Big difference between GPU temps

[Davidwr80]

Hi,

Yesterday I made a post about my GPU temps in my new custom loop, which I found to be a bit high (53C lower card, 65C top card). Since then, I decided to flush the loop, I changed the thermal paste on both GPUs and checked if all the thermal pads were cut to their correct sizes. Now, my lower card sits at 39C under load, but the upper card is now at 68C. I use an EK serial terminal to connect the graphics cards, the water enters the lower one first and the goes to the upper card. Therefore, I expected that the upper card would run a bit hotter, but not 27C hotter. Any ideas why that is? could it be that my pump is too slow or has not enough pressure? 

[Legolessed]

22 minutes ago, Davidwr80 said:

Hi,

Yesterday I made a post about my GPU temps in my new custom loop, which I found to be a bit high (53C lower card, 65C top card). Since then, I decided to flush the loop, I changed the thermal paste on both GPUs and checked if all the thermal pads were cut to their correct sizes. Now, my lower card sits at 39C under load, but the upper card is now at 68C. I use an EK serial terminal to connect the graphics cards, the water enters the lower one first and the goes to the upper card. Therefore, I expected that the upper card would run a bit hotter, but not 27C hotter. Any ideas why that is? could it be that my pump is too slow or has not enough pressure? 

I have a few ideas. There are air bubbles in your loop. Or the two you suggested. also If your coolent has a low heat capacity it will heat up very quickly in your loop. 

[PhysicalVocalist]

Well for starters, the loop order doesn't make a difference for fluid temperature, so the second card shouldn't technically be any hotter as all the fluid in your loop will be heating up nigh-simultaneously.  That being said, a 27C difference is DEFINITELY a sign that something is up.  I don't have much experience with GPU liquid cooling, but I would maybe double check that the top card has the water block securely in place, and is making good contact.  Apart from that, maybe check to see if you have it running at a higher clock speed?  

[Davidwr80]

Ok, so you think there is air left in the GPU blocks or in the loop in general? how do I get air bubbles out of a GPU block? I use EK-CryoFuel

[Davidwr80]

1 minute ago, PhysicalVocalist said:

Well for starters, the loop order doesn't make a difference for fluid temperature, so the second card shouldn't technically be any hotter as all the fluid in your loop will be heating up nigh-simultaneously.  That being said, a 27C difference is DEFINITELY a sign that something is up.  I don't have much experience with GPU liquid cooling, but I would maybe double check that the top card has the water block securely in place, and is making good contact.  Apart from that, maybe check to see if you have it running at a higher clock speed?  

Ok, I thought that for a parallel setup there has to be a temp difference. I did check both GPU blocks last night, they both make full contact with all the important parts. But the PCB of the top card flexes slightly where the power connectors are, that said, all the components are still in full contact with the block.

[Legolessed]

5 minutes ago, Davidwr80 said:

Ok, so you think there is air left in the GPU blocks or in the loop in general? how do I get air bubbles out of a GPU block? I use EK-CryoFuel

You can tilt your case around to try and get them into your reservoir 

[Davidwr80]

Just now, Legolessed said:

You can tilt your case around to try and get them into your reservoir 

If I am going to do that, I have to close the reservoir, but then the air has nowhere to escape. Are the bubbles going to stay in the reservoir if I tilt the case, even if it's closed?

[PhysicalVocalist]

1 minute ago, Davidwr80 said:

If I am going to do that, I have to close the reservoir, but then the air has nowhere to escape. Are the bubbles going to stay in the reservoir if I tilt the case, even if it's closed?

Yes.  The total volume inside your loop remains constant, the air is just moving to a different spot.  Ideally though, once you get all the air bubbles out, you may need to add a bit more fluid into the reservoir.  But that assumes that air bubbles are, in fact, the issue.

[Davidwr80]

12 minutes ago, PhysicalVocalist said:

Apart from that, maybe check to see if you have it running at a higher clock speed?  

 Both cards are running at stock speeds right now.

[Davidwr80]

Just now, PhysicalVocalist said:

Yes.  The total volume inside your loop remains constant, the air is just moving to a different spot.  Ideally though, once you get all the air bubbles out, you may need to add a bit more fluid into the reservoir.  But that assumes that air bubbles are, in fact, the issue.

Ok, I will try that, thx.

[Legolessed]

36 minutes ago, Davidwr80 said:

If I am going to do that, I have to close the reservoir, but then the air has nowhere to escape. Are the bubbles going to stay in the reservoir if I tilt the case, even if it's closed?

The reservoir is meant to have air in it so the air will just collect in the reservoir. Sorry for the late reply.

[Legolessed]

34 minutes ago, Davidwr80 said:

 Both cards are running at stock speeds right now.

Gpu boost?

[Davidwr80]

37 minutes ago, Legolessed said:

Gpu boost?

Both cards boost to 1924 Mhz and level off at 1898.

[Davidwr80]

1 hour ago, Legolessed said:

You can tilt your case around to try and get them into your reservoir 

Litte Update on that: I tilted my case in all possible directions and air came out. Quite a lot actually. I topped off the reservoir again and did my temp test. Sadly, the temps are pretty much the same. Lower card sits at 38C, the upper card at 68C again.

[baconborn]

What is the utilization of each card in real-world? 

[For Science!]

7 hours ago, Davidwr80 said:

Litte Update on that: I tilted my case in all possible directions and air came out. Quite a lot actually. I topped off the reservoir again and did my temp test. Sadly, the temps are pretty much the same. Lower card sits at 38C, the upper card at 68C again.

Have you had a look at this thread?

[Davidwr80]

2 hours ago, For Science! said:

Have you had a look at this thread?

Yes, I saw that, but since I already replaced the thermal paste on my blocks, I thought that would not be the issue. Maybe I screwed the thermal paste application two times up :3