GPU Temp with Waterblock, not Yet Connected to Loop

[Cavalry Canuck]

Good Day,

 

I put a water block on a GPU. It's my first time using Liquid Metal as a TIM on a GPU (have done a couple CPUs now).

 

I didn't want to go through the hassle (again) of installing a watercooled GPU into a loop, only to find that the TIM application was insufficient. I don't have a second pump to build a mini-loop with to test it properly prior to rigging it in to the main loop. I was hoping that running it with just the water block would offer some indication of whether everything was good to go or not. Information is scarce, and what little I've read suggests that GPUs with water blocks will idle as low as 35c at a dead idle when not connected to a proper loop.

 

Card: MSI GTX 1080Ti Gaming X

Block: EKWB EK-FC1080 GTX TI TF6 (I re-attached backplate using longer screws than what came with the kit.)

TIM: Thermal Grizzly Conductonaut.
Temps: Settled at around 63C after writing this post and otherwise being allowed to idle for 10 min.

Does this sound like I did it well enough to confidently proceed to installing and testing it on the main loop?

 

Thanks, Matt

[An0maly_76]

No circulation = no cooling. I wouldn't do it unless the original fans can still cool it. So I wouldn't run it long, but if it never peaked 63C, sounds like it should be okay.

Edited June 27, 2022 by An0maly_76
Revised, more info

[RONOTHAN##]

1 hour ago, Cavalry Canuck said:

It's my first time using Liquid Metal as a TIM on a GPU (have done a couple CPUs now)

This is a bad idea. Liquid metal has the bad habit of getting under the substrate and destroying the BGA unless conformal coating and other insulation is absolutely perfect. Plus the performance uplifts from LM aren't that massive on GPUs and aren't worth the risks. I would clean it off and just do a normal GPU paste. 

[Somerandomtechyboi]

45 minutes ago, RONOTHAN## said:

This is a bad idea. Liquid metal has the bad habit of getting under the substrate and destroying the BGA unless conformal coating and other insulation is absolutely perfect. Plus the performance uplifts from LM aren't that massive on GPUs and aren't worth the risks. I would clean it off and just do a normal GPU paste. 

So LM can destroy bga solder balls even w a normal application on the die?

[RONOTHAN##]

1 minute ago, Somerandomtechyboi said:

So LM can destroy bga solder balls even w a normal application on the die?

LM has the nasty habit of moving after being applied to the die. It might not affect it for 6mo+, but eventually it will end up off the die unless it's specifically engineered to not to. 

[Cavalry Canuck]

On 6/26/2022 at 11:49 PM, RONOTHAN## said:

This is a bad idea. Liquid metal has the bad habit of getting under the substrate and destroying the BGA unless conformal coating and other insulation is absolutely perfect. Plus the performance uplifts from LM aren't that massive on GPUs and aren't worth the risks. I would clean it off and just do a normal GPU paste. 

I coated everything on the PCB near the die with a proven non-conductive nail polish. It would have to travel very far off the die and contact plate before it has the opportunity to come into contact with anything on the PCB. 

[For Science!]

Really hard to say unfortunately. I also would say that LM on GPU waterblocks yields very little gains and the risk is not worth it.