). I left it off and unplugged for 4-5 hours. When I came back, plugged it back in and powered it on I noticed the first PCI-E lane switch was back on. I powered the computer back down, plugged the monitor into the first card and powered it back up. What in the flipping hell?!?! the damn thing is working again!

I remember somebody referring to scrypt mining as black magic and I am now convinced it is nothing less. The guy that created cgminer is most definitely a warlock.

Even more odd, my cards all prefer different settings in cgminer compared to my first installation (I installed the exact same drivers in the exact same order as my first installation). Drastically different I might add.

Seriously though. What in the world could have caused this? The only thing I did differently was leave the computer unplugged for several hours as opposed to only a few minutes before (why would this make a difference?).

 

[Rizen]

I have no idea but I have fixed computer issues before by leaving the PSU unplugged for a while. So.. sometimes black magic works.

 

[clownshoes]

I have no idea but I have fixed computer issues before by leaving the PSU unplugged for a while. So.. sometimes black magic works.

Ha ha, yeah. I guess I should just be happy everything is working now considering what a pain in the ass it is to tear down the water loop I built.

This was my first time water cooling a build and although everything in the loop works great, I definitely need to make some modifications so it isn't such a nightmare to add / remove components.

Anyway, thanks again to everyone who offered advice.

 

[Rizen]

Ha ha, yeah. I guess I should just be happy everything is working now considering what a pain in the ass it is to tear down the water loop I built.

This was my first time water cooling a build and although everything in the loop works great, I definitely need to make some modifications so it isn't such a nightmare to add / remove components.

Anyway, thanks again to everyone who offered advice.

Install a drain tube. It's not too much work and it makes draining and refilling the loop 100x easier.

 

[Karlitos]

Any updates?

I also have Quad 290x and I cannot run my cards in Crossfire / Tri / Quad with the display port. It loses the signal. But I can run the display port on each single card.

Very weird.

 

[Thor]

I know you got it working again, but I cant help but feel like you might want to expect a PSU failur in your future. Watch your voltages.