Unless you are Mr.Moneybags I only recommend swapping the face of the card. You can get away with using a single 2mm sheet. US link: amzn.to/3MC2pO0 Canada: amzn.to/3Xf62y2
Exactly.. Memory temps reduction by 16°C is a huge success, it doesn't matter if the performance rises or not, the important thing is the memory now isn't overheating what should potentially increase the gpu lifespan, or rather the lifespan of the memory chips. The sad thing is the gpus don't come like this from the factory, the late stage capitalist big business corporatist thinking nowadays is that it will survive long enough so the warranty expires and then you'll simply buy new gpu if it manages to die on you. Some call this planned obsolescence but they don't really plan it, they just use the same materials they used for decades now and only address the issues ie replacing the materials or doing other improvements if the product would be dying way too often under warranty what costs them money.
Exactly! Not to mention that with the lower temps, I was able to oc my 3080 and run fans slower and quieter, SOOOO, I gained huge performance advantage.
Looking at the GPU market prices, lowering temps may not give you more gaming performance, but it will make your card last longer due to the low temps, which I'd say is worth it.
Yep! I plan on using my RTX 2080 ti at least until 2030 .. seriously. and I am gonna replace the paste on the GPU by a Carbonaut pad and the thermal pads by Gelid ones. As the memory was an bigissue with the 2000 series, lowering the temperatures on the mem could have a huge impcat on longevity. Also I of course heavily undervolted the card.
Thats why i ordered new pads for my 3070 :) For the test, i repasted the chip and the temp in-game dropped with 10 degrees. After that i decided to replace the pads as well.
These mirror my results exactly. Performance improvement was within margin of error, but memory temperatures were decreased substantially, and that was with cheap "blue" amazon thermal pads. Performance gains were only realised with undervolting (using the Volt/Freq curve in Afterburner). Great video!
@@Gojee_ Not that I recall, not substantially anyway. I was also playing around with custom fan curves at the time, so I can't comment on how fan speeds were affected, or by how much.
the coagulated feel of the pad is called "wet-out" pad. They are extremely conforming to the shape of the location you apply them to. This is because any gaps/scratches/deformities can be filled in and eliminate the air pockets, ensuring good contact. Its usually held in shape by the plastic sheets/casing it comes in. You need to let them cure/set before the catalogue properties/performance can be expected.
Just did the same thing on my 3080ti FE. Oct 2021 date on the cooler, reddit suggested changes on the later production 3080(ti) FE coolers should use 1.5mm instead of 2mm. Used 1.5mm Gelid Extreme on the chip size, 2mm Extreme on the back place. Temp results are fairly aligned with the your test. Thank you for the info!
Hey, it might be late for you to remember, but what thickness did you use for VRAM and MOSFET pads? I have a 3080 TI that has a hot spot reaching 100 degrees when gaming, so I am planning to replace pads/thermal paste to start with and see if anything changes...doing some research first regarding what pads/what thickness to use for different locations on the GPU though.
It was pretty useful! Especially respect for making separate test with temperatures with new paste, but no new pads, since it could change the results. Thanks for the video!
@@flawns yes but still it helps with the peace of mind that your card is not running at near limits. also, it helps keep room temps down especially if you're not using ac.
@@flawns By a f**king shitload. There are plenty of studies which have tested the difference of running at thermal throttle temps (100% of the time) vs capped at 90% of thermal throttle temps and it almost doubles the lifespan....considering his memory was being throttled (100 deg) ....yeah....huge difference.
@@Lancia444 For real, you're kinda aggressive but this factor seems really under emphasized. Heating and cooling is expanding and contracting the material, which wears it out. All materials have durability limits and can only do this so much. The higher temps push all metal and silicone further to it's limit, and like most thing's in nature it isn't linear, but moreso exponential, specifically to closer to the Fibonacci pattern than any other pattern. Basically the lifespan can be thought of as a result of build quality taken to the square root of the temperature it's used at. That lifespan dropoff gets a way faster increase in every incremental step of it's reduction for each consecutive degree a material is taken toward the final max temperature that it will simply break at immediately.
@@mangosteen. "it helps keep room temps down especially if you're not using ac" That's not true. The card will output the same amount of heat no matter what cooler/pads you put on it. It's just that the heat from the card will be transferred into the air more efficiently. Unless they've invented a new form of GPU cooling solution that sucks heat and then teleports it into a black hole at the other end of the universe.
These results are so misleading and wrong. The temperature reduction, yes, thats expected and obviously intended. But the reason you do this is 1) to reduce throttling and overall heat if its an old card, and 2) to increase overclocking headroom. Reducing the core temperatures of the vram/vrm mosfets means you have even further safe overclocking headroom while still maintaining those good temperatures due to no more artifacting introduced by intense heat. I ran a .993v undervolt @ 1995mhz with a +1000mhz memory overclock on my 2070super. Max 62~c in time spy. Prior to the undervolt method and just attempting +100/+1000 etc directly in Afterburner, Fire Strike would artifact noticably and it'd get close to 85c. Aka it performed worse at stock voltage curve and ran hotter/downclocked a lot. Anyway, back to the point - after replacing both paste and pads with gelid extremes, i can run 1.043v @ 2055mhz with +1200(!) Memory clock and be max 45~c by the end of Time Spy. I have zero idea why you would expect changing the thermal pads would give you increased fps when leaving it stock. New paste would because of gpu boost 3.0, but the vrm and vram obviously only affects your memory clock which is static unless you overclock.
This is the one I tested too, I agree with you, if you don't overclock, the only real effects are lower temepratures and quieter fans. Overclocking is never a good idea for a recent board or one under warranty.
Auto like and subscribed because you told us the results in the beginning. Skimmed through the video anyway to see your methodology but the information up front is always appreciated.
There's a LTT video where they try out a thermal pad replacement in gel form. The pad replacement was largely irrelevant, but the repaste turned out very interesting. There wasn't a direct FPS increase, but hte min 1% and min 5% fps increased by about 20%, which is huge. They didn't test this very thoroughly or even note it in the video. I'll probably try this out myself, but it may or may not be a good video topic for you as well.
To get a better idea of performance improvement in FPS or synthetic benchmarks, you would want to overclock the GPU to a stable setting _before and after_ swapping the thermal material
Yep exactly, you won't see big difference in performance, as long as the GPU was functioning properly and normally, he is only lowering the temps, but what that does is to give you more space in terms of overclocking the GPU and that is when you will get a boost in performance
I just replaced the thermal pads on my Asus TUF 3080 over the weekend. I was getting a GPU temp of 64 which isn't bad but then a hotspot temp of 90C and Memory Junction temp of 110C. After replacing the pads with Gelid pads, and knocking two resistors off the board, and dying inside until Monday when I got the GPU fixed, I was back up and running. My GPU temp dropped about 8 degrees using Arctic MX-4, Hotspot dropped to 72C and most impressive of all is Memory Junction temp which dropped about 50 degrees to 65-70C. This was with my GPU overclocked too. I do have an undervolt profile saved but haven't checked what the temps are for that but I'm sure temps would probably be a tad bit above idle temp.
I thought 1.5 mm was the ideal pad size for these things... that's what I used on my 3090. Same deal, temps dropped significantly.. I don't game, but I do feel safer when my card is rendering in octane for several hours.
Better pads = Lower temps on card, less heat in the PC tower in general, less stress on the card which means better longevity in general. Couple this with a slight undervolt and you're golden. Also don't forget that some cards come with really crappy thermal pads. Subbed by the way 😀
You're also makign it quieter. I got a 3090 yesterday and noticed it never ramps up for GPU temperature: the huge cooler deals with that pretty well, but sometimes it seems to go into turbojet mode, and if I look in HWINFO, it's the memory junction at 104-106C. Apparently damage temp is about 120C, but if you're spiking to 110, that's uncomfortably close. Going to repad this week, because I want my new card to last for at least 4-5 years.
Got a 3070 2nd hand and it was having really bad temp issues. Had to underclock my GPU otherwise it would completely black out after 10 minutes. Replaced pads and paste following this advice and also the undervolting video and it works a treat now getting temps ~60°C when previously it was capping out at 83°C.
Fan speed references the core temp, I think it might ramp up if VRAM is thermal throttling, so the difference would be pretty minor, not nothing, but small
I dont think many people looking for thermal pad/paste replacement to improve performnce from stock (unless defect), rathet for longevity and thermal improvement Thank you for the video
A question I have is, while it may lower the mem temps and higher benchmark score, how about the fanspeed? Did they become more quiet? That can also be an interesting factor to look into for those who like a silent operation under heavy load =)
I'd recommend slapping a fin heatsink on the back of GPU. I've done that to my 1060, and it's actually drawing away a lot of heat. The backside heatsink I installed is always hot to the touch, I'd guess around 40C. With a more power hungry card like 3000 series, I'd imagine the benefits would be even greater.
@@sscultima Nah, the 5600x has a maximum temperature of 95C before it starts to throttle. Mine was dropping almost 30 FPS when it was 110. That was my first hint that something is wrong since I don't check my CPUID HWMonitor every day.
Those 30serie cards run super hot. My 4080 super runs at idle core/memory 32~37, underload core/memory max 68~70. I was looking to change the thermal pads to improve temps but I'm super surprise how great they are compare to older gens.
What about your fanprofile? If its stock it will just run the fans lower speed when temps drop so you have to keep the same fanspeed all the time to get accurate results.
The fan curve certainly skew the results in a way that bring the numbers closer together. But most people are going to have a fan profile that isn't flat so I like to test in a way that will show what actual day to day results look like. I published a video about undervolting. th-cam.com/video/9JsqKREvukQ/w-d-xo.html You can see how the fan curve keeps the temps closer than they would be with a flat profile. The power differences suggest the gaps in heat should be bigger.
I replaced the pads and the paste on my three-year-old 3080 and now have 8 to 10°C less GPU temp. The card is much quieter and has 400 points more in TimeSpy with the same settings as before. For me, the 30€ investment was worth it.
I just ordered a 3080 that the reseller failed because it crashes after an hour of furmark, it's a Dell OEM card that is known for high memory temps (thanks Dell design team) if I'm lucky pads and paste make it a very cheap 3080 (or what a last gen good card should actually cost rather), if I'm unlucky it's an expensive brick. Either way it's getting the thermal mod it deserves to find out.
I think there is an advantage for the gamer from doing this. In that the card will likely run for longer before eventual failure. While I understand that the specs for the vRAM may be max temp 110'C, it's better to run cooler. It is after all, heat that destroys silicon. I think the 3'C drop on your GPU core is worth it too, because the fans will run fractionally quieter. If someone is OK with forgoing the warranty then it's a good idea all around in my opinion. Additionally though, some manufacturers do allow customers to change the thermal paste, without voiding warranty. Honestly found your vid vaulable. At the very least it means that you could run your gaming rig in a warmer room. Which also means you can go longer before reaching for the air-conditioner. Hence a bonus for carbon footprint. Lastly though a cooler card means less heat emitted toward the CPU cooler. Like you mentioned the temp of the backplate on your card. When I worked on my 1080 Ti, the backplate touch-temperature dropped from scorching to much more bearable. That will keep the whole card that bit cooler, and prolong the life of all components. It will also reduce the heating-cooling cycle extremes, which will help with component lifetimes. EDIT: Possibly reducing power consumption fractionally too.
A cooler internal component does not mean less heat going inside the PC case. It may actually mean more. Let’s say that your CPU thermal paste is REALLY bad, worse than no paste, and it only takes like 40W until it throttles. Well, it’ll only throw 40W into the heatsink. Now you repasted it, and it can run a full 120W without throttling. Guess what? Now your heatsink will be toasty. Heat = energy, but not necessarily temperature. That’s why a car can have a much more powerful engine with a much bigger radiator, still maintain itself at 90°C, and function as a space heater much better than the lower power car
@@satsumagt5284 You're working from theory of an unlikely event. If you have no thermal paste the GPU will heat up and throw heat in to the case. From experience I saw the following. One of my GPUs used to run around 70-73'C. One time when I was repasting, and I noticed and issue with two inductors next to each other, One inductor did not have a thermal pad on. Meaning no pad between inductor and GPU heatsink. Once I put a pad on, and replasted with the same paste I always used preaviously with it. I noticed a considerable drop in temps on the card. Card ran between 60'-70'C. The result, and my argument was this. I went from needing to run my rear case fan on full, to running on medium. (I could even add a small core clock OC without any more power/voltage needed if I wanted.) Considering that the GPU fans were having to deal with more heat, that was interesting. Clearly what matters is the temperature of the components over all else. The coolers can do their job even when given more load. What they can't do is compensate for a component like an inductor gettting hot. Then that inductor spreading heat through the board to the GPU chip. I.e. using better pads and paste will mean slower GPU fans and case fans, over poorer pads and paste. Poorer pads and paste mean your GPU will get hot, and the GPU fans will run very fast. Then you'll over-compensate with case fans to try cool it.
@@greenbow7888 The Graphics card’s fans used to go crazy because they usually use the GPU temperature to determine their speed. They can have more inputs too, of course. By repasting the GPU, you made the heat transfer from GPU to heatsink more efficient. With all that said, your heatsink is surely getting hotter now that it receives more energy from the GPU and the inductor. GPU at 70°C almost never equals heatsink at 70°C, especially if the conductivity is poor. Proof of that is my laptop: Before repasting it, the case itself would barely warm up, CPU would hit 88ºC easily and the fans were annoyingly loud. That’s because it was only drawing like 23W and reaching the throttle temperature. Nowadays, the CPU can even draw 30+W with the GPU under load without reaching 88°C, the case and heatsink gets hot AF, fans are notably quieter and the performance is on another level, since it can actually keep itself at 4GHz constantly (On a cold day I don’t even need Beast Mode to avoid some slight throttling while gaming).
100 degrees is high, but it probably isn't shortening the realistic lifespan of the card, even under heavy load. It isn't uncommon for industrial IC's to reach 105 degrees and run for ten years continuously at that temperature. But it's probably better to be safe than sorry. I would definitely prefer to see lower temperatures on these kinds of cards. If you are really concerned about longevity, making sure the card isn't overclocked, and possibly underclocking and undevolting the card, is the way to go.
My poor 3080 was hitting about 101 to 105c on the memory chips but the gelid extreme pads helped. I also put some on the backplate and now you can fully load the gpu and you get maybe 80c
These had better bring the NVS 2100M away from that 100c range. It came stock like this, I repasted both the CPU and GPU and while the CPU did improve temps, GPU got worse. I suspect that since the GPU came with pads, I should just get NEW pads. Ones that really perform instead of the old blue malleable stuff. I'm gonna have to guess the thickness so I'll get both the 1mm and 2mm before I take apart my Tecra A11 S3540 again.
hello. I have bought 1.5 and 2mm. From what I have read, the 1.5 would be for the memories and the 2mm for the rear part. Is that correct? thanks for the video and greetings
Lower temps would help stability, wouldn’t it? About to try padding up an old GTX 970 that crashes under heavy gaming loads but seems fine on regular duty/temps …
considering how much temperature drop you got, I'd say it's totally worth it. A GPU is one big expensive piece of electronic. I'd take any opportunity to increase the life span of that thing.
So, huge thanks for proving that when you have a hot spot of 110 degrees Celsius on your 7900XTX then thermal pads MUST be replace. Huge thanks for your test.
Better temps only leads to better gaming only if you’re currently dealing with thermal throttling. Any other case, it is just healthy for the gpu to operate way below the throttle limit
Well for mine 3080 ti FE 2mm pads are too thick i just switched on gpu and fans went crazy so I opened and thermal paste didn't even touch cold plate...1.5mm maybe even not sure feels like 1mm could be best choice..Damnit i need order new pads 😂😮💨
2 years ago i bought a second hand aorus xtreme gtx 1080, and i changed only the thermal paste, but now since the gpu is 6+ years old i am also thinking of replacing the thermal pads as well when i open it in a few days, those 6+ year old pads are probably worth replacing with fresh new ones, not for gaming, but to extend the life of the gpu as much as possible
I'm curious what the ambient temperature of the room was during the tests... Perhaps those in hot climates would see even more benefit from changing thermal paste/pads
i have 3080ti but TUF from asus, and first thing i do is to undervolt the gpu, and my temps drop to 60st in metro exodus in 4k with no hit in performance :D, now im waiting for MX-6 and gelid ultimates and will drop even more temps from those manufacture stock paste and pads
I think its maybe the Founders edition Design cant be bring the temperature out so good,if you do the test on a custom design. You will be get a better result. I do that with my 2080 rog,new paste and new pads, and i dropped from 83c to 70c maxed out 1 hour pubg all max settings. idle temps was over 32c too and i dropped to 27c
Replacing thermal paste is an amazing solution to older cards (2070 Super in my case) having fans ramp up and down constantly. Fixed my issues and dropped the temps by 10+ degrees. Maybe performance has also gone back up to closer to when it was knew, but that's not really the reason to mess with the paste, its the irritating fan noise.
@@LEK Exactly that yes. No fan controller/setting worked. But changing the paste worked completely. I did not change the pads, I was careful not to damage them. Maybe I should have done them as well, but the results were still very good.
@@crispycarrots Yeah had the same issue on my 2070 Super. They charge 700 bucks for the card but cant put the effort in to properly apply quality thermal paste and pads which arent that expensive at all.
@@LEK Dude, that is happening to my 2070 super as well! and after tried all imaginable under the sun fan control software, it did not matter :D. I guess it is time for paste/pads change.
thanks for this test this test amazing work i change thermal paste to memory i have 60 c to memory after change that after 2 years on rtx 3080 ti trinity
people dont upgrade the thermal paste and pads for performance, it's for extending the life of the GPUS by years. Heat is a major culpert in product lifetime and dropping memory temperatures by nearly 14-16C is huge. Also, if you want performance, with so much more headroom with lower temperature this opens up the door for overclocking the GPU which is another reason as to why people do these mods.
Asus in their infinite wisdom only covered half the memory with pads on my old 2060. They also seems to be low thermal conductivity. A $15 (Canadian dollar) multipack of basic pads allowed me to redo the VRM, add pads to the backplate that it didn't previously have, and fix the memory all at once. I still have more than half of the pad material left. It's a bit sad that budget pads could be improve what was stock on a "premium" card. The strix models are a scam.
I would believe these tests if every test were on par with the stock pads that had been in use for 6 months compare to the other tests who were fresh install.
I had to use new thermal pads because the old ones were torn down. It was not my GPU. The original owner was smoking. The "smoke" "glued" everthing together. There was no option than replace old ones.
there new pads the 15w/m-k are good. i use them for my sapphire toxic 6900 xt (pushing that air cooled timespy record). the biggest difference is within the hotspot. where i was maxing 105c on runs. but now with same settings it barely cracks over 100. thanks for the vid tho
hey i want ask u something about thermal pads gelid I bought a thermal pad gelid-ultimate and that arrived gelid extreme did u think there is a difference between the two and that was cheating right? so I called them to return that right or i can keep it?
Changed thermal pads on my gpu yesterday.. Before i had temps around 62-65 and hotspots over 100° when doing heaven 1440p ultra which lead to the gpu crashing.. Now i get hotspots around 75 after 10 min with 50-70% fans almost the same "gpu temp". So if you run into issues with high hotspots etc, either rma it or open it up and swap pads / paste took me around 1h
Unless you are Mr.Moneybags I only recommend swapping the face of the card. You can get away with using a single 2mm sheet. US link: amzn.to/3MC2pO0 Canada: amzn.to/3Xf62y2
@Tech Illiterate Is this 3080ti FE new? like bought within the last couple months? People are saying the spec changed with the newer cards
@@spaghettir101 bump ^
@@spaghettir101 his is old board
@@spaghettir101 In 3080ti Fe have 5 month old more or less, whats thermal pads I need? Is 1.5mm one side and 2mm other side?
@@mdmt111 probs 1.5mm on the vram and 2mm on the vrms. Leave the back, its not needed
It's not just about gaining performance, it's also prolonging the life of the card.
Exactly.. Memory temps reduction by 16°C is a huge success, it doesn't matter if the performance rises or not, the important thing is the memory now isn't overheating what should potentially increase the gpu lifespan, or rather the lifespan of the memory chips.
The sad thing is the gpus don't come like this from the factory, the late stage capitalist big business corporatist thinking nowadays is that it will survive long enough so the warranty expires and then you'll simply buy new gpu if it manages to die on you. Some call this planned obsolescence but they don't really plan it, they just use the same materials they used for decades now and only address the issues ie replacing the materials or doing other improvements if the product would be dying way too often under warranty what costs them money.
Exactly! Not to mention that with the lower temps, I was able to oc my 3080 and run fans slower and quieter, SOOOO, I gained huge performance advantage.
Not just that but fan noise is bad when the temps are very high.
and he didnt or couldnt monitor vrm temps, im pretty sure a huge differnce in temps on them too
Might not affect performance much BUT it could help reliability and longevity.
Looking at the GPU market prices, lowering temps may not give you more gaming performance, but it will make your card last longer due to the low temps, which I'd say is worth it.
no one knows by how much
. were making our best guess that could be the case
@@flawns lower is always better.
Don't forget the noise of the fighter jet that doesn't come out off my case ;)
It will improve latency, low1% i notice
Yep! I plan on using my RTX 2080 ti at least until 2030 .. seriously. and I am gonna replace the paste on the GPU by a Carbonaut pad and the thermal pads by Gelid ones. As the memory was an bigissue with the 2000 series, lowering the temperatures on the mem could have a huge impcat on longevity. Also I of course heavily undervolted the card.
I think high temps can degrade more quickly components. That -20ºC maybe help in therms of longevity. Great vid.
The rapid rise and fall of temperatures affect the life span of the card. High but not critical constant temperatures are ok.
Am I blind, where do you see -20?
@@serban2139 read the graph, the bonus test section. u can see it on the graph (mem avg/max) do you even watch the whole video?
lower temps = lower fan rpm = quieter pc. Well worth it imo.
Thats why i ordered new pads for my 3070 :) For the test, i repasted the chip and the temp in-game dropped with 10 degrees. After that i decided to replace the pads as well.
How have I only just discovered your channel. Exceptionally well explained and produced.
Keep making the videos, do not be discouraged by the less views. U will catch up soon. Consistency is the key.
I am elated that anyone watches at all :)
Thanks!
These mirror my results exactly. Performance improvement was within margin of error, but memory temperatures were decreased substantially, and that was with cheap "blue" amazon thermal pads. Performance gains were only realised with undervolting (using the Volt/Freq curve in Afterburner). Great video!
did the fan sound also decrease?
@@Gojee_ Not that I recall, not substantially anyway. I was also playing around with custom fan curves at the time, so I can't comment on how fan speeds were affected, or by how much.
For what it’s worth gelid is very close pronunciation to the Arabic word “جليد” “galeed” which means frozen/ice
the coagulated feel of the pad is called "wet-out" pad. They are extremely conforming to the shape of the location you apply them to. This is because any gaps/scratches/deformities can be filled in and eliminate the air pockets, ensuring good contact. Its usually held in shape by the plastic sheets/casing it comes in. You need to let them cure/set before the catalogue properties/performance can be expected.
except if the pads are crap then you won't get very significant temp changes when they do cure/set.
Just did the same thing on my 3080ti FE. Oct 2021 date on the cooler, reddit suggested changes on the later production 3080(ti) FE coolers should use 1.5mm instead of 2mm. Used 1.5mm Gelid Extreme on the chip size, 2mm Extreme on the back place. Temp results are fairly aligned with the your test. Thank you for the info!
Hey, it might be late for you to remember, but what thickness did you use for VRAM and MOSFET pads? I have a 3080 TI that has a hot spot reaching 100 degrees when gaming, so I am planning to replace pads/thermal paste to start with and see if anything changes...doing some research first regarding what pads/what thickness to use for different locations on the GPU though.
It was pretty useful! Especially respect for making separate test with temperatures with new paste, but no new pads, since it could change the results.
Thanks for the video!
I really appreciate you doing this testing and collecting data. I have been thinking about redoing the thermal pads on my graphics card. Thumbs up!
A graphics card at better temps is worth it, as it probably might increase the longevity of the card…
no one knows by how much
@@flawns yes but still it helps with the peace of mind that your card is not running at near limits. also, it helps keep room temps down especially if you're not using ac.
@@flawns By a f**king shitload. There are plenty of studies which have tested the difference of running at thermal throttle temps (100% of the time) vs capped at 90% of thermal throttle temps and it almost doubles the lifespan....considering his memory was being throttled (100 deg) ....yeah....huge difference.
@@Lancia444 For real, you're kinda aggressive but this factor seems really under emphasized. Heating and cooling is expanding and contracting the material, which wears it out. All materials have durability limits and can only do this so much. The higher temps push all metal and silicone further to it's limit, and like most thing's in nature it isn't linear, but moreso exponential, specifically to closer to the Fibonacci pattern than any other pattern.
Basically the lifespan can be thought of as a result of build quality taken to the square root of the temperature it's used at. That lifespan dropoff gets a way faster increase in every incremental step of it's reduction for each consecutive degree a material is taken toward the final max temperature that it will simply break at immediately.
@@mangosteen. "it helps keep room temps down especially if you're not using ac" That's not true. The card will output the same amount of heat no matter what cooler/pads you put on it. It's just that the heat from the card will be transferred into the air more efficiently. Unless they've invented a new form of GPU cooling solution that sucks heat and then teleports it into a black hole at the other end of the universe.
These results are so misleading and wrong.
The temperature reduction, yes, thats expected and obviously intended. But the reason you do this is 1) to reduce throttling and overall heat if its an old card, and 2) to increase overclocking headroom.
Reducing the core temperatures of the vram/vrm mosfets means you have even further safe overclocking headroom while still maintaining those good temperatures due to no more artifacting introduced by intense heat.
I ran a .993v undervolt @ 1995mhz with a +1000mhz memory overclock on my 2070super. Max 62~c in time spy.
Prior to the undervolt method and just attempting +100/+1000 etc directly in Afterburner, Fire Strike would artifact noticably and it'd get close to 85c. Aka it performed worse at stock voltage curve and ran hotter/downclocked a lot.
Anyway, back to the point - after replacing both paste and pads with gelid extremes, i can run 1.043v @ 2055mhz with +1200(!) Memory clock and be max 45~c by the end of Time Spy.
I have zero idea why you would expect changing the thermal pads would give you increased fps when leaving it stock. New paste would because of gpu boost 3.0, but the vrm and vram obviously only affects your memory clock which is static unless you overclock.
This is the one I tested too, I agree with you, if you don't overclock, the only real effects are lower temepratures and quieter fans. Overclocking is never a good idea for a recent board or one under warranty.
Auto like and subscribed because you told us the results in the beginning. Skimmed through the video anyway to see your methodology but the information up front is always appreciated.
Well explained and presented. Makes sense to extend the life of any GPU whatever way you look at it.
There's a LTT video where they try out a thermal pad replacement in gel form. The pad replacement was largely irrelevant, but the repaste turned out very interesting. There wasn't a direct FPS increase, but hte min 1% and min 5% fps increased by about 20%, which is huge.
They didn't test this very thoroughly or even note it in the video. I'll probably try this out myself, but it may or may not be a good video topic for you as well.
Literal answer in the first 5 seconds, no "watch to the end to find out"? Watching the whole video and subbing, fuck yeah.
"this video is brought to you by my pocketbook" got me faster than ever before. Subbed.
To get a better idea of performance improvement in FPS or synthetic benchmarks, you would want to overclock the GPU to a stable setting _before and after_ swapping the thermal material
I totally agree with this. The real test is how the GPU will perform at heavy load.
Yep exactly, you won't see big difference in performance, as long as the GPU was functioning properly and normally, he is only lowering the temps, but what that does is to give you more space in terms of overclocking the GPU and that is when you will get a boost in performance
Thank you for answering the question right away, love the no-bullshit presentation. Now I had to watch from beginning to end, of course!
With lower temps you can overclock the memory more. Much more.
I just replaced the thermal pads on my Asus TUF 3080 over the weekend. I was getting a GPU temp of 64 which isn't bad but then a hotspot temp of 90C and Memory Junction temp of 110C. After replacing the pads with Gelid pads, and knocking two resistors off the board, and dying inside until Monday when I got the GPU fixed, I was back up and running. My GPU temp dropped about 8 degrees using Arctic MX-4, Hotspot dropped to 72C and most impressive of all is Memory Junction temp which dropped about 50 degrees to 65-70C. This was with my GPU overclocked too. I do have an undervolt profile saved but haven't checked what the temps are for that but I'm sure temps would probably be a tad bit above idle temp.
I thought 1.5 mm was the ideal pad size for these things... that's what I used on my 3090. Same deal, temps dropped significantly.. I don't game, but I do feel safer when my card is rendering in octane for several hours.
Every cooler has different dimensions, 1.5mm may be the correct size for this card, but it’s not universal
Better pads = Lower temps on card, less heat in the PC tower in general, less stress on the card which means better longevity in general. Couple this with a slight undervolt and you're golden. Also don't forget that some cards come with really crappy thermal pads. Subbed by the way 😀
You're also makign it quieter. I got a 3090 yesterday and noticed it never ramps up for GPU temperature: the huge cooler deals with that pretty well, but sometimes it seems to go into turbojet mode, and if I look in HWINFO, it's the memory junction at 104-106C. Apparently damage temp is about 120C, but if you're spiking to 110, that's uncomfortably close. Going to repad this week, because I want my new card to last for at least 4-5 years.
Got a 3070 2nd hand and it was having really bad temp issues. Had to underclock my GPU otherwise it would completely black out after 10 minutes. Replaced pads and paste following this advice and also the undervolting video and it works a treat now getting temps ~60°C when previously it was capping out at 83°C.
what paste and pad did u used ?
Yes, they are. Went from thermal throttling at 95mh/s on a 3080 TUF to 95F at 101mh/s after Gelid pads all around the board.
Did you use a similar method to this video for the 3080 TUF? Same pad thickness etc?
May not help for overall FPS but will surely help for long term LIFE of the card..
Fan speed references the core temp, I think it might ramp up if VRAM is thermal throttling, so the difference would be pretty minor, not nothing, but small
This is what I thought. Today, after watching you, I am convinced that there is no need to change. Thank you
I dont think many people looking for thermal pad/paste replacement to improve performnce from stock (unless defect), rathet for longevity and thermal improvement
Thank you for the video
Thank you for giving the answer at the very beginning!
your beard grew through the testing even if not intentional, that is awesome, hahaha
Swapped the thermal pads on my new MSI 3080 with 2mm gelid pads. Mem temps went from 90 to 80 under identical conditions.
Beautiful bonus creation…‘it’s aliveeeee’ 😂😂🎉
i wouldn't call the decrease in memory temps mild. thats a massive improvement.
Watched the first 3 seconds, already know the answer. Best YT video ever
Using the correct thickness of thermal pads is of the utmost importance.
A question I have is, while it may lower the mem temps and higher benchmark score, how about the fanspeed? Did they become more quiet? That can also be an interesting factor to look into for those who like a silent operation under heavy load =)
I'd recommend slapping a fin heatsink on the back of GPU. I've done that to my 1060, and it's actually drawing away a lot of heat. The backside heatsink I installed is always hot to the touch, I'd guess around 40C. With a more power hungry card like 3000 series, I'd imagine the benefits would be even greater.
Hi could you please tell me exactly how to do that to which area exactly I should put the fin heatsink
@@KimRTo the area that generates the most heat
Jesus Christ. 110 Celsius? My Ryzen 5 5600x was that hot when my AIO water cooler pump broke and wasn't cooling it at all. lol
isnt the max temp on that chip 90c? how was it still running, just super throttled if at all.
@@sscultima Nah, the 5600x has a maximum temperature of 95C before it starts to throttle. Mine was dropping almost 30 FPS when it was 110. That was my first hint that something is wrong since I don't check my CPUID HWMonitor every day.
Those 30serie cards run super hot. My 4080 super runs at idle core/memory 32~37, underload core/memory max 68~70. I was looking to change the thermal pads to improve temps but I'm super surprise how great they are compare to older gens.
great video, reminded me of the 680 gtx times I've gone full macgyver on that one. before modding it i was able to make some very nice Steaks.
What about your fanprofile? If its stock it will just run the fans lower speed when temps drop so you have to keep the same fanspeed all the time to get accurate results.
The fan curve certainly skew the results in a way that bring the numbers closer together. But most people are going to have a fan profile that isn't flat so I like to test in a way that will show what actual day to day results look like.
I published a video about undervolting. th-cam.com/video/9JsqKREvukQ/w-d-xo.html You can see how the fan curve keeps the temps closer than they would be with a flat profile. The power differences suggest the gaps in heat should be bigger.
I replaced the pads and the paste on my three-year-old 3080 and now have 8 to 10°C less GPU temp. The card is much quieter and has 400 points more in TimeSpy with the same settings as before. For me, the 30€ investment was worth it.
I just ordered a 3080 that the reseller failed because it crashes after an hour of furmark, it's a Dell OEM card that is known for high memory temps (thanks Dell design team) if I'm lucky pads and paste make it a very cheap 3080 (or what a last gen good card should actually cost rather), if I'm unlucky it's an expensive brick. Either way it's getting the thermal mod it deserves to find out.
I know this post is 8 months old, but I'm curious to know if you had success or not?
@@handenbramilton it's my daily driver, works fine. It wasn't memory temps as it turned out though, the core thermal paste application was bad.
I think there is an advantage for the gamer from doing this. In that the card will likely run for longer before eventual failure. While I understand that the specs for the vRAM may be max temp 110'C, it's better to run cooler. It is after all, heat that destroys silicon.
I think the 3'C drop on your GPU core is worth it too, because the fans will run fractionally quieter.
If someone is OK with forgoing the warranty then it's a good idea all around in my opinion. Additionally though, some manufacturers do allow customers to change the thermal paste, without voiding warranty.
Honestly found your vid vaulable. At the very least it means that you could run your gaming rig in a warmer room. Which also means you can go longer before reaching for the air-conditioner. Hence a bonus for carbon footprint.
Lastly though a cooler card means less heat emitted toward the CPU cooler. Like you mentioned the temp of the backplate on your card. When I worked on my 1080 Ti, the backplate touch-temperature dropped from scorching to much more bearable. That will keep the whole card that bit cooler, and prolong the life of all components. It will also reduce the heating-cooling cycle extremes, which will help with component lifetimes.
EDIT: Possibly reducing power consumption fractionally too.
A cooler internal component does not mean less heat going inside the PC case. It may actually mean more.
Let’s say that your CPU thermal paste is REALLY bad, worse than no paste, and it only takes like 40W until it throttles. Well, it’ll only throw 40W into the heatsink.
Now you repasted it, and it can run a full 120W without throttling. Guess what? Now your heatsink will be toasty. Heat = energy, but not necessarily temperature. That’s why a car can have a much more powerful engine with a much bigger radiator, still maintain itself at 90°C, and function as a space heater much better than the lower power car
@@satsumagt5284 You're working from theory of an unlikely event. If you have no thermal paste the GPU will heat up and throw heat in to the case.
From experience I saw the following.
One of my GPUs used to run around 70-73'C. One time when I was repasting, and I noticed and issue with two inductors next to each other, One inductor did not have a thermal pad on. Meaning no pad between inductor and GPU heatsink. Once I put a pad on, and replasted with the same paste I always used preaviously with it. I noticed a considerable drop in temps on the card. Card ran between 60'-70'C.
The result, and my argument was this. I went from needing to run my rear case fan on full, to running on medium. (I could even add a small core clock OC without any more power/voltage needed if I wanted.)
Considering that the GPU fans were having to deal with more heat, that was interesting. Clearly what matters is the temperature of the components over all else. The coolers can do their job even when given more load. What they can't do is compensate for a component like an inductor gettting hot. Then that inductor spreading heat through the board to the GPU chip.
I.e. using better pads and paste will mean slower GPU fans and case fans, over poorer pads and paste. Poorer pads and paste mean your GPU will get hot, and the GPU fans will run very fast. Then you'll over-compensate with case fans to try cool it.
@@greenbow7888 The Graphics card’s fans used to go crazy because they usually use the GPU temperature to determine their speed. They can have more inputs too, of course. By repasting the GPU, you made the heat transfer from GPU to heatsink more efficient.
With all that said, your heatsink is surely getting hotter now that it receives more energy from the GPU and the inductor. GPU at 70°C almost never equals heatsink at 70°C, especially if the conductivity is poor.
Proof of that is my laptop: Before repasting it, the case itself would barely warm up, CPU would hit 88ºC easily and the fans were annoyingly loud. That’s because it was only drawing like 23W and reaching the throttle temperature. Nowadays, the CPU can even draw 30+W with the GPU under load without reaching 88°C, the case and heatsink gets hot AF, fans are notably quieter and the performance is on another level, since it can actually keep itself at 4GHz constantly (On a cold day I don’t even need Beast Mode to avoid some slight throttling while gaming).
@@satsumagt5284 Am not reading your post. You are telling me I am wrong. I am right. You are wasting my time.
@@greenbow7888 Yesn’t. I only disagree with your last paragraph of the original comment. Everything else, agree
Would’ve been nice to see a VRAM overclock to see if you could gain more significant performance with the headroom.
100 degrees is high, but it probably isn't shortening the realistic lifespan of the card, even under heavy load. It isn't uncommon for industrial IC's to reach 105 degrees and run for ten years continuously at that temperature. But it's probably better to be safe than sorry. I would definitely prefer to see lower temperatures on these kinds of cards.
If you are really concerned about longevity, making sure the card isn't overclocked, and possibly underclocking and undevolting the card, is the way to go.
Yeah I think you're right 100p.
What about the fan speed? With such a huge drop on memory temps of around 15 degrees vs. stock the fans should operate at fewer RPM.
My poor 3080 was hitting about 101 to 105c on the memory chips but the gelid extreme pads helped.
I also put some on the backplate and now you can fully load the gpu and you get maybe 80c
These had better bring the NVS 2100M away from that 100c range. It came stock like this, I repasted both the CPU and GPU and while the CPU did improve temps, GPU got worse. I suspect that since the GPU came with pads, I should just get NEW pads. Ones that really perform instead of the old blue malleable stuff. I'm gonna have to guess the thickness so I'll get both the 1mm and 2mm before I take apart my Tecra A11 S3540 again.
hello. I have bought 1.5 and 2mm. From what I have read, the 1.5 would be for the memories and the 2mm for the rear part. Is that correct? thanks for the video and greetings
There's 1 big reason to upgrade cooling: cooler = quieter!
Lower temps would help stability, wouldn’t it? About to try padding up an old GTX 970 that crashes under heavy gaming loads but seems fine on regular duty/temps …
considering how much temperature drop you got, I'd say it's totally worth it. A GPU is one big expensive piece of electronic. I'd take any opportunity to increase the life span of that thing.
welcome back handsome!
I will be going a step further by using thermal putty and copper heatsinks on memory chips and see what the results are.
So, huge thanks for proving that when you have a hot spot of 110 degrees Celsius on your 7900XTX then thermal pads MUST be replace. Huge thanks for your test.
Lower temps are good, it will live longer. Nice graph!
I'm more convinced that this will be a good mod (as in replacing the thermal pads) used on laptop VRMs and memory chips covered by thermal paste.
well done, adds longevity, I miss the old artic freezing GPU coolers that would do this and reduces noise
Better temps only leads to better gaming only if you’re currently dealing with thermal throttling. Any other case, it is just healthy for the gpu to operate way below the throttle limit
I need this for a 980 TI. Just because to pads are original and it should get redone and I like to keep things as cool as posable.
You get nice temp drops, that alone is worth the work and the relatively cheap price of the pads.
Well for mine 3080 ti FE 2mm pads are too thick i just switched on gpu and fans went crazy so I opened and thermal paste didn't even touch cold plate...1.5mm maybe even not sure feels like 1mm could be best choice..Damnit i need order new pads 😂😮💨
Yes i just experienced the same thing. 7 months later xD
I'm curious how would it be with a laptop... Gonna check it out in this week 👌
Need to measure the clearance in relation to the gpu die to determine the thickness of thermal pads needed.
Thank you , for the video.
how do you know what thickness for thermal pads to use?
TYVM for the video.
One more test could have been noise reduction i think that will change
I use MP14 thermal pads and they work great too!
2 years ago i bought a second hand aorus xtreme gtx 1080, and i changed only the thermal paste, but now since the gpu is 6+ years old i am also thinking of replacing the thermal pads as well when i open it in a few days, those 6+ year old pads are probably worth replacing with fresh new ones, not for gaming, but to extend the life of the gpu as much as possible
Great video... Would you recommend 2.5mm pads on back plate? Really don't want have bow on back plate from 3mm pads
I wouldn't really recommend pads on the back. But if you're keen to I think the 2.5 would work. The bow has subsided since mind you.
I'm curious what the ambient temperature of the room was during the tests... Perhaps those in hot climates would see even more benefit from changing thermal paste/pads
Yeah Surely would depend on the time of year for me. It was about 21-22 C for these tests
But if we overclocked memory after we would get increased performances ?
Indeed. Not my wheelhouse...yet.
Wish you did this on a older card, i'd like to know if the thermal pads degrade over time
U Forgot the Final test
Which is to add thermal paste on pads
i have 3080ti but TUF from asus, and first thing i do is to undervolt the gpu, and my temps drop to 60st in metro exodus in 4k with no hit in performance :D, now im waiting for MX-6 and gelid ultimates and will drop even more temps from those manufacture stock paste and pads
I think its maybe the Founders edition Design cant be bring the temperature out so good,if you do the test on a custom design.
You will be get a better result.
I do that with my 2080 rog,new paste and new pads, and i dropped from 83c to 70c maxed out 1 hour pubg all max settings.
idle temps was over 32c too and i dropped to 27c
Replacing thermal paste is an amazing solution to older cards (2070 Super in my case) having fans ramp up and down constantly. Fixed my issues and dropped the temps by 10+ degrees. Maybe performance has also gone back up to closer to when it was knew, but that's not really the reason to mess with the paste, its the irritating fan noise.
I assume your 2070 Super was also just randomly jumping up to 4000rpm, ignoring any fan control you have set right?
@@LEK Exactly that yes. No fan controller/setting worked. But changing the paste worked completely. I did not change the pads, I was careful not to damage them. Maybe I should have done them as well, but the results were still very good.
@@crispycarrots Yeah had the same issue on my 2070 Super. They charge 700 bucks for the card but cant put the effort in to properly apply quality thermal paste and pads which arent that expensive at all.
@@LEK Dude, that is happening to my 2070 super as well!
and after tried all imaginable under the sun fan control software, it did not matter :D. I guess it is time for paste/pads change.
@@GameWithTonyYT I can guarantee that this the fix. Doesnt even have to be the pads. Paste will suffice. They use super low quality thermal paste.
What about a test with no pads and paste ?
thanks for this test this test amazing work
i change thermal paste to memory i have 60 c to memory after change that after 2 years on rtx 3080 ti trinity
people dont upgrade the thermal paste and pads for performance, it's for extending the life of the GPUS by years. Heat is a major culpert in product lifetime and dropping memory temperatures by nearly 14-16C is huge. Also, if you want performance, with so much more headroom with lower temperature this opens up the door for overclocking the GPU which is another reason as to why people do these mods.
Asus in their infinite wisdom only covered half the memory with pads on my old 2060. They also seems to be low thermal conductivity. A $15 (Canadian dollar) multipack of basic pads allowed me to redo the VRM, add pads to the backplate that it didn't previously have, and fix the memory all at once. I still have more than half of the pad material left.
It's a bit sad that budget pads could be improve what was stock on a "premium" card. The strix models are a scam.
How does the Copper Shims mod compare to new thermal pads?
From my research copper shims are better hands down.
Lower temps might lead to a longer life for those components. I think it's worth it in the long term. At some point you can resell your GPU.
I would believe these tests if every test were on par with the stock pads that had been in use for 6 months compare to the other tests who were fresh install.
the lower memory temp is not the worth it - longevity ?
Are gelid thermal pads good quality? How long do they last?
I had to use new thermal pads because the old ones were torn down. It was not my GPU. The original owner was smoking. The "smoke" "glued" everthing together. There was no option than replace old ones.
there new pads the 15w/m-k are good. i use them for my sapphire toxic 6900 xt (pushing that air cooled timespy record). the biggest difference is within the hotspot. where i was maxing 105c on runs. but now with same settings it barely cracks over 100. thanks for the vid tho
hey i want ask u something about thermal pads gelid I bought a thermal pad gelid-ultimate and that arrived gelid extreme
did u think there is a difference between the two and that was cheating right?
so I called them to return that right or i can keep it?
Any improvement is worth it.
7 months old lol I bet it would help temps much more on my 6 year old 1070 😅 I'm gonna try it
What thickness of pads have been used on the front and back of the Gelid Extreme?
2mm on the front 3mm on the back
Changed thermal pads on my gpu yesterday.. Before i had temps around 62-65 and hotspots over 100° when doing heaven 1440p ultra which lead to the gpu crashing.. Now i get hotspots around 75 after 10 min with 50-70% fans almost the same "gpu temp". So if you run into issues with high hotspots etc, either rma it or open it up and swap pads / paste took me around 1h