Issues w/ Alphacool 5080 Reference Block 10248

[labizoni]

Hi folks!
So I've installed the 10248 block in a 5080 Shadow, which shares the same pcb with of Ventus cards and also, the Shadow is listed as compatible w/ the 10248.
However, there's a issue. The pcb warps once the block is installed.

If the nut standoff makes contact with the pcb, as per picture below...


The pbc warps a lot like the picture below.. Lucky it did not damage the card, but once the pc loaded into windows I got artifacts straight away..



If the screws are not tightened as they should be (because I believe this space in between the pcb and the nut standoff is not by design), the card warps way less..





The thermal pads are (were) the ones supplied. Once I noticed the artifacts, I OFC disassembled the whole thing wasting PTM and put together again. The pcb was still warping. Took apart again, and used putty (thinking it might be a thermal pad thickness issue), which also warped. Took apart again wasting more PTM (for the fourth time) and used very soft liard tflex hd90000 1mm and did not tight the screws well and the result is like the pictures above, a good gap in between the pcb and the standoffs and just slightly warped. - but no artifacts. I'd like to stress the fact this whole thing could have damaged the card.

So, what to do?

 

[Eddy]

The first step should have been to contact support directly before attempting any modifications or troubleshooting on your own. This way, assistance could have been provided immediately, or the necessary information could have been requested, which is now missing. The pictures provided are not very informative for us, as they only show the issue in specific areas. A photo of the entire PCB, including the applied thermal pads, and a measurement of the height of the standoffs would be essential. The support team would definitely need this information. Could you please send this to info@alphacool.com? We are not familiar with such an extreme case.

The tolerances for GPU height with Nvidia are extremely broad, far exceeding the tolerances specified by Nvidia. However, it still should not look like this. We had the card in-house; such an issue would have been immediately noticeable.

 

[labizoni]

Hey @Eddy hope you are well.

I did open a ticked and also posted here afterwards.

No modifications were made. The card was warped so I took apart and then I needed to use diferente thermal pads(1mm as well)/putty as the original ones were not usable anymore.

I'm going to send the pictures I have to the email provided.

 

[Eddy]

Why three email addresses? You only need to send it to info@alphacool.com, unless additional colleagues were involved from there.

I believe you when you say you haven't modified anything. In the end, it's all about understanding where the problem lies. That's why the pictures and information are important. We've had various cases where there were suddenly different PCBs, or tolerances were completely off, or the cooler actually had the wrong spacers attached. Although these spacers should be the same length in all coolers. What we definitely know is that Nvidia is not keeping to its own tolerances for GPU height. It could be off by as much as 0.6mm, which makes a big difference. But in your case, it seems like much more.

But if you've already sent a ticket, I'll stay out of it to avoid any chaos in communication.

 

[labizoni]

That's a typo, I meant "to the email provided", I've corrected it.
Course, no modification was done. A lot of thinking yes but I couldn't get to a consensus about what's wrong.

Once I get an answer from the ticket I'll update this.

 

[labizoni]

As I said I would update this, here we go..
Well, support via the ticket was not helpful at all, which honestly I would not expect that. I'm not asking for a "personalised support experience" but no intent to sort the issue was shown apart from "you can use 1.5mm pads", which if any of my emails were read properly anyone with two neurons working would say that would just increase the gap in between the pcb and the standoffs because by analysing just the bare card against the block is possible to see the standoff are far away from the pcb.

Anyhow, moving on..

I've sent that card back, and purchased another one (tried my luck).. Took the whole think apart and placed the pcb on top of the block with nothing in between them. Not even a little bit of paste left, no residue of thermal pads left on the pbc as well. Just the bare pcb and clean block.
While applying pressure right on the opposite side of the core I could clearly see the standoffs again don't make contact with the pcb, which causes the card to warp and would warp the new card as well. Is possible to see the gap in the picture, but in real life the gap looks bigger than that.



As mentioned above, and tested, 1.5mm very soft thermal pads did not work with the previous card, and it was not going to work with this new one too, no thickness of thermal pad would work if the bare card + block show gaps in between standoff and pbc. The only possible reason for that if the gpu core tolerance is higher than what it should be.. However, MSI confirmed via they support the Shadow and the Ventus share the same pcb, same tolerances, same everything.

Anyway, I need to close this gap so I used 0.5mm washers from a Bykski 6900XT block I still have here, trimmed couple of them to do no overlap some caps..



Applied putty in some of the vram and vrm just to test, and used a thin layer of thermal paste and assembled the card using the screws of the Bykski 7900XT I also have here, because the head of the screws is a little broader than the Alphacool's ones and they would not sink below the surface of the back plate.
At this time, with washers and different screws (thanks to STS @ YT for the tip as he also had issues with a alphacool block) the card mounted perfectly flush, no warp across the card, having the same distance across the card in between the back place and pbc also in between the pcb and the block, and with the screws tightened to a satisfactory torque, unlike before.



Opened up the card to check the imprint of the putty and thermal paste.. All nice and squished with the core making contact (+0.5mm offset from the washers) with the block.





Then reassembled the card using 0.17g of putty for each of the vrms around the card and 0.65g for each of the vram + ptm in the core.



And bingo.
Steady 46-49C in the core and 44-45C in the vram with no warp, solid mount, good toque in the screws, pushing 400W with core at 3.2ghz and vram at 2.8ghz.

 

[labizoni]

@Eddy

 

[Jackysi]

In my case, I cannot see any bending. I do, however, have an issue after 3-5min gaming card crashes. Like, there isn't enough pressure on the GPU die.