in

RTX 3090 FE thermal pads and paste replaced + lessons learned + results

​

[NZXT H510, RTX 3090 Founder’s Edition, Intel i9 10900K, Gigabyte z590 Aorus Elite, G.Skill Trident Z Royal 3600mhz, EVGA Supernova 850w G5 Gold](https://preview.redd.it/dtejl7ux22b71.jpg?width=1152&format=pjpg&auto=webp&s=9e3d564a04c1e596bda1b9dd2864491d43b14d77)

Got a pre built computer from NZXT that has an RTX 3090 Founders Edition. The main purpose of the computer is for gaming but with etherium mining being hot, i figured I would mine some etherium when i’m not gaming. I maybe play once a week at best so it’s mostly just a mining computer for now.

Soon as I started mining, i realized the memory junction temps were through the roof at 110 degrees celcius! There is no way I wanted to let it mine with those temps, so I ordered the Odyssey 1.5mm thermal pads to try the thermal pad mod. I was hesitant on opening up the card but with so many people having success, I figured, how hard can it be? lol.

After a few hours, I had thermal pads replaced along with new thermal paste (Thermalright TF8) for the GPU. Fired it up and my memory temps were stable around 98-100c. The gpu core was at 65-67 degrees and my mining rate was around 118 MH/s. I thought that was pretty good but my gpu temps went up. I wasn’t too happy about that. I waited 2 weeks thinking that that paste may need to cure a bit and might improve but it never did.

I took the card out and opened it back up again to re-apply thermal paste again. This time I ordered Thermalright TFX. When I took the PCB out to reveal the thermal paste, I realized it didn’t squeeze out at all like the factory paste when I first opened the card. When i had applied the TF8, I had spread it on like I had previously done on CPUs. Big mistake. Apparently the pressure on the GPU isn’t as high as coolers on CPUs even with the screws tightened all the way. This time, I applied a lot more thermal paste using the X method. Put the card back together and the GPU temps were in the 55-57 range. Thats a 10 degree improvement!

Unfortunately I made a dumb mistake when putting everything back together. I firmly pressed on the card cover plate to make sure “everything was tight”. That caused the plate to flex a little. This flexing indented the thermal pads and since the pads don’t inflate again after being pressed, it must have left a gap between the plate and the thermal pads. My memory temps were back to 110.

I ordered pads again, this time I got the Gelid 1.5mm Ultimate pads because they had a slightly better thermal conductivity rating than the Thermalright Odyssey pads. Took the card apart for the 3rd time and noticed the thermal compound just barely squeezed out from all sides similar to the factory paste. That was a great sign, since I must have applied just the right amount of paste. I put the new pads in place and fired up the card again. I made sure to not press on the memory pads at all! I simply screwed everything nice and tight without pressing down on anything. Memory temps were in the 94c range while the GPU was still a cool 55-57 degrees. Success!

One more thing to mention, when applying thermal pads, make sure they’re not oversized, specially near the edge where the cover edge sits. If the edge of the cover sits on some extra thermal pad, it won’t apply proper pressure on the thermal pads that are over the important areas. I’ve included pics of how much thermal paste I put on the GPU along with the thermal pad layout on both sides.

I also read somewhere that you shouldn’t cover the resistors (R005) with thermal pads so i avoided them when applying the pads. Not sure the reasoning behind this but I figured its not something that gets unnecessarily hot so I made sure to avoid them. So far everything is working as expected.

**Thermal Paste:** Thermalright TFX on the GPU (X application method)

**Thermal Pads:** Gelid Ultimate 1.5mm

**Memory Overclock:** +1100mhz

**GPU Overclock/Underclock:** 0mhz (Underclocking made zero difference on power draw but caused instability with cuda core occasionally crashing)

**Power:** 83%

**Memory Junction Temps:** 96c

**Power Draw:** 290 watts

**Efficiency:** 415 KH/w

**Fan Speed:** Target 62c GPU core (Results in 50-55% fan speeds which keeps things silent!)

**Mining Rate:** 120.5 MH/s

I recommend starting with a +1000mhz clock on the memory and bumping it up 50mhz a day if you’re not crashing. I saw some minor instability when going over 1100 so I stopped there. You might max out at +1,000 or get lucky and reach +1,300 without crashing. Keep an eye on your MH/s to see if you’re actually getting a difference in your mining rate with each bump in memory clock. The mining rate delta is very small with each 50mhz increase. For example, at +1,000 memory clock, my mining rate is 119.1 MH/s which is only 1.4 MH/s lower than at a memory overclock of +1,100. In the end, stick with the clock speed that provides you with the most stability. This isn’t a sprint, its a marathon.

I set all my overclock settings (fan, power, memory, gpu, etc.) in the t-rex config file just to keep everything in one place. This also basically sets the clock settings only when the miner is running so I can just turn the miner off and game with the factory clock settings. Targeting a GPU temp of 62c keeps my fan speeds at about 52%. I think this is great considering I have an nzxt h510 case which isn’t the most airflow friendly case. The computer is in my bedroom and is dead silent. Its rock solid with no GPU crashes or reboots for days.

TLDR;

– Use more thermal paste than you think you need and don’t spread it!

– Don’t press on the thermal pads when you lay them on the memory chips

– Don’t press on the cover when you’re putting the card back together. You don’t want to indent the thermal pads more than what the screws are doing already.

– Don’t underclock the GPU core, its useless when mining etherium and can cause instability.

– Don’t oversize the thermal pads.

​

[RTX 3090 FE with Gelid 1.5mm Ultimate thermal pads and Thermalright TFX paste (GPU side)](https://preview.redd.it/k8muerq8l1b71.jpg?width=1536&format=pjpg&auto=webp&s=a3fde2f39f7dbdbdee36ba60c40a281b7ea4f839)

​

[RTX 3090 FE with Gelid 1.5mm Ultimate thermal pads (back side)](https://preview.redd.it/c1wdhuz9l1b71.jpg?width=1536&format=pjpg&auto=webp&s=4aa1b528d7c840e69bdae11e1b201f2d76db4b74)

​

https://preview.redd.it/sqdy0yabl1b71.png?width=803&format=png&auto=webp&s=11285398c650d8af962d729697be7dce859ce955

What do you think?

Leave a Reply

Your email address will not be published. Required fields are marked *

GIPHY App Key not set. Please check settings

12 Comments

  1. I’ve never done anything like this, even spreading thermal paste on a cpu. Do you think I’d be able to do it if I take my time? How difficult is it? There is a guy who I can pay $150CAD to do it all for me, but that seems expensive

  2. Thank you SOOOO much for this post. I have the 3080 FE (the models before the hash limiting happened) and I was wondering why my core temps shot up by 10 degrees. I even opened up the card again and repasted with slightly more thermal paste but no effect. However one mistake I made which I only just realised after reading your post is that I spread my thermal paste evenly across the die area and not in the X pattern which you have done and ended up giving you great results.

    Will definitely try this soon when I get the time. My GPU currently sits at 47C on 80% fan speed and all the windows open in my room and a case without the side panel attached. I’ll probably be able to get around 40-45C on 70% fan speed or even lower.

  3. I have done this now for 1 x 3090 FE, and 5 x 3080 FE cards. I also used the Gelid Ultimates… I use 1.5 mm on the GPU side, and 2.0 mm on the back side.

    Note: I had a card that after swapping the thermal pads still had a high VRAM temp… when I opened the card back up (for the 3rd time) I noticed no indention for some of the memory chips on the side close to the PCIe slot. I ended up putting a couple of 2.0 mm thermal pads on those that were not making good contact and it helped tremendously.

    My highest VRAM temp is now around 92 degrees C…. and my lowest is 76, 2nd lowest is 78.

    Good job.

  4. I’m curious as to why the resistors shouldn’t be covered. I mean I understand the reasoning of better safe than sorry, but is there an actual fact based answer?

  5. Try locking core clock, if it’s not already my 2070 saw a massive rise in stability and temp, dropped 8c and 40watts in power, E rose from 300 to 420 W/Kh.
    Initial -800 on core and +800 on mem, power limit 65%
    T Junction sat at around 105c before I locked my core
    (mined like that for a month like a utter moron)

  6. Having replaced the thermal paste and pads on two 3090s, I disagree a little on the spread part.

    You said don’t spread the thermal paste, because this made your temps bad. Not sure how spreading it could fuck it up, as I have spread both times, and on dousens of other gpus, and gotten good core and hotspot temps. This sounds more like user error to me, spreading should be fine, just make sure to give enough paste and cover the die properly.

BRMT Airdrop

➡️ What are differences between APY & APR ??