High core/hotspot on Eisblock Aurora Acryl RTX 4080 Strix

kubko

New member
Hi,
I recently installed waterblock on 4080 Strix. I followed instructions but made one difference. On memories instead of supplied pads I used Thermal Grizzly Minus pad Advance, however they are same thickness - 1mm. All other pads on the PCB are the one supplied. What I noticed is that TG pads are pretty stiff and maybe not very compressible like the pads supplied with block.

My flowrate was about 100l/h with 25% pwm on VPP Apex pump
Water temperature did not raise over 27C

So the behavior with temperatures I've observed, higher wattage draw makes GPU's core temperature grow exponentially. For example if my GPU pulls 250-300W the temperature is around 58-65c what is slightly worse than Strix air cooled. However once my GPU gets to 350W I was seeing 70-75C (core not hotspot!) what I think is really high. What is interesting is that my memory peaked at around 45C what I think is really cool. Do you think my memory pads could cause issue with die pressure even though they are same thickness? Are here any specific pads you recommend, or should I try order some putty for memories?

I'm dumb I didn't take photo, because I tried to remount the block because I thought something was wrong with paste application or screw pressure but got same result. Couldn't evaluate paste contact on die, however on pads I could see memory chips imprinted on surface really well.
 
Last edited:

Eddy

Iceman
Staff member
1 mm pads are not all the same. There’s a reason ours are extremely soft. If you use harder ones, the contact pressure won’t be correct. The pads need to be able to compress down to about 0.7 mm. Their Shore hardness is 25 Shore 00. If the pads are harder than that, it’s no surprise you’re getting poor temperatures.
 

kubko

New member
Didn't know there are so many variables with pads sometimes its PITA to even find correct thickness leave alone hardness, my bad.. I wanna try Honeywell HT10000 putty, that should compress perfectly on memories. Rest of the pads should be fine as those are all that came precut with the block.

Once I get to draining loop and disassemblying, I will update with photo of cooler to evaluate if I had poor contact on the die. Thanks for now!
 

Eddy

Iceman
Staff member
Putty is no problem. It’s very soft and adapts well to everything. But with pads, it’s not just the thickness that matters, the hardness is important too. For GPU coolers I would always choose pads that are as soft as possible. Hard pads are not an advantage here.

And with the RTX 3000 and RTX 4000 series, there is the added issue that GPU heights can vary a lot and often even fall outside Nvidia’s own specifications. The height differences are usually around 0.2–0.4 mm, which sounds tiny at first, but it already makes a huge difference.

That’s why soft pads are always better for GPUs. In the end, the thermal conductivity barely matters.
 

kubko

New member
Hello! Just a follow up post. Finally got to draining my loop and tearing GPU apart. I'll continue with loop assembly tomorrow, but I took some pics after cooler disassembly. Does it look like my core had not the best contact? Also I realized one thing, I had my inlet on incorrect port 🤦‍♂️ not the one that goes through jetplate but the second one that should've been an outlet. Not sure how much that one could affect temps though.. I put PTM7950 and putty, hopefully the contact will be good this time

 

Eddy

Iceman
Staff member
Does it look like my core had not the best contact?
Frankly, that’s hard to say. You’ve completely drowned the GPU in thermal paste, but it does look a bit uneven. Swapping the IN and OUT ports makes a noticeable difference in cooling performance. Depending on the graphics card, heat output, and chip size, it can be anywhere between 5–10°C. I can’t say exactly how much it would be on a 9070 XT, but I’d estimate you’re losing at least 5–8°C under full load.
 

kubko

New member
Okay, very good news 😎 used PTM7950 with putty on memories/vrm, reassembled the loop with correct in/out orientation and the magic happened, during the same benchmark (Timespy Graphics test 2) now I hover around 55c max, while before I was peaking at low/mid 70's under same benchmark.. GPU has no problem sustaining 350W now, I tried to run Furmark, that one pulled around 420W (thats the strix power limit) and temps stayed in high 50's, that's crazy wattage for 4080, so I assume I have very good cooler contact now, hotspot delta is around 10c too, thanks!

Now I just need to do some savings to get another Alphacool 360 rad on top to expand thermal headroom coupled with some nice fans and my build will be perfect 😁
 

Fruchti

Member
Okay, very good news 😎 used PTM7950 with putty on memories/vrm, reassembled the loop with correct in/out orientation and the magic happened, during the same benchmark (Timespy Graphics test 2) now I hover around 55c max, while before I was peaking at low/mid 70's under same benchmark.. GPU has no problem sustaining 350W now, I tried to run Furmark, that one pulled around 420W (thats the strix power limit) and temps stayed in high 50's, that's crazy wattage for 4080, so I assume I have very good cooler contact now, hotspot delta is around 10c too, thanks!

Now I just need to do some savings to get another Alphacool 360 rad on top to expand thermal headroom coupled with some nice fans and my build will be perfect 😁
Very clean PC:love:
 
Top