You're applying thermal paste wrong! Here's how to do it right!
Vložit
- čas přidán 14. 05. 2024
- There's a debate that has raged since the beginning of PC building - What's the right way to apply thermal paste to your CPU? Well, we aim to put that debate to rest! We reached out to our community and asked them how they apply paste and then we tested to see how those patterns actually performed! Check out our testing rig, tell us which patterns you'd like to see tested, and maybe we'll make another video in the future!
We'll take you through our testing process, from rig design, to thermal paste patterns, to testing and find which design is the best way to apply thermal paste to your CPU!
*PRODUCT LINKS!*
Arctic MX-4 Thermal Paste: amzn.to/3HqcTMX
Noctua NH-U12A Cooler: amzn.to/3zgCuoI
Glowforge Laser Cutter: glowforge.us/f/NJRAYDBS
*GET SOCIAL AND MORE WITH US HERE!*
Join our Discord server! It's a great way to chat with us!
🎮 / discord
Follow us:
📸 / 2guystek
🐦 / 2guystek
💻 / 2guystek
Visit our store!
🏬 www.amazon.com/shop/influence...
If you would like to support us in other ways, please become a Patreon
😁 / 2guystek
*TIMESTAMPS!*
0:00 Introduction
0:40 The test rig and how we test thermal paste patterns
2:00 The patterns we're testing in this video
3:12 The results!
5:57 The winners!
6:12 Digging in deeper!
6:43 Final thoughts and caveats!
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites. - Věda a technologie
Do you have a different pattern you use that you want to see tested? Tell us how you apply thermal paste and maybe we'll test your pattern!
Yey, my comment made it in the video.
I think the pea sized method used to work just fine when the cpu chip was more in the centre of the ihs compared to now with Ryzen and it’s chiplet design.
Btw, love the methodology and use of acrylic.
2gt acrylic keychains I hear are coming next ;)
Oh @toosas mentioned one that I also wanted to see. A cross with a a dot on each side of the X.
SIDE BUSINESS HERE WE COME! 😂
This video should have 1 million views, Very well done and interesting actually being able to see it.
For the past several builds(10+ years) I have not had to apply any as either the AIO or the air cooler has had pre applied paste and it usually is covering the whole area which backs up your video.
15+ years ago I would do the thin spread covering whole area with a throw away business card out of the mail.
Just curious what a small square for AMD or small rectangle for the new Intel cpu's would achieve, Next time you do an update maybe give those a try.
I give you a thumbs up and I'll subscibe, Will come in quite handy next week as I am about to finally upgrade my old Haswell to Alder Lake and am using the same Noctua cooler, Guess it's back to the old days of the thin even spread. ;)
Sorry for the long post but I really enjoyed this video, Seeing is believing the use of arcylic was genius.
Full coverage but with point finger, not spatula ;-)
Strange to judge success based on strength of adhesion rather than operating temperature.
I've always done the full coverage thin spread method. It just always made the most sense to me.
Good video. Thanks for sharing. I've always stressed to anyone that would listen that a full spread super thin, transparent if possible, is the absolute best. Sticking your pointer finger in a thin plastic baggy to spread was the best. CPU and heatsink surfaces are not perfectly smooth under a microscope and all the thermal paste's job is to fill in those microscopic peaks and valleys to allow 100% contact between the CPU and heatsink. I owned a computer shop for 10 years and now dabble as a hobby here and there for the last 6. I've always gotten great temps albeit a business laptop, gaming laptop, office pc or gaming rig over the years.
All the paste has to do is fill microgaps in the metal between the two surfaces, that's it. Every good cooler and board will allow the heatsink to sit perfectly flush with the CPU so there is no need for it to fill any kind of large air gap, that would only be there from a mismount..
I urge anyone to try this mega thin method as you will do no harm. Worse case your CPU thermally shutdowns as it's designed to do. If you can afford the waste, try the method and then before even booting take the heatsink back off. You will see perfect thin coverage. Then clean and reapply and boot her up. I've had temps some didn't believe until they saw it for themselves, near ambient at idle. A great cooler helps as well. Watch reviews on them. Xigmatek had amazing air coolers when they first popped up back in the day. Not sure how they are now. Anyway, hope that helps someone.
Overall, most of the methods that give good coverage used too much thermal paste. I use Gelid Extreme which is thick pasty stuff, so I start by sticking the tube under my arm to warm it up so it can be spread more easily. Once warm, I use the pea method, and the little spreader that it comes with and do the full spread method (I've always used full spread and I have been doing this since heat sinks first became available) with as little paste as possible. Most people forget that the whole point of the paste is to fill those tiny imperfections in the two surfaces. No matter how well you lap them, it will never be perfect. It's not supposed to look like cake icing... less is more here.
Much respect. Great work. I also love the humility to own a mistake and adapt. Increased credibility. I think I will be trying the 3 line from here on out. I'll do full coverage when I want to be OCD after all testing is done.
Yes, but you didn't measure the temps. In the end, as long as you get a decent area covered out from the centre, all temps will be very similar.
yes i hate they didnt do that
Exactly so it doesnt really matter too much.
Common sense says you should have the whole IHS of the CPU covered, or you can get hot spots. Not to mention, imperfections exist all over the IHS, so you are leaving areas exposed that are not devoid of imperfections. Tests are meaningless and do not tell the full story, because an even, thin layer will always work better than a "heavy center but missing/thin outer layer" method like the pea method. It will bring the heat down faster and this is hard just by measuring load or idle temps. In other words, it does matter!
temps of what a plastic 😂 go look. Theirs videos out for that.
Exactly, this video does nothing for anyone, unless you test the temps after applied.
This also came out right in time for me! I needed to reseat the processor to the heatsink. Thank you to this video I used the three line pattern and had amazing results! Thank you guys for this amazing video!
I have to reseat mine aswell, computers not working and I've got a dram light on the MB. Hopefully, this will fix it...
I use the "square" (AMD) or "rectangle" (INTEL) method to apply my paste, they work perfectly (when i take the CPU cooler off there is full coverage with minimal bleed).
I draw the square/rectangle with thermal paste about the same thickness as you did with the first of your 3 parallel lines.
Noctua recommends pea size in the middle for medium sized cpu's. 5 dot pattern for larger cpus. You'd think they'd have done a lot of testing? I'm assuming they go with the theory that you don't want too much paste because you want max metal to metal...not metal to paste to metal. The paste is only meant to fill up the micro pits in the two metal surfaces.
I don't think you're supposed to place the outer dots right on the edge of the CPU heatspreader, unlike these guys.
I have used both the pea sized dot and the grain of rice short line. Both work great with the old Intel CPU cooler which had a circular copper slug in the center and my Cooler Master G100M which has a large circular copper slug/ heat pipe in the center. The pea sized dot with OK coverage and no bleed out the side would definitely be the way to go with a conductive thermal paste like Arctic Silver. I use Arctic MX-4 myself.
I use 5 dot method all the time and I have no problem getting the paste spread evenly with very minimal spillage over the edge.
This came out right on time, I'm just waiting for my AIO to arrive. I watched tons of videos about this but the majority only give you a "almost everything is fine... with the right amount", not useful at all. This is concise and with a clear advice, that's the only thing I needed, thanks!
So happy to hear this helps! Thanks for the feedback! Consider subscribing!
I've used the pea and the manual spread methods. With the spread, I use an old plastic membership card, credit card sized, but flexible. I wouldn't use the small ice cream scoop spreaders that come with some pastes, too likely to make an uneven spread.
1 dot in the middle has always worked fine for me. Never over heated once.
Great video - Noctua recommend a pea-sized blob in the middle with 4 smaller blobs in each of the corners (a combination of the pea and the 5-dot but with a bigger blob in the middle).
The spread is the way to go I had pea method screw me today and I had to re-paste with the spread to get decent thermals
They make good fans but their paste sucks and the recommended application for it is a complete joke
@@shorty808100
@@chronicalcultivation
I spread a complete thin as possible layer on both cpu and cooler. Always had the lowest recorded temps.
Very professional, like watching an episode of Mythbusters, and the information is very helpful. Good job!
I put an x with very thin lines then spread it into a square with a small gap to the edge of the IHS.
100% coverage, hardly any excess being squished out the sides.
The point of starting with the x and spreading it out from the middle is to ensure the square I created is even thickness across.
Cool Video! 😉
Wait the 5 dots woulda been dialed if y’all didn’t put the outer dots so far out right? Make the 5 dot closer together I’d bet it’s similar to X
I use the 841 dot method - equal size dots in a 29x29 dot pattern. It works great!. It does take most of an afternoon to do the application step. But, no one ever says perfection is easy.
Thanks, this was very useful
You're welcome!
A full, thin spread on CPU and heatsink seems to work OK for me.
I've used the full coverage method for 30 years, always every time.
That MX-4 seemed to be a bit thicker than, say, the NT-H1. I used the 5-dot method with it and got excellent coverage. I suggest you try again with two more pastes - one that is less viscous and the other that is more than the MX-4. Also, try a phase-change TIM like Honeywell PTM7950.
the smiley face my go to, you want your processor to be happy so it doesn't get heated
What we need to know is what happens after it settles, say after 200 hours of CPU use, results would be totally different?
what if you did a 3 line with both sides are just half lines on and the middle one being full line ??
I do X with lines from end points so a Square X and combo ? XD seems to be doing fine for years now
Your 5-dot design used insufficient paste/dot and the dots were too close to the edges.
After watching this video, I'm wondering if a small dot inside a rectangle pattern would strike the magic balance between coverage and application ease? The X and Y dimension of the rectangle pattern should be about 70% of the CPU X and Y dimension respectively, so there is roughly equal area outside the rectangle vs. inside the rectangle. Would it be possible to test this pattern?
Great question! We’ll add it to the (growing) list!
Well heck. I wonder if I should re-do my paste now. I'm the pea sized dot adherent, but maybe three lines is the answer.
Instead of the 5 dot pattern you should try the 9 dot pattern 3x3 to get better coverage.
I just used a crosair kit that had a Stancel and a scraper that was a full coverage. Just wondering what yall think of that. So far I'm impressed with the kit after seeing this video
If the coverage was good after using the stencil was good then I’d say use it! The only real downside to full coverage is the amount of time it takes to do a good and complete job. So if the stencil makes quick work of it, use it!
Please test anti-seize as TIM next! I have used this in a pinch and it worked wonderfully. Had it in my 9600k system for over 6 months before removing it and it looked brand new. Only reason I removed it was to install a new CPU cooler. The material itself is made of very finely powdered aluminum or copper mixed with grease. You guys could do a couple awesome videos on this to show the differences between the two types.
Mine is called copper grease
I've not used it on computers. Only on motorcycles...
I did this today I used pea method I started PC it ran hot, I removed my IHS and paste did the spread and my idle was down to 40 from 60C the pea spilled over the edge and insulated the cpu making it run hotter you don’t want spillage
If you went this far I feel like you guys should have tested temperatures. I have seen another video and despite the Pea method not fully covering the CPU it had lower temperatures than other methods.
You did 5 dot pattern test wrong. The outer dots should have been closer to the middle, which would have give a much better result!
I’m gonna try the X with 4 tiny dots on the inside angle, or an X with a smaller “pea” in the middle. The X reaches the corners which the least alone doesn’t.
Overage doesn’t hurt anything, but if you are using the conductive (Liquid Metal) types, I’m assuming it would be really bad.
Correct, conductive pastes are total danger zone stuff - If you're going to use it, you need to be very careful.
I didn't see the video yet, but from what I have seen in other experiments is that the whole thing is cpu dependant, in other words you might get different conclusions with different cpus, a 12900KS is thermally very different from a Ryzen 3600X.
Long ago on an obscure PC cooling Forum.
The agreed method was full spread. And less was more..... Even to a point they would be lapping the cooler's contact surface to use as little thermalpaste as possible. The idea being more direct between "hot" surface and Cooling element surface. And the thermal paste only as a void filler rather than a full heat transfer medium.
Seems like the consensus today is thermal paste a heat transfer medium.
What changed?
We lapped because the integrated heat spreaders or coolers weren't all that flat back then and it caused bad contact and high temps that even copious amounts of paste could not solve. If you had bad luck with your cpu IHS lapping could shave up to 10c or more off the max temp. What changed ? If i had to guess newer people have to do and know less these days. Back then you had to be more knowledgeable about you parts/settings/overclocking in general or you could kill your stuff pretty easily.
How about the Intel 12th gen bending problem on the middle of the IHS? Should we apply more in the center when we choose to go with the full covered thermal paster?
You certainly can, however I'm not convinced you'd need that much more to make up the difference. The bow in the IHS isn't that dramatic.
@@2GuysTek Thanks for your reply, yes, in my understanding also if we put too much paste on it, it will degrade the performance.
no processor has the cores placed in the same place, the best thing is to know where they are by looking at a diagram and applying your paste according to that
The "pea" method and "5 dot", neither is good for big square AMD and rectangle but smaller Intel. I personally use and recommend the "X pattern" or manually spreading it with a thin layer (which is best overall, but you must not create air bubbles). I had a AMD Ryzen 9 5900X before and now own the Intel Core i9 13900K, both had the AIO coolers removed and the X pattern got the full coverage both times.
The 5 dot pattern is supposed to have the middle dot as big as the pea sized method. That's why it didn't cover the full CPU. Now add the pea sized to the 5 dot and you have full cover.
I'm not saying it's the best, I'm just saying you guys did it wrong lol.
wouldn't the heat from the CPU cause the paste to spread a bit more? I think the many of these would work better in that case right?
Thermal paste doesn’t melt under temperature.
But pump out is a thing.
coverage only matters so much... temps are what really matter.
The 5 dot results do not make sense. If the single dot covered 80%, how does the 5 dots with a center dot yeild such poor results.
Hmm. One dude did test with paste and w/o on ryzen 3900x. The results were that it doesnt really matter. Paste is used only for surface imperfection. Personaly next time i change it on my 3900x i will use just a bit and spread it manually.
And by a bit i think even on 6:14 were applied too much. Radiators surfaces became much better during last years, processors plates too. 2 almost perfectly finished parts wouldnt even need a paste probably.
Always used full ;)
What amount of paste is appropriate for a small rectangular cpu? Thank you!
A generous pea sized amount evenly spread across the IHS of the CPU.
What about a 2 line pattern
"Too much/little" is a function of how thick you applied, regardless of pattern, no? For example, the eight-armed asterisk would have fit perfectly if drawn a little smaller.
Also, air pockets?
ima try 3 line is better i gotta a good felling bout it is fantastic good job
It’s what we have adopted as our go-to.
When in votech training for my A+, my instructor said to use a blob half the size of a grain of rice. He would fail us during an assembly exam if we didn't do this. What an idiot.
why no one tried the spread method with an additional dot in the middle
one type missing - X pattern with a dot in each triangle which I imagine should be pretty good
We'll add it to the list! If we get enough suggestions we'll run another test! Thanks for watching and consider subscribing!
12th and 13th gen are more rectangle. what is the best for that?
An even spread is always the best, followed by 3 evenly spaced lines that run the length of the IHS.
im affraid to have air bubbles if i dont spread it good. i guess the 3 lines will do 4 me
Nice plexiglass tool👌 I also wonder whether how thermal pad would perform (plexi test), as the perfect way as generally expected!? Nice to know since some thermal pad as eg Gelid (at least theoretical) even better perform these days, as compared to eg MX-4🤔
the full is the best imo but very difficult ... theres WAY too much thermal paste used there in this video .. the tinner youre able to get the layer the better the heat transfer will be .. its art xD
I don't know guys. Have you tried the thermals? Because the temperatures may show little to no difference, and that's what matters the most.
Good question! Thermals are most important, but poor coverage will mean poor thermal transfer between the CPU and the cooler. We hope that our testing will help people make a better informed decision on how they apply thermal paste for the best thermal transfer. Thanks for the comment and watching! Consider subscribing!
@@2GuysTek I'm not sure. Although full coverage of the IHS is important, IIRC during the Ivy bridge era the "line method" was good because the cpu die was actually a rectangle underneath the IHS. So if you covered that part of the IHS you would little to no difference from a full cover. I still think a thermals test should be done before coming to conclusions.
Edit: I thought I was already subscribed, my bad 👍
This speaks to the need for a follow up video!
@@2GuysTek yes please! Rly looking for it 👌🙂
5 dots you applied too little on a single dot that was the problem
Same for 12th gen intel hasa different shape
i usually put 2 lines
why not test how much difference the different method coverage actually makes to cpu thermals?
Why not apply the 3 line pattern and then spread that out? Looks like the best of both worlds.
I use x with tiny dots in the triangles.
Honestly I thought of a similar idea but I just simply didn’t have A big CZcams channel so I doubt if the effort would inform anybody
Guys the video is good and the work is appreciated.
However, you did not apply the 5 dot pattern correctly.
As per Noctua and others:
4 mm center pea
Four 2 mm dot applications near the vertices
After application, the center pea will cover a big portion of the cpu in the typical circular pattern. The other dots will expand and cover the remaining areas near the vertices.
Coverage is complete with no overage.
The 5 dot pattern cannot translate to 5 EQUALLY sized dots as that will result in the pattern you demonstrated.
Just and heads up.
Pattern doesn't really matter...its the amount you put. In the 5 dot pattern, place the edge dots a bit closer to center, and more paste in middle dot like in the pea sized one. You'll see its way better than what you did.
Won't spreading it create air pockets?
If there are any bubbles they wouldn’t cause any significant thermal difference.
Results of the line and dots will change with the amount you use though
With enough thermal paste you can cover anything.😅
Instead of paste, try using various clear adhesives of similar viscosity, to prove the concept. Air could be trapped within. You may be deceived due to your paste adhering to the visible surfaces. For 15 years I've worked in engineering in an industry focused on the prevention and removal of air bubbles between adhesives and vinyl and PE films.
Sorry, but no matter how many videos I watch about it, there are so many results, this is like a religion, believe what you want. I haven't seen any stress tests here though, this is just a visual test, whatever anyone likes. Overflowing paste is also sometimes not good because of short circuits and overflowing means a lot of paste and too much thermal paste can create air bubbles. Stress tests have shown that there is a difference of 0.5 or 1.5 degrees Celsius between some methods, which is nothing. I'm also looking at the 5-point method, sorry, but what I saw covered almost the entire space. Full coverage - sometimes an air bubble will form in the center, whether you like it or not, just to be sure, put a small dot of paste in the center. I have good experience with the X method, which you can't go wrong with.
Linus techtips one said, its not needed to cover all die. Just make a decent dot and with pressure its perfect. i do this since years like linus said. never had issues
Just draw a 69 with your thermal paste, that's always perfect
🤣😂
I use full coverage method, using a credit card to spread a minimal amount of paste, after all, you're not filling up wishing wells here.
You all seem to just be going off of looks. Why not add an objective test, like testing the temperatures with different thermal past patterns? Then we can have some usable data.
full coverage is worst for temp. don't fall for it. see the tests that measures temps to. thank me later
Upvote#1.
*you and me
Should have repeated all patterns with thermal testing and you’d see not a lot of difference in temps unless too much or too little paste was used.
3 line wins.
Coverage is not everything. You have to test it with temperature values.
The full coverage / spread methods were discouraged due to it causing air bubbles / air pockets in between the CPU and the Heat Sink.
They can not be seen!
100%
Fibonacci Spread
We’ll add it to the next round! Thanks for the idea and for watching!
Failed to show temperatures.
Stupid test! Acrylic not so tough as copper and after heating one dot cover all cpu, because thermal paste is silicon oil
this test is very bad for various reason: first when you do the "x" for exemple you let lot of air on the middle of you paste so lot of buble on your paste.
second point you never plug the CPU to show the temperature of it so you can say anything about a good circle or everything, we have no data so this test is obsolete
What you're doing here isn't science, it's coffee fortune telling.
The edge coverage isn't even so important. There can't be anything under the edge of the heatspreader anyway, that's a keepout zone where the heatspreader is just glued down onto the substrate PCB.
Thickness however might or might not matter. But you wouldn't know since you measured nothing.
Besides that quincunx application obviously wasn't correct. If you put a dot right on the corner of the CPU heatspreader, 3/4th of that paste escapes, who would have thought! Instead you're supposed to place the dots in the middle of each quadrant.
Gamersnexas did a better job and has a pressure test Machen as well as real good heat testing psu testing load and now there working on testing fans for pc
So you went out of your way to test everything about thermal paste, except actual heat transfer performance................................................................
Amount or Coverage of thermal paste doesnt defien temps lol
Wow, people STILL get this wrong in 2022...
Forget all those patterns. What is the purpose of thermal paste?
-Well, the best conductor of heat in this system, is the metal that the sink consists of, no surprise here, and if we could lap the two surfaces down to optimal contact, i.e. to eliminate air (or vacuum, if you work at SpaceX), we would have the best solution.
In practice, there will be small gaps where the metals are not contacting, so practically we want to replace the air with something as heat conductive as possible:
We then have thermal paste, which is a somewhat good conductor, not VERY good in comparison to the metal bodies we try to interface, but paste is still better than air.
Now; apply a smear to the whole of the mating surfaces with whatever tool you prefer, then drag a razor blade across each, to get AS MUCH PASTE OFF AS POSSIBLE.
Now, to answer the question at the start; "What is the purpose of thermal paste?": If you think of thermal paste as a thermal insulation, that is unfortunately necessary to apply to eliminate the air, you have a good understanding of the process.
the best way is the british flag ...
X
These youtube videos have different opinions. I dont know who to believe. First some say the paste should not reach beyond the square but in this video they prefer the paste to overflow the square.
Way too much past used
With all of these methods you used way too much paste. You are supposed to use a thin layer, just thin enough to fill the irregularities. How did this knowledge get lost over the past two decades?
Honestly who cares lmao. Clearly people have been using all sorts of methods with plenty of success otherwise they wouldn't use them. So they all work fine.
@user-mk4or5yu9r again, all sorts of methods work just fine. Stop trying to look intelligent.
This is way more serious than you realize. Children have been disowned by their parents for not applying thermal paste properly.
@@xuimod well they shouldn't have been little brats.
@user-mk4or5yu9r again bud, plenty of methods get the same results. So yeah, who cares. Cry about pointless things more.