Core and Hotspot temps problem with Gigabyte 4070 Ti Super Gaming OC

[smk03]

Hello,

I am having a problem with my new waterblock model number 13729, for the Gigabyte 4070 Ti Super Gaming OC 16GB (GV-N407TSGAMING OC-16GD).

I have core temperatures rising above 60C, with hotspots reaching 75-80C, at around 280W (My previous card, a RTX 3090, only reached 48C with a hotspot of 62C at over 360W+ with an Alphacool model 11943 waterblock!) This is also approaching the air cooler temps that was on the 4070 card, and is much worse than my previous, hotter card! (Water temp is 30C GPU load) - Thing is, the 4070TiSuper will idle at 29C with a hotspot delta of +11 degrees.

I have remounted the block twice, and after the first mount found that there was little to NO contact on any of the memory, or vrm pads that were included, (listed as 0.5mm) and in fact, there was an easily visible tiny gap between the thermal pad and the block on all these components when looking at the block with a flashlight:



I also found there was seemingly a lack of full contact on the gpu core as well (visible thermal paste edge around the die). I have ensured that I fully tightened all screws, and on visual inspection, all standoffs are contacting the PCB like they should.

On my second mount, I used 1mm pads to test on the vram and vrm components, and saw that the memory junction temp is now being reported as 42C! While the core is over 60c, so clearly core contact is a big issue with this block I received.

I have verified my card's model number and the model number of the cooler sent to me, and they match. I am afraid either I have a defective unit with improper standoff heights, or somehow the Z-height of the GPU die changed? What SHOULD the length of the standoffs be?

Please help. Thank you.

 

[smk03]

Since I haven’t received a response, I will add these images of the most recent die contact pattern:

https://imgur.com/a/C58H0uh

To me, it appears the block is not contacting the GPU die correctly, and isn’t making contact on the top left portion of the die.

Again, what are the correct lengths of the installed standoffs, and how can this be rectified?

 

[astaaa]

Hi, could you please send us a picture of your pcb where we can see the whole card?

 

[smk03]

Hi, could you please send us a picture of your pcb where we can see the whole card?

Yes, here as requested:

https://imgur.com/a/JrBhhTp

Thank you.

 

[Vanzin]

Hi,

the imprint of the GPU Core doesn´t look very good. Did you mount the water block as shown in the manual and did you tighten the four screws around the GPU DIE clockwise and not over cross, as often done? If you have done it, can you please send a photo of your GPU, where we can check the imprints of the thermal pads and their placement, too.

 

[bender]

Hi, I have the same setup...

For me the GPU is around 45-50C and the Hotspot around 50-57C. Although I'm experiencing that my VRAM temperatures are around 85-90C after 15min Fullload (GPU + CPU = Furmark + Prime95 Small FFTS).

I think with the original FAN the VRAM temperatures were better or at least in the same range - what might be the problem?

 

[Eddy]

That sounds like the RAM modules don’t have good contact with the cooler. You’ll need to check the thermal imprints.
Without looking at those, there’s not much we can say other than guess — sorry.

 

[bender]

Ok weird I'll check. The temperatures were stable overnight with Furmark + P95.

After checking for the imprints I'll need new thermal pads right?

edit:
Before I'll disassamble it - I have checked from the side and it looks good to me. Are those temperatures way off?

 

[Eddy]

No, you won’t need new thermal pads — unless the imprints suggest otherwise. In some cases, they might need to be replaced with thicker ones.
The GPU and hotspot temperatures look great, but the RAM temperatures are too high. That’s not how it should be. They’re not critical yet, but definitely too high for a water cooling setup.
Since the GPU and hotspot temps are excellent, I’d almost assume that the pads on the RAM don’t have enough contact pressure — meaning the GPU height tolerance is too large (a typical Nvidia issue).

 

[bender]

OK, but then it seems to be the best to have the thicker pads ready - right?

 

[Eddy]

That’s just a guess. Without seeing the imprints, I can’t say for sure. You might need thicker pads, but it’s also possible that you don’t need thicker ones — just harder pads.

 

[bender]

But probably I have to change them, because I don't see any other possibility to improve the contact since I'm limited by distance nuts.

 

[bender]

I have disassambled the cooler - seems that probably the 0.5mm thermal pads are too low... 0.75 would have been better.

Although after reassambly the temperatures are now at max 84C.

 

[Eddy]

Are those our pads? Ours are 1 mm, not 0.5 mm.
If those aren’t our pads, what made you decide to use 0.5 mm ones?
The spacing tolerance between the cooler and the RAM is 0.25 mm — 0.5 mm pads can't establish proper contact at all.

 

[bender]

Those are the pads from the originally packed AlphaCool Cooler which I have installed last week... it was originally packed and if I remember correctly it was also written on the pads "0.5mm Cooler Side" or something and "3mm Backplate Side". I dont have the bags anymore otherwise I would have made a picture, but I measured it.

So maybe a packaging or a measurement issue? I'd be glad if you support me in solving the temperature issue...

 

[Eddy]

We don’t have any 0.5 mm pads at all in our factory. So they can only be 1 mm pads. Are you sure you measured them correctly?
3 mm on the backplate side is correct.

Please measure the pads again carefully. If they really are only 0.5 mm thick, then something clearly went wrong during production. In that case, we would need to send out replacements — but for that, please contact our support team directly at: info@alphacool.com

Either way, it's obvious that the pads have no or only minimal contact. But it’s really important for us to know the actual thickness of the pads in your case.

 

[bender]

Ok - next time my thermal paste will be empty :/

I have used a calipers gauge and measured from the top of the thermal pad to the cooler. I have measured at the edge it is around 0.55-0.75 thick, but never 1mm, maybe also because I have now disassambled the cooler 2 times, but at least it obvious that it just has minimal contact point.

I can try to use the thermal pads from the original cooler those are 1.5mm.

 

[Eddy]

Okay, 0.55–0.75 mm is a bit imprecise, but that does seem too thin.
1.5 mm pads will likely be too thick — the 1 mm pads are designed to compress down to about 0.75 mm. You can’t compress 1.5 mm pads that much, which would result in poor GPU temperatures.

I completely understand if you’re frustrated and annoyed right now — I’d feel the same way in your position.
All I can do is sincerely apologize for the inconvenience.

 

[bender]

So what would be your suggestion?

I'd actually prefer to have the right thermal pads for my VRAM.

 

[Eddy]

Write a mail to our support: info@alphacool.com with a lilnk to this treah and a short description of the problem.