Bubble causing bad GPU temps?

[penti01]

So, I've got a Rampage V Ed 10 mobo, i7 6800k, and two 1080 Ti running on an EK D5 pump, one 360 and one 480 rads.

Now, I've got CPU temps running around 50C, the second 1080 Ti running 40C, while the first 1080 Ti is running 70C on idle..

I've set it up running parallel, so in left of GPU 1 > out/in left and right to GPU 2 > out right of GPU 2.

This is my first custom LC so please be patient with me, but could this be caused by a trapped air bubble in the GPU 1 (top)?

I've been leak testing for 48 hours, and have just booted my PC for 2 hours.

Would rather try getting air bubbles (if any) out before draining, reassembling GPU block, etc..

[Foxxer]

Tip and tilt the case. Yes go extreme, remember air loves going upwards. 

 

 

[0ld_Chicken]

Could you post pics of the loop?  also what bridge or form of connection did you use between the cards?

[penti01]

I just used two flexible tubing to connect both GPUs..

[For Science!]

A photo will be very helpful, perhaps the parallel configuration is wonky especially if you haven't used a bridge.

[For Science!]

This is a correctly routed parallel loop, in my case the water comes from the bottom, but you can do it in reverse without too much impact on performance. You can also switch the position of the stop plugs and the in/outs  and also be in the okay. Normally the two GPUs should be very close in temperature is both are idle/loaded and so the huge discrepancy may indicate that one card's water is not flowing

 

[penti01]

Pic is bad, but basically this is the flow of my loop

[For Science!]

1 hour ago, penti01 said:

-snip-

Apart from the fact that technically the flow in the GPU block is in reverse (right to left), it looks good. EKWB says the temps should not be too different, so I wouldn't bother swapping them around until your next scheduled draining.

 

Then perhaps indeed is a matter of rocking the case to further bleed the system. Based on the flow direction and block design the order which I think the tilting needs to be done is (relative to the picture poster above):

 

- Rock to left

- Rock towards you (without returning to the center, so it is on the left front feet)

- Rock to right (without returning to the center, so it is on the right front feet)

- Rock away from you (without returning to the center, so now on the right back feet)

- Rock to left (now on left back feet)

 

You should observe air moving through the system and hopefully accumulate in the reservoir.

 

Alternatively you can try a similar motion with the PC on its back and that should also promote the air to move upwards.

[penti01]

8 hours ago, For Science! said:

Apart from the fact that technically the flow in the GPU block is in reverse (right to left), it looks good. EKWB says the temps should not be too different, so I wouldn't bother swapping them around until your next scheduled draining.

 

Then perhaps indeed is a matter of rocking the case to further bleed the system. Based on the flow direction and block design the order which I think the tilting needs to be done is (relative to the picture poster above):

 

- Rock to left

- Rock towards you (without returning to the center, so it is on the left front feet)

- Rock to right (without returning to the center, so it is on the right front feet)

- Rock away from you (without returning to the center, so now on the right back feet)

- Rock to left (now on left back feet)

 

You should observe air moving through the system and hopefully accumulate in the reservoir.

 

Alternatively you can try a similar motion with the PC on its back and that should also promote the air to move upwards.

Really, didn't know that the GPU had specific IN/OUT ports.

Anyway, will try further bleeding in the system to see if that helps..

[penti01]

So, I've tried bleeding the system but to no avail, so drained the loop, and changed the GPU flow from parallel to serial.

However, while I've seen better temps on the aforementioned card (now down to 85C under load), it's still too high. Running one card, and stress testing the GPU would spike the GPU temps to 45C max.

 

I'm not certain, but could this possibly be due to incorrectly installing the GPU block?

I remember the block shifting a bit during installation (had to remove GPU block for a minute but didn't reapply the thermal paste since it barely made contact for a minute and it didn't have time to dry out yet). Could this be the possible culprit?

[penti01]

Just an update to anyone reading this in the future.

 

I opened up the GPU block, and saw that the thermal paste did not set (still wet, block popped right out).

I'm not sure how long the EK thermal paste takes to set, but from my experience, a week would result in slight caking already.

 

Either way, popped the GPU block off, and apparently a small plastic was preventing good contact from the GPU to the block (it most likely came from the middle cutout of the plastic washer since it was about the same diameter).

Cleaned it off, applied new thermal paste, reassembled the loop, and currently leak testing again just to be sure.

 

Should be able to test in a few hours, and hopefully everything is running nice and cool.

 

Update 1:

Seemed to do the trick, but GPU 1 (the problematic one) now runs at 50C under load, while GPU 2 runs at 40C. There's a 10C difference between the two, but I'm thinking this is due to inadequate bleeding since there's significant bubbles seen in GPU 1's block.

 

Hopefully, can bleed it thoroughly over the weekend.