Hi @chuck_petras 

and welcome

You are right to be worried about voiding your warranty.

Generally speaking, Alienware Service techniciens are not Mod'ers and I would expect this kind of mod is not Dell's realm for gaming consumers, worlds apart I think, just like if you want to convert your GPU to liquid cooling, the warrantly will probably be voided.

Just a heads up, I once did this with a GTX 1070 and had to buy a new one. After I did the thermal pad thing, the card would crash in 3d loads within seconds.

I even re-did it twice. I think the issue was the thermal paste on the GPU getting to the side of it when I tried to remove the old one.

It's now a nice paper weight.

Also, for mining you have to under volt the card. Even thermal pads will not really save you from vram destruction.

There's a good tutorial on Toms hardware about what to do to get the temperatures in check and still obtain a good hash rate.

I'm pretty sure this would void the warranty on the Dell OEM card. Some third party manufacturers might allow for repasting, but changing components and repasting I'm pretty confident would void the Dell warranty. 

I've repasted and swapped out thermals pads and paste a number of times without issue. Never spent close to $60 on pads though.  

Be careful as the die can chip or crack if the heatsink is overtightened or removed unevenly. That would brick the card. 

Also, overapplication of electrically conductive paste can also fry it. 

Perhaps two reasons why self-servicing of the GPU's internal components is not (like 99% confident here) covered under warranty. 

Having the same issue, have 2 80mm fans sitting top on the card, keeps it stable under 106.

Hello Plaidjedi,

Could you please post a photo of your supplemental fan setup?

I'm considering doing something similar, but 106 still seems high.

May go the therma pad route + back fan

This was very interesting about 3000 Series cards and vram temps. Also if your a miner.... I’m not. But while gaming/DCS my vram on my 3090 sits around 85-92c. Idles @ 72-74c

https://youtu.be/YklybEdoKIM

Also he said that the Noctua fan he used through a bios error, bypasses the error by hitting ESC.

Ty for posting that video - super clear and crazy impressive results.

Your welcome ! It’s that everyone is doing thermal PADs where he clearly IMO did the right fix out of so many videos I have seen with this vram issue by applying very little paste.

yeah I did the thermal paste after removing the serial stickers as well on my Dell 3090. Similar results with a 10-12 C drop, but still not great.

My 3080 FE saw a 15-20 C drop from pads and paste replacement. It is crazy how they can’t come from the factory with these simple improvements.

I don't think its nothing to do with Dell/Alienware factory. All nvidia 3080 and 3090's have high vram issues and I blame the design flaw on the engineers. Im just gonna run my 3090 until it smokes and Im really ... really tried of peaking at the vram temps too. I mean I'm not even using half of the memory on this card so was this a waste of money ?   Heck no !  lol !

I have Dell 3080 and Dell 3090.

I have no luck with 3090.. No matter what i did , I can;t get the temp below 104c.

1st  0.5mm on 4 sides of copper, repaste chip  with kyronaut

remove the barcode sticker . 1.5mm on backplate  temp 104c 12x mh

2nd try removes the 0.5mm, thermal paste with Mx4 on 4 sides of copper, kyronaut the chip

1,5mm thermalright on backplate, left of gelid pad on center chip (backplate side)

104c

3rd try, remove only backplate, repad with 1.5mm pad direcltly on mem, retain the gelid .

temp hit 106-108c.

Are you using MSI afterburner for the gpu fan control? It will also show you if both gpu fans are operating. Also a good program to use for monitoring gpu is GPU-z free program as well. You need to check if both fans are working. Maybe take the side panel off a boot a game up to see if both fans are spinning. That temp is really high for gpu.

on my 3090’s it’s 1,5mm on 3 and then on the right four chip side half the chip is 1.5mm and the other half of that chip bank is 2mm.  (The side that does not have the copper over your pads)

Also put your pads pattern side up, put the backplate back on tight. Run it an hour or so watching memory junction temps.  Remove backplate very gently again and you should no longer see any of the thermal pad pattern.   If you see pattern that means the backplate isn’t touching those spots.  Whoever designed these was super not smart heheh.