Easy mod for RTX cards with hot VRAM... (paste instead of pads) 3070s 3080s 3090s

Roflcopter_Down

[H]ard|Gawd
Joined
Jan 7, 2008
Messages
1,535
Reapply thermal paste to the chip and add paste the heatsink areas next to the chip. Most places say to buy pads, but ime you only need thermal paste, which everyone already has.
My cards mine much faster, and my gaming card runs notably cooler and smoother. It might void your warrenty, some cards don't have stickers so you might be able to get away with it.
I think it's important to share because almost everywhere says to buy thermal pads but thermal paste works just as well or better.
If I am missing something I'd like to know.
 

hititnquitit

[H]ard|Gawd
Joined
Apr 16, 2018
Messages
1,386
What kind of tim did you use and on what cards specifically?
What kind of contact were you seeing? Did you take any pics?
Were you spreading the tim thick, thin, average?
You have to have good contact obviously or youll cook components. Thats why we use pads and why ive asked so many questions.
 
Joined
Apr 9, 2021
Messages
748
Thermal paste gets very runny when hot. I would be concerned about it working well as first, and then leaking out over time. Kudos for taking a path less travelled, though. I'm interested in hearing about the long-term results.
 

kirbyrj

Fully [H]
Joined
Feb 1, 2005
Messages
29,090
The only way this works is with non-conductive TIM obviously. I've seen stuff like K5 Pro which is essentially thermal pad replacement paste, but I don't know that I'd use regular thermal paste in this way personally.

The only cards you need to worry about are ones with GDDR6X. Otherwise, GDDR6 runs cool enough on its own even when mining.
 

Falkentyne

[H]ard|Gawd
Joined
Jul 19, 2000
Messages
1,816
Reapply thermal paste to the chip and add paste the heatsink areas next to the chip. Most places say to buy pads, but ime you only need thermal paste, which everyone already has.
My cards mine much faster, and my gaming card runs notably cooler and smoother. It might void your warrenty, some cards don't have stickers so you might be able to get away with it.
I think it's important to share because almost everywhere says to buy thermal pads but thermal paste works just as well or better.
If I am missing something I'd like to know.
This only works with TG PP10 Thermal Putty. Absolutely do NOT use thermal paste! Regular paste will only work with shims on the VRAM or VRM's. Paste is absolutely NOT meant to fill a 0.5mm + gap between heatsink and IC's! It will perform terribly and may leak out. Use TG PP10 which is on digikey.
 
Joined
Apr 9, 2021
Messages
748
This only works with TG PP10 Thermal Putty. Absolutely do NOT use thermal paste! Regular paste will only work with shims on the VRAM or VRM's. Paste is absolutely NOT meant to fill a 0.5mm + gap between heatsink and IC's! It will perform terribly and may leak out. Use TG PP10 which is on digikey.

I'm still waiting for my sample request of PP10. I might just have to say screw it and order it off digikey.
 

Roflcopter_Down

[H]ard|Gawd
Joined
Jan 7, 2008
Messages
1,535
Didn't think to take pics. I found this from some youtube videos. I used Notura paste. From what I read only the super cheap pastes run, although maybe it is more of a crapshoot. Some people who first took apart their cards noticed the stock paste covering just the chip was already running. That putty looks really interesting however. I understand that paste isn't supposed to fill a large gap, but it's funny, the Dell/oem 3080s don't even have thermal pads, there was nothing. Gigabyte at least has some cheap looking pads...and ironically the stock Dells stay cooler in every way I can tell, compared to the Gigabytes.
I have another Gigabyte 3080 to do, so Ill take pics then.

This is what I followed mostly

There are a lot of good comments. Also for the record, I removed the old paste unlike the video =)

This video is where I heard about it first. This guy messed up applying pads and paste at first, so I went to find other peoples' experience to be sure. Iirc at the end, he ultimately recommends just paste, pads being superfluous, at least in this case.


It's been over a week and still running cool. I'll remember to check on the cards for possible paste runoff.

Not sure I will try the 3090 fe. Might try a ghetto heatsink+fan on top of the card setup.
 

Falkentyne

[H]ard|Gawd
Joined
Jul 19, 2000
Messages
1,816
Didn't think to take pics. I found this from some youtube videos. I used Notura paste. From what I read only the super cheap pastes run, although maybe it is more of a crapshoot. Some people who first took apart their cards noticed the stock paste covering just the chip was already running. That putty looks really interesting however. I understand that paste isn't supposed to fill a large gap, but it's funny, the Dell/oem 3080s don't even have thermal pads, there was nothing. Gigabyte at least has some cheap looking pads...and ironically the stock Dells stay cooler in every way I can tell, compared to the Gigabytes.
I have another Gigabyte 3080 to do, so Ill take pics then.

This is what I followed mostly

There are a lot of good comments. Also for the record, I removed the old paste unlike the video =)

This video is where I heard about it first. This guy messed up applying pads and paste at first, so I went to find other peoples' experience to be sure. Iirc at the end, he ultimately recommends just paste, pads being superfluous, at least in this case.


It's been over a week and still running cool. I'll remember to check on the cards for possible paste runoff.

Not sure I will try the 3090 fe. Might try a ghetto heatsink+fan on top of the card setup.

It's just extremely risky.
It can work if the pad thickness is <0.5mm, and its the pads that help limit the GPU core contact pressure (rather than the other way around). Because then, by removing the pads and using paste, you get better contact pressure on the GPU core to its heatsink, and then the gap space between VRAM/VRM and pad plate is even lower, which, in a circular way, end up making thick thermal paste work.
However the amount of these cards where this would work is very limited. Witness the people who tried 1mm thermal pads on their founder's edition cards and ended up with overheating VRAM because the pads weren't touching the plate, because they were too thin. Paste would be even worse, and the only option would be direct TG PP10 thermal putty, or possibly a thin custom shim + PP10 (which some people have done very successfully, but any shim requires you make sure it does NOT compromise core to heatsink die contact).

You could seriously destroy a card by doing direct paste however and I can NOT recommend anyone even try this. Look at all the eVGA 1080 Ti's which caught fire because the pads didn't make contact with the chips! Just be smart--if you want to use paste, buy TG PP10 thermal putty and don't cut corners trying to save money. If you can afford an expensive video card, you can afford $30 for putty.

https://www.digikey.com/en/products/detail/t-global-technology/TG-PP10-50/6204863
 

Roflcopter_Down

[H]ard|Gawd
Joined
Jan 7, 2008
Messages
1,535
Thanks for the input and putty shilling. I ordered some, excited to try it out. I'll take pics. Any more advice for using putty? Ive never heard of it before.
 

WilyKit

Limp Gawd
Joined
Dec 18, 2020
Messages
264
Thermal paste gets very runny when hot. I would be concerned about it working well as first, and then leaking out over time. Kudos for taking a path less travelled, though. I'm interested in hearing about the long-term results.

Not sure about all that. how do they get away with using it on the GPU die which gets hotter than the VRAM If that’s the case?
 
Joined
Apr 9, 2021
Messages
748
Not sure about all that. how do they get away with using it on the GPU die which gets hotter than the VRAM If that’s the case?

There's still a pump-out effect between the die and the heatsink, but it's much slower because there's direct contact.
 
Joined
Apr 9, 2021
Messages
748
Thanks for the input and putty shilling. I ordered some, excited to try it out. I'll take pics. Any more advice for using putty? Ive never heard of it before.

Just ordered some myself too. I read that you're supposed to roll it into little balls, put each one on the center of each component you need to cool, and then they flatten out when you mount the heatsink. I also read that you should wear latex gloves.
 

mnewxcv

[H]F Junkie
Joined
Mar 4, 2007
Messages
8,710
Some cards use 3mm thermal pads. You need pads not paste. Though I do wish board makers would make the cooler closer to the ram. No real reason to need a 3mm gap vs a 0.5mm gap.
 

WilyKit

Limp Gawd
Joined
Dec 18, 2020
Messages
264
Some cards use 3mm thermal pads. You need pads not paste. Though I do wish board makers would make the cooler closer to the ram. No real reason to need a 3mm gap vs a 0.5mm gap.
Tolerances is the reason. The more precisely machined and aligned a piece has to be the more expensive it will be to produce.
 

DaeviousMax

Weaksauce
Joined
Jun 8, 2021
Messages
110
Tolerances is the reason. The more precisely machined and aligned a piece has to be the more expensive it will be to produce.
Which since the msrp of these cards is much higher now they could easily afford to do that....too bad it’s all going into margins
 

WilyKit

Limp Gawd
Joined
Dec 18, 2020
Messages
264
Which since the msrp of these cards is much higher now they could easily afford to do that....too bad it’s all going into margins

Would you? I doubt it. Why would you eat the cost of a more expensive manufacturing process when a cheaper process works “well enough”

but that’s beside the point. The point is there is a reason.
 

Nenu

[H]ardened
Joined
Apr 28, 2007
Messages
19,917
Would you? I doubt it. Why would you eat the cost of a more expensive manufacturing process when a cheaper process works “well enough”

but that’s beside the point. The point is there is a reason.

Depends whether its exposed as a bad practise.
I'm not happy with the temp of my GDDR6x ram despite owning one of the most costly 3090 cards.
 

DaeviousMax

Weaksauce
Joined
Jun 8, 2021
Messages
110
Would you? I doubt it. Why would you eat the cost of a more expensive manufacturing process when a cheaper process works “well enough”

but that’s beside the point. The point is there is a reason.
No you are absolutely right. Invest in more expensive heatsink manufacturing or pass the increased margin to the shareholders. I’d go with the ladder, but still I’m right that they could conserve their normal margin yet make better heat sinks with the increased msrp
 

Roflcopter_Down

[H]ard|Gawd
Joined
Jan 7, 2008
Messages
1,535
You'd think...since almost all these cards have volcanic vram temps, one company could make a super extreme cooler max edition that actually has $2 of pads on the ram, market it as the one card with cool vram! Msrp could be ebay teir and still sell out.
How are gpus manufactured? Are the coolers attached by hand or machine?
edit: daeviousmax posted right before and and much more eloquently
 
Joined
Apr 9, 2021
Messages
748
No you are absolutely right. Invest in more expensive heatsink manufacturing or pass the increased margin to the shareholders. I’d go with the ladder, but still I’m right that they could conserve their normal margin yet make better heat sinks with the increased msrp
Better yet, save the slightly better heatsink for an uber-clocked version, factory OC the memory +50MHz and change an extra $100 for it.
 

Andrew_Carr

2[H]4U
Joined
Feb 26, 2005
Messages
2,322
Didn't think to take pics. I found this from some youtube videos. I used Notura paste. From what I read only the super cheap pastes run, although maybe it is more of a crapshoot. Some people who first took apart their cards noticed the stock paste covering just the chip was already running. That putty looks really interesting however. I understand that paste isn't supposed to fill a large gap, but it's funny, the Dell/oem 3080s don't even have thermal pads, there was nothing. Gigabyte at least has some cheap looking pads...and ironically the stock Dells stay cooler in every way I can tell, compared to the Gigabytes.
I have another Gigabyte 3080 to do, so Ill take pics then.

This is what I followed mostly

There are a lot of good comments. Also for the record, I removed the old paste unlike the video =)

This video is where I heard about it first. This guy messed up applying pads and paste at first, so I went to find other peoples' experience to be sure. Iirc at the end, he ultimately recommends just paste, pads being superfluous, at least in this case.


It's been over a week and still running cool. I'll remember to check on the cards for possible paste runoff.

Not sure I will try the 3090 fe. Might try a ghetto heatsink+fan on top of the card setup.


I did this on my 3080 that was having issues. Nothing leaked out. Not sure where the risk is if you use non-conductive paste.

https://hardforum.b-cdn.net/data/attachment-files/2021/06/465435_thermal_paste_on_die_and_hsf.jpg
https://hardforum.b-cdn.net/data/at...34_thermal_paste_on_die_rtx_3080_6_8_2021.jpg
 

Roflcopter_Down

[H]ard|Gawd
Joined
Jan 7, 2008
Messages
1,535
I puttied up my main gaming card, some Gigabyte 3080. It went from ~63mh/s to 93-100mh/s! Best of all, poorly optimized games like Escape from Tarkov, no long overheat the card and cause it to throttle...no more stuttering!
For comparison the thermal paste 3080s cards only do like 87mh/s (although they are in a warm room close to other cards, while my gaming card is alone).
I'd recommend putty over paste, but paste seems way better than nothing. My pasted up cards still look fine. I applied putty to the front and back of memory. I didnt change the other thermal pads or anything else. The backside might have been a bit of a waste since the 3080s dont have backside fans like the 3090s. I used almost all the putty from one jar on this card. I probably used a little too much, but I didnt want to have to open it up again.

Didn't take pic of the backside putty. I was also out of latex gloves so it wasnt easy to make lil balls of putty.
 

Attachments

  • IMG_20210811_140609-cleaned.jpg
    IMG_20210811_140609-cleaned.jpg
    516.9 KB · Views: 0
  • IMG_20210811_145044-cleaned.jpg
    IMG_20210811_145044-cleaned.jpg
    493.1 KB · Views: 0
Last edited:

Andrew_Carr

2[H]4U
Joined
Feb 26, 2005
Messages
2,322
I puttied up my main gaming card, some Gigabyte 3080. It went from ~63mh/s to 93-100mh/s! Best of all, poorly optimized games like Escape from Tarkov, no long overheat the card and cause it to throttle...no more stuttering!
For comparison the thermal paste 3080s cards only do like 87mh/s (although they are in a warm room close to other cards, while my gaming card is alone).
I'd recommend putty over paste, but paste seems way better than nothing. My pasted up cards still look fine. I applied putty to the front and back of memory. I didnt change the other thermal pads or anything else. The backside might have been a bit of a waste since the 3080s dont have backside fans like the 3090s. I used almost all the putty from one jar on this card. I probably used a little too much, but I didnt want to have to open it up again.

Didn't take pic of the backside putty. I was also out of latex gloves so it wasnt easy to make lil balls of putty.
What is that putty? I have a 3080 I can't figure out the correct pad thickness for so maybe I'll try that. (It's a zotac OEM card that is similar to that alienware 3080).
 

Roflcopter_Down

[H]ard|Gawd
Joined
Jan 7, 2008
Messages
1,535

Nenu

[H]ardened
Joined
Apr 28, 2007
Messages
19,917
I dont disagree. But the point is thermal paste everyone already has can make these cards (at least some models) run much much better.
Perhaps this needs a prohibition for those with cards mounted vertically.
As the paste heats up it can run down the card leaving memory without cooling.
I'm not sure its a good idea long term on a flat mounted card either, unless a very sturdy paste is used.

ps I dont mind be proven wrong.
 

legcramp

[H]F Junkie
Joined
Aug 16, 2004
Messages
11,858
  1. Buy thick pads ~2.5mm-3.0mm in thickness (depending on your card)
  2. Put new pads on VRAM
  3. VRAM temp good but GPU core temp high? (pads on VRAM too thick, less contact on GPU)
  4. Take new pads put into sandwich ziplock bag
  5. Take heavy book
  6. Put bag under book, stand on book
  7. Take pads and put on VRAM again
  8. Rinse and repeat until all temps are great
  9. Enjoy great GPU and VRAM temps

You're welcome, this works fantastic for cards with weird sizes for their thermal pads like 2.8mm etc, the more you step on book, the thinner pads will be!
 

Roflcopter_Down

[H]ard|Gawd
Joined
Jan 7, 2008
Messages
1,535
Perhaps this needs a prohibition for those with cards mounted vertically.
As the paste heats up it can run down the card leaving memory without cooling.
I'm not sure its a good idea long term on a flat mounted card either, unless a very sturdy paste is used.

ps I dont mind be proven wrong.
I see what you mean, worth noting, I am trying to find videos of paste running down but it's not easy phase to search.
Has anyone experienced modern thermal paste dripping?
 

Falkentyne

[H]ard|Gawd
Joined
Jul 19, 2000
Messages
1,816
I see what you mean, worth noting, I am trying to find videos of paste running down but it's not easy phase to search.
Has anyone experienced modern thermal paste dripping?
Paste needs a gap of far less than 0.1mm to be effective, and needs two mediums that have -direct- physical contact with the actual chips. It's made for conducting heat through imperfections in chip and heat block surfaces. It's not made as a gap filler. Trying to use thermal paste on a CPU, for example, where there is a 0.1mm gap between CPU and heatsink will result in atrocious temps very quickly, with very high core temp deltas on the DTS sensors on multi-core chips, with a slow increase in those deltas and average temps over time. That is your "paste runoff"--paste dripping off the chip from constant thermal expansion and contraction, and then what's left over drying into a powder type substance from massive heat and hotspots. This has already been very well known and documented on notebookreview forums. Not going to do your searching for you, but there are tons of links about people dealing with this on the Alienware, MSI, Acer, and even Clevo sections. There's no need for "videos of paste dripping off"--temps show everything and this isn't any type of new discussion. There are people even dealing with degrading temps on heatsinks that directly contact the chips, all because of balancing problems with the heatsink (e.g. Tripod bullshit), or convex or warped plates--VERY common in the laptop world.

You can look here.

http://forum.notebookreview.com/thr...nd-apply-traditional-and-liquid-metal.806840/

Thermal paste cant function with gaps between the two surfaces. If there are gaps, you need thermal gap filler (Pads or putty).
I think you're just mixing up paste on VRM/VRAM with people using actual copper shims, and THEN using paste. There, the shims are the gap filler, and the paste creates the perfect bond between the two surfaces, as copper has a much higher w/mk than pads, so copper + Paste would do more than pads.

The huge difficulty is finding shims of the correct size--which again--is why you need shims that are --slightly-- too thin (like 0.5mm too thin) and thermal putty, or otherwise, thermal putty. The big risk is if you have a material that is too thick, this prevents a main heat plate from touching the primary GPU, which is always an issue on any integrated heatsink where the heat plate and VRAM cooling are part of the same block.
 

Shadowarez

Gawd
Joined
Jul 8, 2019
Messages
746
anyone try this with 3090 founds Cards im thinking of doing this just so i can say for 100% its been pasted nicely and has fresh pads. the gpu core runs at 45-50c under gaming load 23 - 25 idle i just dont like uncertainty. if i change this all out can i avoid the same issue steve had with that junction temp or what it was that made it worse after opening.
 

GoodBoy

2[H]4U
Joined
Nov 29, 2004
Messages
2,188
Reapply thermal paste to the chip and add paste the heatsink areas next to the chip. Most places say to buy pads, but ime you only need thermal paste, which everyone already has.
My cards mine much faster, and my gaming card runs notably cooler and smoother. It might void your warrenty, some cards don't have stickers so you might be able to get away with it.
I think it's important to share because almost everywhere says to buy thermal pads but thermal paste works just as well or better.
If I am missing something I'd like to know.
You are missing something. THIS DOES NOT APPLY TO EVERY MAKE/MODEL OF CARD.

He is using paste to mate metal to metal, and stated that he tried thermal pads and they were too thick. Obviously, that heatsink is not designed to use thermal pads!

He used paste where it is safe and fairly obvious to do so. He used paste where there either was none, or previously it had paste, but not as a replacement for thermal pads. You should really edit the thread title, it's misleading and inaccurate.

How the heatsink contacts the ram, the spacing, must be taken into account. Paste should only be used when the surfaces are flat and designed to make direct contact with each other.

I did a thermal pad replacement on my 3090FE, 1.5mm pads. You cannot use paste on a 3090FE for the memory. Of course I did use thermal paste on the GPU. My temps since then have never exceeded 90C on the ram, previously it hit 104C and I had the side of the case off.

My Max readings after hours of Serious Sam 4:
1629173830183.png


Underneath that metal coverplate, I bet there are thermal pads... lol.

Thermal pads almost HAVE to be used, as the VRAM chip heights, VRM height, and GPU height off of the circuit board all differ. Not only that, if the manufacturer has to change brands of say VRM's after a few months production, the replacement component might be a different thickness.. So over the life of the manufacture of a video card, the heights of the various sections could change more than once.
Thermal pads solve all of that by being compressable.

Another thing to note: Thermal pads have an impressive mW/k rating as compared to paste. In fact the only paste that exceeds the thermal conductivity are the liquid metal based pastes.
Typical paste: 6 mW/k
Cheap Thermal Pads: 6 to 8 mW/k
Arctic silvers: 7 to 10 mW/k across the various versions
Liquid metal types: 35 to 38 mW/k
Thermalright Odyssey Thermal Pads: 12.8 mW/k (This is what I used on my 3090 to get the results above)
FujiPoly Thermal Pads: 17 mW/k These are pretty expensive

There's no way in hell you can use a liquid metal based TIM on the VRAM of a video card, and all other pastes are a downgrade from the decent thermal pads.
 

DaeviousMax

Weaksauce
Joined
Jun 8, 2021
Messages
110
You are missing something. THIS DOES NOT APPLY TO EVERY MAKE/MODEL OF CARD.

He is using paste to mate metal to metal, and stated that he tried thermal pads and they were too thick. Obviously, that heatsink is not designed to use thermal pads!

He used paste where it is safe and fairly obvious to do so. He used paste where there either was none, or previously it had paste, but not as a replacement for thermal pads. You should really edit the thread title, it's misleading and inaccurate.

How the heatsink contacts the ram, the spacing, must be taken into account. Paste should only be used when the surfaces are flat and designed to make direct contact with each other.

I did a thermal pad replacement on my 3090FE, 1.5mm pads. You cannot use paste on a 3090FE for the memory. Of course I did use thermal paste on the GPU. My temps since then have never exceeded 90C on the ram, previously it hit 104C and I had the side of the case off.

My Max readings after hours of Serious Sam 4:
View attachment 385594

Underneath that metal coverplate, I bet there are thermal pads... lol.

Thermal pads almost HAVE to be used, as the VRAM chip heights, VRM height, and GPU height off of the circuit board all differ. Not only that, if the manufacturer has to change brands of say VRM's after a few months production, the replacement component might be a different thickness.. So over the life of the manufacture of a video card, the heights of the various sections could change more than once.
Thermal pads solve all of that by being compressable.

Another thing to note: Thermal pads have an impressive mW/k rating as compared to paste. In fact the only paste that exceeds the thermal conductivity are the liquid metal based pastes.
Typical paste: 6 mW/k
Cheap Thermal Pads: 6 to 8 mW/k
Arctic silvers: 7 to 10 mW/k across the various versions
Liquid metal types: 35 to 38 mW/k
Thermalright Odyssey Thermal Pads: 12.8 mW/k (This is what I used on my 3090 to get the results above)
FujiPoly Thermal Pads: 17 mW/k These are pretty expensive

There's no way in hell you can use a liquid metal based TIM on the VRAM of a video card, and all other pastes are a downgrade from the decent thermal pads.
Actually thermal pads have worse thermal conductivity than paste and the reason why is in the scientific value of thermal conductance itself. W/m*K or watts per meter kelvin. The m is the particular reason because M is the distance through the thermal interface material that the thermal energy must travel to go from heat source to heatsink, so the thicker the material the less effective it is at conducting heat. Ex: 10 W/mK paste spread out to 0.0001m thick (1/10th of a mm) vs a 10 W/mK pad at 0.001m (1mm) thick, the paste will have 10 times higher thermal conductance (assuming K, the difference in temperatures of the heatsink and heat source, is the same)

basically thermal pads suck at conducting heat compared to paste but as you explained above are designed for a use case that paste cannot fulfill. So it’s a trade off, either great thermal conductivity requiring tight tolerances or worse thermal conductivity allowing looser tolerances between heatsink and heat source.
 

GoodBoy

2[H]4U
Joined
Nov 29, 2004
Messages
2,188
Actually thermal pads have worse thermal conductivity than paste and the reason why is in the scientific value of thermal conductance itself. W/m*K or watts per meter kelvin. The m is the particular reason because M is the distance through the thermal interface material that the thermal energy must travel to go from heat source to heatsink, so the thicker the material the less effective it is at conducting heat. Ex: 10 W/mK paste spread out to 0.0001m thick (1/10th of a mm) vs a 10 W/mK pad at 0.001m (1mm) thick, the paste will have 10 times higher thermal conductance (assuming K, the difference in temperatures of the heatsink and heat source, is the same)

basically thermal pads suck at conducting heat compared to paste but as you explained above are designed for a use case that paste cannot fulfill. So it’s a trade off, either great thermal conductivity requiring tight tolerances or worse thermal conductivity allowing looser tolerances between heatsink and heat source.
Makes sense. But somehow it still doesn't paint the whole picture. The partially unknown number is how much the thermal pads get compressed, but I think there is more to it.

On the 3090FE, users have reported up to 25C improvement on VRAM temps, going with the Thermalright pads over stock. And there is an even better pad available.

Thermal pads basically are thermal paste... Just do a thermal pad replacement and touch one, try to remove it. It's basically thermal paste bound to a framework of some sort, strikes me as clothlike.
 

Nenu

[H]ardened
Joined
Apr 28, 2007
Messages
19,917
Thermal pads basically are thermal paste... Just do a thermal pad replacement and touch one, try to remove it. It's basically thermal paste bound to a framework of some sort, strikes me as clothlike.
Maybe some are still, I remember pads as you describe over 10 years ago.
Those I experienced recently have no cloth element, they are simply sticky and squidgy, consistent throughout.
ie in the last few months, the pads that came on my 3090 and Gelid Extreme.
I didnt replace my 3090 pads but had it apart to change the GPU paste, the Gelid was for another project.

A simple way to check is if the pad edges have any cloth thread poking out or not.
All pads are cut from a large sheet so there should be some visible somewhere.
 

DaeviousMax

Weaksauce
Joined
Jun 8, 2021
Messages
110
Makes sense. But somehow it still doesn't paint the whole picture. The partially unknown number is how much the thermal pads get compressed, but I think there is more to it.

On the 3090FE, users have reported up to 25C improvement on VRAM temps, going with the Thermalright pads over stock. And there is an even better pad available.

Thermal pads basically are thermal paste... Just do a thermal pad replacement and touch one, try to remove it. It's basically thermal paste bound to a framework of some sort, strikes me as clothlike.
You are right about the consistency of the pad and the squish effect but still my analysis is correct in that with a thermal pad there is still a thickness that is magnitudes higher than the microscopic ridges and valleys of two metal surfaces touching and the paste filling in the microscopic voids. Just know that you could never squish a pad to the microscopic level so even if the pad could be squished to maximum it still won’t be as thin as paste. A good experiment would be making 2 custom heatsinks using the same materials, method, and design except one has the same heatsink to mem chip distance as your 3090 and use the best thermal pad and the other with tight tolerances and using a low end thermal paste, the paste would win. Then repeat by using the tight tolerances heatsink and swap between the best thermal pad and low end paste, the paste would still win because you can only squish the pad so far before it becomes essentially solid and can no longer be further compressed and that is still factors thicker than paste.

I’m not knocking thermal pads by any means, they do a good job on anything that doesn’t have a microscopic gap between heatsink and heat source. I also don’t mean to discount your DIY prowess, it’s awesome you got 25C reduction on your memory chips! Just goes to show that GC manufacturers really use crap pads
 

GoodBoy

2[H]4U
Joined
Nov 29, 2004
Messages
2,188
Actually thermal pads have worse thermal conductivity than paste and the reason why is in the scientific value of thermal conductance itself. W/m*K or watts per meter kelvin. The m is the particular reason because M is the distance through the thermal interface material that the thermal energy must travel to go from heat source to heatsink, so the thicker the material the less effective it is at conducting heat. Ex: 10 W/mK paste spread out to 0.0001m thick (1/10th of a mm) vs a 10 W/mK pad at 0.001m (1mm) thick, the paste will have 10 times higher thermal conductance (assuming K, the difference in temperatures of the heatsink and heat source, is the same)

basically thermal pads suck at conducting heat compared to paste but as you explained above are designed for a use case that paste cannot fulfill. So it’s a trade off, either great thermal conductivity requiring tight tolerances or worse thermal conductivity allowing looser tolerances between heatsink and heat source.
The more I think about this, I am not sure you are interpreting what the value of mW/k means, regarding how the thickness affects the thermal conductivity and heat transfer. I am sure that the thickness plays a role, but it doesn't seem logical that something having say .2mm thickness vs a paste at .001 thickness, is somehow 200 times worse at conducting heat. It's probably a logarithmic or algebraic scale, but I don't think it's linear.

But we are just nitpicking at this point. The thermal pads are pretty decent when you upgrade them and offer significant results. Thermal paste cannot be used in their place.
 

Nenu

[H]ardened
Joined
Apr 28, 2007
Messages
19,917
The more I think about this, I am not sure you are interpreting what the value of mW/k means, regarding how the thickness affects the thermal conductivity and heat transfer. I am sure that the thickness plays a role, but it doesn't seem logical that something having say .2mm thickness vs a paste at .001 thickness, is somehow 200 times worse at conducting heat. It's probably a logarithmic or algebraic scale, but I don't think it's linear.

But we are just nitpicking at this point. The thermal pads are pretty decent when you upgrade them and offer significant results. Thermal paste cannot be used in their place.
It will be linear.
The properties of the material do not change with thickness.
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
6,652
With an EVGA 3090 XC3 with upgraded cooling using EVGA hybrid cooler and backside thermal pads replaced with added heat sinks to the backplate and a somewhat dirty old fan laying on top: 9+ hour period mining as shown in HWinfo and Awesome miner, max junction temperature was 92c with a HR around 109MH/s mining Ethereum plus all 24cores/48threads for CPU mining Monero:

DDR6xCooling3090.jpg

MiningTemps2.png
It took a combination of factors to get the temperature to get down into the low 90's while mining ETH with over 100MH HR. Cooling the backside with heatsinks and fan was probably the biggest but the Hybrid cooler was pretty significant as well. I really did not like the idea of pushing 100c plus on the DDR6x. I like to keep graphics card 3-5years.

Edit: 21:9 images are downsized automatically and makes them hard to see, so here is a cropped 16:9 image

MiningTemps3.png
 
Last edited:
Top