2/16/21 Update at bottom!
TL;DR I still DO NOT RECOMMEND MESSING WITH THE HEATSINK!

*Despite having been an EVGA member since my 1080 Ti FE I realize as a new forum poster I'm restricted from posting links for XYZ.
Once, I have the ability to post links I'll edit my reference materials and the album back into the post.

Hello all,

After picking up some casual Nicehash mining, I've decided on changing out my pads and paste based on a few different threads discussing and executing this mod.
In short, based on my own choice of thermal pads and the time it took me to install, this mod is largely unnecessary for the 3090 FTW3 and may even worsen your general temps depending on your pad compression. Not by a whole lot, but enough to reconsider the entire effort. So if you've tried this and were largely successful, I'd love to know what was different in your mod!

Apologies for not being more scientific about this and taking actual standardized benchmarks so in lieu of those please my reference my VRAM temps during Nicehash Excavator daggerhashimoto mining.

Materials used:

• Thermal Grizzly Kryonaut
• Fujipoly/mod/smart Ultra Extreme XR-m Thermal Pad
  • 60 x 50 x 1.0 
  • 17W/mK
  • GREY in photos

• Thermal Grizzly Minus Pad 8 Thermal Pad
  • 100 × 100 × 2.0 mm
  • 8W/mK
  • BROWN in photos
• "AI AIKENUO" Thermal Pad
  • 100 x 100 x 2.5mm
  • 6W/mk
  • BLUE in photos

Posts referenced:

• [will add link when permitted]
• [will add link when permitted]
• [will add link when permitted]

Photo album with minor captions (expanded upon in this post):

• https://imgur.com/a/LlzY0Vm

Stock Photos

• 
  • EVGA Uses very soft thermal pads and thermal putty in order to bridge the unfortunately uneven gaps between the PCB and heatsink, according to product manager EVGA_JacobF, "Memory is 2.25mm, VRM 2.85mm, back VRAM is 2mm." [will add link when permitted] These were soft, and very stuck to the board making a great deal of contact.
  • 

Mod Photos

• 
  • As you can see, grey pads are Fujipoly 1mm, brown are minus8 2mm, blue are the "AI AIKENUO" brand 2.5mm pads. In the side view it looks like they all make quite good contact. Unfortunately wrong.
    • Fujipoly do not compress much if at all, but are very pliable
    • minus8 are the softest of the three, but the stock thermal pads are still quite a bit softer and much tackier
    • AI AIKENUO are in between and when used on the backplate below the MOSFETs (which I realized after the fact were not VRM) result in temp throttling due to not allowing the backplate to compress the VRAM pads on the back enough. It's strange because the overall temps are better, but due to the insufficent backplate compression I guess it causes the internal readings to be much hotter than what is reported, resulting in thermal throttling according to both Nicehash and GPU-Z per this article [will add link when permitted]. 

• 
• After removing the blue 2.5mm pads on the backplate, the thermal issues subside but backplate compression is still comparatively unimpressive despite EVGA_JacobF's 2mm pads for VRAM back suggestion and other users in my referenced post using 3mm pads on their VRMs! While 2.5mm on mine cause the plate to bulge the slightest bit, enough to upset the card and cause it be thermally throttled.
  • Applying a wood dowel on each the backplate quadrants with some force to compress the panel in results in a 4C improvement! For this reason, I am slightly considering bending my backplate the slightest perceptible bit so it is concave enough when screwed down to apply the necessary force. Alternatively, any suggestions for soft 2.5mm pads that allow for good compression matching the 8W/mK minus8 pad spec's would be appreciated!

So for the more adventurous, what would you have done differently?

Updates:

• 
  • Placed 14x14x6mm aluminum finned heatsinks over VRAM and heatpiped areas. In conjunction with the removal of 2.5mm VRM/backplate pads, VRAM/backplate pads were able to make contact and temps are now a few degrees hotter than stock, and peaks at 76C VRAM temps due to thermal throttling in Nicehash. I've also asked EVGA Tech Support for the brand of pad/paste/putty used during assembly in case I'm unable to meet or beat stock temps.
    • Unfortunately, EVGA cannot provide a manufacturer for replacement TIM, nor can they accept cards for TIM reapplication.
• I've ordered some Thermalright Odyssey Thermal Pads in varying thicknesses to undo my poor TIM job.

2/12/21 Update:

• Just about 2 weeks later, the Odyssey pads from AliExpress have arrived!
• GPU now set up accordingly:
  
  • Thermalright TFX (14.3W/mk) on GPU die
  • 2.0mm + 0.5mm stacked Odyssey Pads (12.8W/mk) on VRAM to meet minimum 2.25mm spec per EVGA_JacobF
  • 2.0mm + 0.5mm stacked Odyssey Pads (12.8W/mk) on BACK VRAM for adequate compression as 2.0mm minus8 Pads only made about 40% contact based on VRAM impressions left on the pads.
  • 3.0mm Odyssey Pads (12.8W/mk) on VRM lines to meet minimum 2.85mm spec per EVGA_JacobF
  • 1.0mm Odyssey Pads (12.8W/mk) on MOSFET lines to meet minimum 0.85mm spec per EVGA_JacobF
• Due to the increased thickness in the VRAM pads, the port-side MOSFET line does not make enough contact to the heatsink. The temps are largely inconsequential and under 80C so I'm not worried, but it is a pain to look at.
  
• I can now mine using NiceHash as before with good temps below 72C and an increased hashrate with power increased to 71-72% but still nowhere near stock heatsink TIM level temps. 
• With the card now redone entirely Thermalright Odyssey pads and TFX paste my temps have largely improved when it comes to mining from the previous attempt, but have unfortunately gained over 10C increase in GPU1 / GPU die temps when gaming. Cyberpunk 2077 Luxury apartments fast travel point was 73C prior to this, 83C now. I suspect the uneven TIM thicknesses results in less than ideal GPU die / component cooling one way or another since there are no reputable 0.25mm thermal pads of similar W/mK quality to balance things out as far as I'm aware of.

For now I am emotionally finished with this project. I once again ask you to reconsider redoing your heatsink/PCB side TIM, and see how adding heatsinks to the backplate work for you instead. These temps are well within safe ranges, but given the time, money, and effort I've put into just one card for overall less than satisfactory temps I'm throwing in the towel lol. 

This will be my last update for a while until I can get some solid suggestions on what TIM is the absolute best for the FTW3 heatsink design to remedy my imperfect solution.
EVGA, I would gladly pay you to repaste/repad/reputty my card and in the future I hope you can find a way to standardize the distance between the components and heatsink. 
Thank you all again for reading and giving input if you can, I look forward to fixing this in the future! 
 
 

I have a KPE and it looks like we have the same stock pads & compound.  On a whim I quickly swapped out with some GELID and even though GELID is softer than Fuji, it still wasn't ideal.  Although I cannot accurately measure the performance, I will suggest that compared to my EVGA 1080 ti KPE, the current generation of EVGA thermal compounds are pretty good and we are more likely to get better numbers from other solutions such as active cooling on the back.  I may revisit this later on but I only recommend checking the thermal paste on the GPU, mine wasn't making the best contact.

Awesome, thanks! I have an order of minus8 0.5mm pads I’ll be adding on top of the 2mm for the backplate and hope that will make the difference. Will post the update to main post when they arrive.

_Gir_
I have a KPE and it looks like we have the same stock pads & compound.  On a whim I quickly swapped out with some GELID and even though GELID is softer than Fuji, it still wasn't ideal.  Although I cannot accurately measure the performance, I will suggest that compared to my EVGA 1080 ti KPE, the current generation of EVGA thermal compounds are pretty good and we are more likely to get better numbers from other solutions such as active cooling on the back.  I may revisit this later on but I only recommend checking the thermal paste on the GPU, mine wasn't making the best contact.

After many attempts to solve my issue of compression, I've confirmed that for the 3090 FTW3, EVGA's solution is basically the best as far as practicality goes. The pads and paste I used, while performant on paper, lack the tackniess the stock EVGA solution(s) have that not only provide ideal compression, but a significant amount of adhesion to maintain that compression. Fingers crossed EVGA tech support can provide me with some pastes, pads, and putty manufacturers so I can make it "stock" once again!

Great post and pictures.
Wish I could buy a bunch of thermal puddy pre cut from evga!

you VRM area of the card you need 3mm. 
 
2.5mm doesnt enough for back VRM area. I am about to buy GC-Extreme 3.0mm for back. We can see up to -7c drops

streamroller
you VRM area of the card you need 3mm. 
 
2.5mm doesnt enough for back VRM area. I am about to buy GC-Extreme 3.0mm for back. We can see up to -7c drops

In my case the back 2.5mm worsened my VRAM temps by a couple degrees and became thermally throttled despite VRMs having the up to -7C improvement. Removing them and adding some short 14x14x6mm fins to the backplate as a cautionary measure brought me to within 2-3C worse than stock temps. I believe if I put 3mm for VRM it would worsen the VRAM/backplate pad compression as 2.5mm VRM/backplate pads flexed the plate away from VRAM/backplate enough that I could visually see compression if I squeezed it enough. I do have Thermal Grizzly minus8 0.5mm pads on order so I hope placing those between VRAM/heatsink to make the 2.0mm VRAM pads to 2.5MM will alleviate the issue seeing as 2.25mm is the EVGA spec (assuming the core bracket is able to provide enough compression to push past the extra 0.5mm for adequate core contact. That said I'm sure theres some manufacturing tolerances to consider, and/or I just have the wrong set of pads for my card entirely!  

May I know what you've done and the temp improvement youve seen overall, assuming you also have a 3090 FTW3 Ultra?

Hoopladoodle

streamroller
you VRM area of the card you need 3mm. 
 
2.5mm doesnt enough for back VRM area. I am about to buy GC-Extreme 3.0mm for back. We can see up to -7c drops

In my case the back 2.5mm worsened my VRAM temps by a couple degrees and became thermally throttled despite VRMs having the up to -7C improvement. Removing them and adding some short 14x14x6mm fins to the backplate as a cautionary measure brought me to within 2-3C worse than stock temps. I believe if I put 3mm for VRM it would worsen the VRAM/backplate pad compression as 2.5mm VRM/backplate pads flexed the plate away from VRAM/backplate enough that I could visually see compression if I squeezed it enough. I do have Thermal Grizzly minus8 0.5mm pads on order so I hope placing those between VRAM/heatsink to make the 2.0mm VRAM pads to 2.5MM will alleviate the issue seeing as 2.25mm is the EVGA spec (assuming the core bracket is able to provide enough compression to push past the extra 0.5mm for adequate core contact. That said I'm sure theres some manufacturing tolerances to consider, and/or I just have the wrong set of pads for my card entirely!  

May I know what you've done and the temp improvement youve seen overall, assuming you also have a 3090 FTW3 Ultra?

i have 3080 ftw
 
I am about to do this mod aswell but at least 3 confirmed cases uses 3mm pads to BACK VRM area not VRAM area on 3080. So you say 2.5mm make backplate bend ?

Keep in mind I have Silent Wings 3 12cm (slapped MLCC area of gpu) as exhaust  spinning at 1600rpm. Even without pads backplate gets hot and i do move a lot of heat generated from gpu itself. My GPU1 and 2 temps are same.

streamroller

Hoopladoodle

streamroller
you VRM area of the card you need 3mm. 
 
2.5mm doesnt enough for back VRM area. I am about to buy GC-Extreme 3.0mm for back. We can see up to -7c drops

In my case the back 2.5mm worsened my VRAM temps by a couple degrees and became thermally throttled despite VRMs having the up to -7C improvement. Removing them and adding some short 14x14x6mm fins to the backplate as a cautionary measure brought me to within 2-3C worse than stock temps. I believe if I put 3mm for VRM it would worsen the VRAM/backplate pad compression as 2.5mm VRM/backplate pads flexed the plate away from VRAM/backplate enough that I could visually see compression if I squeezed it enough. I do have Thermal Grizzly minus8 0.5mm pads on order so I hope placing those between VRAM/heatsink to make the 2.0mm VRAM pads to 2.5MM will alleviate the issue seeing as 2.25mm is the EVGA spec (assuming the core bracket is able to provide enough compression to push past the extra 0.5mm for adequate core contact. That said I'm sure theres some manufacturing tolerances to consider, and/or I just have the wrong set of pads for my card entirely!  

May I know what you've done and the temp improvement youve seen overall, assuming you also have a 3090 FTW3 Ultra?

i have 3080 ftw
 
I am about to do this mod aswell but at least 3 confirmed cases uses 3mm pads to BACK VRM area not VRAM area on 3080. So you say 2.5mm make backplate bend ?

Yes, in my experience the 2.5mm VRM-to-Backplate was tall enough that the EVGA-spec 2mm VRAM pads were unable to make contact. Here is a short video I made for my friend showing the lack of compression between VRAM-to-Backplate 2mm pads, which I think is caused by the 2.5mm VRM-to-backplate pads that DO NOT come with FTW3 cards stock. You can view this via imgur.com as I'm unable to post links for the time being: /7nSozeY You could also try opening the broken image file below and open in new tab to load the imgur video.