8800GT SLI Artifacts

[pp3rkut]

After hours of troubleshooting the weird issues I was having with my SLI'ed 8800GT's (BSODs, Artifacts, etc...), I believe I have found the solution to my troubles. I noticed that the issues appeared to only happen with SLI enabled, so I was doubting the stability of the second card in the config. I tested each card independently. I was hard pressed to find issues with either card when running in the primary PCIe slot. A first run through suggested that each card was fine... similar 3dmark06 scores, no immediate artifacts while gaming. This really annoyed me, as it would have indicated issues lying elsewhere in my system rather than a bad card.

 

I was determined to find an error with either card so I plugged the first one in and played various games for over eight hours straight. Call of Duty 4, Team Fortress 2, and Command and Conquer 3 all had no issues whatsoever.

 

Today, I plugged the second card in and crossed my fingers hoping I could find something wrong with it. Call of Duty 4 played through fine after about 3 hours of play. However, Team Fortress 2 appears to be artifacting! Not in the same manner that it did in the SLI configuration where various textures turned pink (I assume because of the different way SLI will load the GPU), but with characters appearing to glow.

 

Before I send my card in, I just wanted to post a screenshot of the issue and verify that this does indeed indicate a graphics problem and is not some weird looking effect in Team Fortress 2.

 

Attached is the image of my artifacts (?)

 

Thanks for all the help.

-PP3RKut

Edited February 4, 2008 by pp3rkut

[Shadowfactor]

That glowing of people indicates an Uber charge where a person in the game is like invincible for a very short amount of time.

 

This is part of the game.

[pp3rkut]

Alright, thanks for the help. I guess the issue lies elsewhere in my system. I was thinking that this looked more like part of the game. My previous issues with SLI had multiple textures turning pink, it appears that this doesnt happen with the individual cards.

 

Thanks for your help, saved me a pointless RMA.

[cyclepath]

After hours of troubleshooting the weird issues I was having with my SLI'ed 8800GT's (BSODs, Artifacts, etc...), I believe I have found the solution to my troubles. I noticed that the issues appeared to only happen with SLI enabled, so I was doubting the stability of the second card in the config. I tested each card independently. I was hard pressed to find issues with either card when running in the primary PCIe slot. A first run through suggested that each card was fine... similar 3dmark06 scores, no immediate artifacts while gaming. This really annoyed me, as it would have indicated issues lying elsewhere in my system rather than a bad card.

 

I was determined to find an error with either card so I plugged the first one in and played various games for over eight hours straight. Call of Duty 4, Team Fortress 2, and Command and Conquer 3 all had no issues whatsoever.

 

Today, I plugged the second card in and crossed my fingers hoping I could find something wrong with it. Call of Duty 4 played through fine after about 3 hours of play. However, Team Fortress 2 appears to be artifacting! Not in the same manner that it did in the SLI configuration where various textures turned pink (I assume because of the different way SLI will load the GPU), but with characters appearing to glow.

 

Before I send my card in, I just wanted to post a screenshot of the issue and verify that this does indeed indicate a graphics problem and is not some weird looking effect in Team Fortress 2.

 

Attached is the image of my artifacts (?)

 

Thanks for all the help.

-PP3RKut

You are kidding right? That is part of the game dude. It's when a player goes uber..... look it up...

[yatstv]

ever thought it might be the sli connector??? or the secong sli slot on mobo that is bad. try cards in second slot. i know depending on mobo board that the second slot is sometimes slower, but you should just try the slot out too.

[pp3rkut]

You are kidding right? That is part of the game dude. It's when a player goes uber..... look it up...

 

Sorry, I never play team fortress. I was simply spectating a game under a friends account, as this was the game that caused the pink screens with SLI enabled.

 

Yes, I have considered that it could be the connector or second PCIe slot. I currently don't have a spare connector around, I'll see if I can borrow one from a friend. As for the PCIe slot, I've been told the 680i LT doesn't let you boot in it and I havn't tried.

 

I am still not fully convinced that my card isnt bad, as the errors under SLI occured very infrequently, making this annoying to troubleshoot. I'm sticking the secondary card in a continuous loop of 3dmark06 overnight.

 

A couple questions for you guys:

Can you run SLI in software mode, thus eliminating the possibility of the connector causing errors?

Will booting under the second PCIe slot cause any sort of damage? I was advised against this, however, I would like to rule out the slot as being a cause of the issue.

And does anyone know of any programs that stress the GPU more than 3dmark06 that I could loop, I have noticed that when running 3dmark is doesn't even heat up my cards to their maximum temperatures.

 

I will test the video cards independently more extensively and post back with my findings.

 

Thanks for all the help guys. Hopefully I can figure this out, as I've had this issue bugging me for some time now.

 

EDIT: another note, regarding the PCIe slot. I forgot that these errrors have been occuring on two 680i LT's, I find it highly unlikely that on both boards the PCIe slot was bad. Possible, but unlikely. I will investigate this if all else fails.

 

System Specs:

Q6600 (currently stock while troubleshooting) w/ Zalman 9700

2GB Crucial Ballistix DDR2-800

eVGA 680i LT

PC Power & Cooling Silencer 610W

Edited February 4, 2008 by pp3rkut

[pp3rkut]

I just tested the PCIe connectors, both output a consistent 12.05V under load, so that rules out the power supply as a possible cause. After more testing both cards appear to be fine when running independently, although I will post back after the overnight run through on each card.

[pp3rkut]

Update on my progress so far.

 

I continued testing each card in the primary PCIe slot, and they both seem to check out fine. No artifacts, BSODs, or errors of any sort after extensive gaming and 3dmark06 loops.

 

Basically, I've come up with a list of possible causes for my issues:

 

- Bad graphics card(s)

- secondary PCIe slot bad

- Power Supply

- SLI Bridge

- Driver Issue

- Overheating

 

I am fairly confident that both graphics cards are okay, so for now I will assume that the errors under SLI that I am seeing are not related to a bad graphics card.

 

I have tested the power supply under load as noted above, the voltage seems fine over each PCIe supp. power connector.

 

All of my testing is being done on a fresh install of XP with the latest drivers. The problems under SLI occurred over multiple driver types. I do not believe it is related to the drivers.

 

Using Rivatuner's Hardware monitor, each card maxes out at 80C under full load, which is hot, but not hot enough to cause any issues and it well within operating specs.

 

 

This leaves only the possibilty of a bad secondary PCIe slot, or a faulty SLI Bridge.

 

 

I have now swapped in each graphics card to the secondary PCIe slot and tested it, and have found that under 3dmark06 the scores in the secondary PCIe slot are slightly lower regardless of the card installed. I was wondering if this is due to a bandwidth constraint on the secondary PCIe slot, or rather indicates a fault? If anyone out there has the 680i LT (or 680i for that matter), I was wondering if you could comment on this. It is possible the nForce MCP is slightly restraining the bandwidth for the secondary PCIe slot.