Real tricky problem with watercooling

[Pr0sper0]

I've had my rig for about a year now and something is getting really hot and for the life of me I cannot work out what.

When playing CPU intensive games like ARMA 3 I start to get artefacting after about 5 minutes, which leads me to believe its heat related, it gets worse the longer it's played.

http://www.youtube.com/watch?v=4VJ462H90vU&feature=share&list=UUwPKMvQ-gTxpSMzCBLA5t6g

According to EVGA precision the two 680 GTXs are running at 34 degrees
while the CPU (according to ASUS AI suite) is 38, motherboard at 26 and PCH at 33

I've completely overhauled the watercooling, cleaned everything and even changed the direction so that CPU is straight after radiator and then the two graphics cards.

I've recently replaced the CPU waterblock and even swapped out the motherboard just in case and even completely rebuilt the OS with Windows 8.1 from Windows 7.

Something is wrong, but I'm completely baffled. My suspicions are pointing towards inadequate flow, memory overheating, memory bios settings or a bad CPU.

Really need some help guys...

Aquacomputer D5 pump with 480mm radiator with 4 120mm fans
Two EVGA GeForce 680 GTX with Koolance waterblocks
Intel 3930K at stock speeds at the moment with Koolance waterblock
ASUS Rampage Formula IV
16GB Corsair Dominator GT Quad channel at 2300

 

[Meeho]

How is the GPU memory cooled? Good contact with the water block?

 

[lilfiend]

GPU VRM's/Memory cooled properly? Enough airflow over CPU VRM/MOSFET's/RAM? Southbridge controller overheating due to lack of airflow?

It's also possible that you have a GPU going out, I'd try updating/rolling back drivers too just to be sure.

 

[silverstonec2d]

Could you get some temp readings from running Prime95 or IBT while using realtemp or coretemp to get a baseline idle/load temp for your 3930K. Also, for ARMA 3, try running gpu-z with the logfile option enabled so when you exit the game when the artifacts occur you'll have a file to view and you will know what your temps, voltage, ect are. Also, seeing as your running an X79 chipset board and same cpu as me, I know running memory at that speed (2300mhz) it may be a bit unstable. I think it's "luck of the draw" at those speeds. Ivy-E can typically do higher memory speeds. This is probably not the problem, but since your overclocking at the moment, maybe try lowering your memory frequency to 1600 or so with the SPD and see what happens. Also, any chance of an air pocket in your loop??

 

[Arcygenical]

Try underclocking the gpu and vram by 100mhz and see if that helps at all.
Throw your CPU at stock too, just to see if that has anything to do with anything.

 

[Spazturtle]

A cheap laser thermometer would help you find exactly where is overheating.

Could you remove all but a single stick of ram and try that please?

 

[TigerLord]

What case is this

 

[Jorona]

I'd run one card at a time. That kind of graphical corruption is usually core related, not memory. You might just have a dying card. Especially since its like flashing corruption. Card 1 is displaying its frame correctly, then card 2 sends its fucked up frame.

 

[Pr0sper0]

Fantastic responses, thanks ever so much. I've isolated the problem

Running the cards individually, see results from card 1 and card 2

Now I'll strip the waterblock from card 2 and reapply thermal compound, hopefully its just that or something similar and not the card itself
although it's under warranty.

Just so chuffed, thank you guys. This one had me defeated!

BTW the case is a Silverstone Temjin TJ-11 which is just my perfect case but there is a more reasonable TJ-07 which I had before and is also a great case just a bit smaller.

 

[TigerLord]

thanks for that, the case is indeed sexy. I was considering a 900D but this looks good.

Glad you solved the issue.

 

[topcaliber]

Nice setup, looks professional.