I use AMD's method on all CPU's (I call it "x-dot" :D). It gets great coverage with small use of paste (5 dots isn't enough for this CPU). I still would use more on mid dot, and smaller ammount on corner dots. That is one HUGE IHS... is it 4-6 times as big as LGA 775 ? An all IHS CPU's history live comparison would be great :)
I always just make a smiley face when applying thermal paste. If the paste is happy, then the cpu is happy, and the cooler is happy too! They'll all be happy friends, unless you set the voltage too high then they all burn to death.
This is still handy almost 4 years later...I just Frankensteined a 1920x system together using spare used inventory I scraped together from the darkest corners of the interwebs. Your results were a perfect match to my field tests (as discovered when I put the cooler on backwards lol). Thanks for such a great collection of very useful info!!!
I have always spread a thin layer of paste over the whole CPU, but in my last 2 builds, i have used the small blob in the center method. I am not sold on the blob method as the temp differs a lot more between cpu cores when idle. Spreading a thin layer over the whole cpu gives me a similar temp on each cpu core when idle. I dont think i am going to use the blob method anymore. It would have been good if we say an even spread over the whole cpu here for comparison.
Personally, because there is 4 sectors to cool on the Threadripper, they should be putting a small diagram on the Threadrippers cooling plate to begin with. They can have their current logo still visible, then also put the 4 rectangles on the cooling plate as an outline of the 4 sectors. Doesn't make sense that we as customers have to use stencils of sorts. Something so simple, yet the end result would be magnificent, then we would surely, more accurately be able to put the thermal paste on the CPU for proper coverage.
Basically, it doesn't matter. You are supposed to have a thin sheet of "paste", to as much of the contact area between the cooling apparatus and the CPU, as possible, ideally the 100% of the supposed contact area. Sure covering just the cores provides the greater efficiency between cooling and quantity of paste and actually is good enough but still having all the contact area (barely) covered will provide an improvement, even if less so if you take into consideration the paste wasted. More paste is detrimental only if it translates to thickness of the sheet OR if it means that the sheet became larger than it should and dropped out of the CPU's surface (which might or might not short circuits depending on conductivity and what it came into contact with).
This is why I subscribe!!! Where else will you see medium blob and heavy blob in a graph? Plus I'm sure tons of people out there are wondering what to do when it comes to this!
I work in power electronics. The rep for the semiconductors we use recommends as thin a layer of thermal compound as possible, only a few molecules thick, spreading with a credit card. This is just to fill the microscopic gaps left in the machining of the heat sink. What he recommends even more than thermal paste is phase change thermal sheets. Just put one on, and when it gets to 52 centigrade the materials phase changes, and it becomes a perfect thermal interface. How about trying this on a cpu? I would be very interested to see the results.
From my optical contacting attempts, starting from a single blob introduces far fewer (insulating) bubbles which is a good reason for why the blob performs the best.
"Air bubbles are not a issue at all; the spread of the thermal paste, a solid, with pressure, and heat cycles, totally kicks out any air between the thermal paste and the surfaces. If that was a issue, placing those silicon gel protections on smartphone displays would be way harder because of the surfaces being drastically different, or doing a display change would be almost impossible to kick air bubbles. Also, not the X method, but a + plus sign method, is the one that spreads the most without yourself spreading it."
@@calypsojordana3274 it is much easier to peristaltically pump the air bubbles out when you have fl3xible surfaces to work with and can see the bubbles you are trying to remove. Mounting a heatsink you have neither
@@calypsojordana3274 It pushes the air out towards the edges if you start from the center. If you push on a bubble from all sides it doesn't go anywhere.
A bit wild to see that a X62 isn't really able to cool this CPU effectively with that power draw... Time for a 420mm AIO cooler with a larger mount designed for TR! Come on NZXT, impress us! :)
It’s the size of the tubing and the thickness of the plate. Threadripper probably could do with a super slim cold plate, and a larger pump block with slightly thicker tubing
Dude, thank you SO much for the videos you upload, specifically for these threadripper ones. I run a small project studio for audio and video work, and I am pre-ordering the 1950x with the Gigabyte AORUS Gaming 7 board. I JUST bought my Corsair H115i cooler a couple weeks ago (before I knew I was going to able to swing the Threadripper setup) and your videos on the cooler/paste topic have helped a great deal! Keep up the great work!
Result: Two straight long lines are the best for THREADRIPPER, *IF* you have a cooler with wide rectangular plate that fully covers it. Worst: "X"-shape and "one single drop in the middle". Medium tier - "five small drops". Runner up - "two lines with mid drops". P.S. Enermax just now has announced LiqTech 240 AIO that has a FULL wide rectangular plate footing for TR4. So there's that.
I like that you do this stuff methodically as humanly possible. Though the webpage articles are more concise, your videos are more comprehensive, and yet don't have a bunch of gimicky fluff like many other techtubers. This vid was extremely informative, and will keep a lot of this info in mind when I build my TR rig.
I'm not a ThreadRipper owner (at least, not yet) but my method: 1st-spread a very thin layer of TIM's on the heat spreader. The idea is to squeeze paste into the microscopic crevices & cracks in the heat spreader. So the best way to do that is to apply & spread the paste then scrape off the excess leaving a very thin layer. When done, then use your usual method of applying TIM's.
method i've always used is an x with four dots, one dot in the middle of each v part of the x. always had the paste spread perfectly across the entire cpu lid.
I am guilty of over spreading, but I also get some high over clocks on air that run fairly cool. My method is a big glob in the middle and spread with a razor blade. I barely have an seepage, and I always run at about 26c/45c no matter the chip. people have always question my temps. I have never killed a cpu so far.
Explain why blob is stupid? The spreader will make it even with pressure. Sure it won't be absolutely perfect, but air bubbles will be less with the pressure method.
even though I have no desire to buy a Threadripper...you sir are freaking awesome to take the time and abuse for doing this especially considering there isn't a proper cooler at launch for these cpus..
I love your videos for what they cover, but they're SO LONG for the information that they present, and often your conclusions could use more in depth analysis of the data you gathered. Keep it up!
Myself i personally put a small dot of paste in the middle then use plastic spreadder to spread through the entire CPU! make sure to do it small bits at a time and not too much of a dot that will leak over, once the metal of the CPU is fully covered with no leakage of the paste, then i know its safe to apply the Heat Sync! Thats my preferred method as this i know u get full coverage of the CPU!
If you want patterns with normal mounting pressure, put cling-wrap on the bottom of the heatsink/cold plate. Wouldn't recommend doing a burn-in with that setup though.
I'm guessing AMD's recommended method is so the load is even among each core. Depending on your compound, a giant blob in the middle could be enough to indent the IHS. Meanwhile, the parallel lines likely create air pockets. I think concave lines would work better.
Great vid. What if going forward for tests like these, you make the plexiglass piece more of the size of the IHS and you mount a cooler on top of the plexiglass for a more accurate spread? Then you just pull it off and boom, more accurate then just pushing down on it with your hands.
Call me crazy, but I combine a (thin) spread with a (small) blob in the middle for standard CPUs. Best of both worlds! (Maybe?) For the Threadripper I'd probably use four small blobs, though.
Doesn't the heat spreader do just that-- spread the heat? under continuous usage, the whole surface comes to an equilibrium depending on how much heat it's getting, so focusing on where the TIM should be, instead of trying to maximize surface area contact for a given heat sink, seems a little silly? If I disable all but one core on my ryzen 1700, after a few hours, temps reported for each core will be within 1-2C.
Hey man I've watched your videos before and always found them useful and full of information. You're doing great work and I finally subscribed to you today. Only one thing annoyed me in this video, the table isn't exactly leveled and I can see everything on the table rocking. Easy fix! :-) Thanks! Keep up the good work!
If full coverage of the bottom of the cold plate is what is desired then it seems like using the spread method on the bottom of the cold plate would be the best and most optimum method.
Any time you want to more more specifics read the article.. "We used an unbranded thermal compound from a liquid cooling factory for all tests; it is the same compound used on Asetek CLCs"
That's why i love you guys. 100% up front about where inspiration comes insteam of stealing entire video scripts wholecloth and trying to be hush hush about it.
That photo of the 5 point method looks like the dots aren't placed very well compared to the dies. It seems like you could have gotten the rest of that edge coverage if they were more in the center of the dies.
As long as it squishes out and covers the whole spreader it hardly matters, more important is using good compound and how well the surfaces are produced, in fact using no compound at all is only a little worse at transfering heat as long as theres good contact it works
With those temps in mind from using that cooler, I think a Ryzen CPU will be the first CPU I buy that will actually benefit enough from a custom cooling loop where the cost is actually justified from performance more than aesthetics.
I suspect that using a center blob would allow all if not most of the air to be pushed out the sides where it can escape where some of the other methods could trap some when the paste fronts meet up during spreading. 2nd point about a single blob is there is a much higher concentration of force on the paste which would spread and thin out faster (of course slowing down once spread some), so while the others are still spreading under load and being tested the blob method is more than likely spread thinner by that point and probably is why the performance in the test is better. The thinner the paste layer the higher the thermal transfer rate. over say a few days or weeks the others should catch up in performance and all should improve as well. My method I found with testing to work best is to spread a good amount paste over each mating face, CPU and cooler and then using a lint free cloth I rub the paste thoroughly into the surfaces without removing it. Then I use a plastic card as a squeegee I scrape off everything I can. What this does is it forces the paste into the microscopic grooves and texture of the surface filling the voids so majority of gases are pushed out and replaces with paste. The surface looks clean but discoloured, darken using most pastes on the market, I use AS5. That is the paste embedded into the voids. Then from there I proceed the normal blob method.
If you use the proper heat sink, water block, or AIO for AMD Ryzen Threadripper like Enermax Liqtech 240 or 360. I recommend use spreading method cover entire CPU lid, like what TechYesCity like to do usually, it spread like perfect evenly all corner.
Man, I am such a nerd. It's Friday night and I am watching a two-year old video of someone testing thermal paste applications to a CPU I have no desire to own.
13:30 Thorough look at it, thank you. Here's an off wall question, does anyone happen to know the torque setting for a third party torque driver when mounting a CPU for the first time, or ever, or just in general for different parts being attached on a DIY build? Thanks!
Is this method any good? Use the thin spread layer method(possibly a little thinner than normal) on the cpu or heatsink surface(whichever is smaller). Press the cpu and heatsink together making sure it's around the point it will be when mounted. This will create an imprint showing the surface area shared by the two surfaces. Lift the heatsink up and use the paste that was already applied to spread a very thin even layer in the space shared between the two surfaces on both the cpu and heatsink. Make sure it's even and covers completely(the shared surface area). Then apply a very small pea on/over each die area. Use as small of a pea as possible. Then press the heatsink back on top of the area imprinted and rotate back and forth while pressing very gently on the heatsink. It should feel like it's floating on the paste. Without lifting the heatsink align and fasten the heatsink and try. If done right it should not leak anything out. I do this on the thought it will fill in micro crevices. This method also scales with any level surfaced varied size die/heatsink and logically produces full, basically exact, coverage. I'm not sure if there is anything bad about it. Are there any problems with it that you know of? I don't really know anything about how thermal paste should work on a technical level. You can do the gentle rotation part while the heatsink is partially attached if needed. It should help spread the pea sized dots around. Also, Is it bad if small amounts of black substance are left on a die surface. Mainly left after cleaning with just a paper towel. I usually then clean with fresh paste itself and a further wiping down to remove as much as possible. But there is some small residue left afterwords normally.
Thanks for the exhaustive testing. But I have a suggestion: What about trying without any thermal paste? This would clarify the issue that some people have with it: after all, the best heat condutor is direct metal to metal contact. If there are not enough air bubbles to justify applying the paste, it actually has a detrimental effect.
Do I something similar but use a piece of plastic food wrap over one finger to make a nice thin, even layer. The paste is only really meant to fill any tiny air gaps in imperfections on each surface. The smoother the surfaces the less you should need.
Great video, as always. I really liked seeing the plexi even if it isn't representative of real world spread; I've always worried about air bubbles being trapped when using anything but the blob in the centre method but seeing the spread with all the methods really put my mind at ease! Thanks!
^^^ Exactly This ^^^ So all the people saying you should have used this or that over the plexi - Or Full Spread - or any of the other nonsense.. This video gives you the real world use case example; not some super long drawn out fringe unneeded complexity.
@@tisjester As far as I can tell, from observing every "Full Spread" benchmark I could find, spreading all over basically always results in the lowest temperatures on the bench given enough compound is used. From a physics standpoint this makes sense because theoretically, if the goal of thermal paste is to fill in tiny air gaps between the cold plate and the IHS, you would want to cover every square millimeter of IHS coldplate contact area right? Then why would you ever use a method than only covers part of the contact area if done incorrectly? Just spread it all over, more is more when it comes to thermal paste. There is a video on YT of a gentleman taking some Arctic Silver 5 thermal paste (literally has silver in it) and putting it UNDER an AMD cpu, installing it and the computer booted up just fine. Full Spread with a blob on top of it if you're worried about "air bubbles'
so basically just make a huge splodge at the middle of the Z. Also DAMN, them temps when Overclocked leave a good bit to be desired, but given the number of cores I'm not surprised, will be interesting to see how the full coverage plates turn out, I suspect some more significant temp changes.
Did you fully burn-in and cool for these results? Otherwise, the gradual increase would be indicative of the progression of testing alongside ambient water temperature...
Great video mate. Such simple tests for any cooler vs cpu thermal coverage. This us a test that can be done on any build to test coverage and temps and to me is a bit of peace of mind for the small cost of a syringe of TP. 10/10 👍😎
Thank you for another really interesting and thorough test! A bit of an offtopic question, is the reason you haven't done any benchmark/performance test videos for this yet because you're not allowed to until release date? I would love to know how this processor handles some very specific x264 presets before i consider getting one of these myself.
Is the EPYC processor able to be used in a ThreadRipper Motherboard? Isn't the EPYC essentially a ThreadRipper with all 4 dies active? So then wouldn't the best cooler solution be to have a heat transfer plate re-designed so as to be rectangular in shape and to coverall 4 die locations so that a cooler solution would be both ThreadRipper and EPYC compatible?
great info however 5c is 41F 15c is 59f you can never get below ambient liquid or air, so were you in a fridge during your tests at idle?? how do you account for this?? or is it the temp difference between idle to load??
My old method is hand spread very thin with a bit of a peak, ridge, or pyramid near the middle. Conceptually like a thin spread for coverage plus dot in the middle to displace air.
i always just used card or something thin and manually spread it thin layer all over the heat spreader then let the cooler press out uneven spots always worked fine for me even gotten some what cooler results then using just blob in middle method since not insanely caked on
I think you should do this again actually bolting the cooler to the socket. You can make up for the thickness of the plastic sheet by using spacers of the same thickness with any bolt-on cooler that uses the 4 corners.
if you made a aluminum frame with a thick plexiglass inset you could tighten it like a cold plate while still getting a good window. If you wanted to do it the clear way with more accurate pressure.
vabese XD. I actually probably wouldnt do the copper grease but car radiator would be good. P.S- using copper grease on core 2 duo(does pretty well 50C maximum with Scythe Ninja 4 passively cooled)
ColtaineCrows I knew a guy from Germany who used a car or truck radiator with his set up. Had it under his desk leant against the wall with a bunch of various fans attached with wire. It performed really well ha.
I'd just stick it outside and let it cool passively, don't want to go too overboard in "sub ambient" cooling. But yeah, it would work well, it's meant to cool off lots of liquid with several kW's of heat getting pushed into it, having to keep that liquid at around 90*C even while the car isn't moving and using only a dinky little fan in many cases. Cooling a mere ~200W or even 500W would be a piece of cake. Anyway, there's the whole mixed metals thing, but a G30 auto-coolant or similar should have the correct corrosion inhibitors to make that less of an issue as long as it's mixed correctly.
have you guys tried reverse application rather than on the cpu on the cooler before it goes on, yes it may be a little more messy but it would optimize contact for the cooler since the cpu's contact is the bigger than most coolers...
Am thinking about swapping my AIO with a noctua air cooler for my 1950x. Curious if youd be willing to revisit this with Threadripper 2 but instead of different patterns, play with the blob method and weigh the thermal paste tube to get an exact gram measure of how much is needed before the point of diminishing returns is reached.
Redo this test after you get a cooler with full TR4 wide rectangular plate footing on your hands. Not that small-ass, wrongly shaped AseTek. Fortunately for you - Enermax just now announced the availability of LiqTech 240 TR4 AIO with a plate that FULLY covers the entirety of THREADRIPPER's socket. This thing should (and highly likely will) work way better than AseTek.
It may have been an example but the problem is people may see that video and continue to think that putting a tiny little blob on paste is still the way to do it. From the size of the blob they put on in that video and comparing to what was shown in this video there will be problems.
I personally highly doubt that anyone who decides to buy an expensive, high quality, high-tier AIO water cooling solution, is so utterly inexperienced or downright stupid as not to do the thing the right way. People such as these won't bother and would just buy an air cooler, instead of "messing with water". Even if it's "just an AIO" and not a hardcore solution that is FCWCL, you still need to have at least some minimal experience accumulated of dealing with computer hardware, before you go and decide to buy yourself one. Air, on the other hand, is, mostly, absolutely "fool-proof" (ESPECIALLY if it's something like a Noctua or a Cryorig product) thing. At least that's how that usually works.
Not everybody who buys these high end systems builds them themselves. I new someone who bought a boutique HEDT system and there were no air intake or outflow fans on the thing - it had an AIO and the builders told them that was enough airflow. Sure enough when it was overclocked it overheated which i cured by installing more fans.
I imagine heat is also a contributing factor to those pastes thinning out and spreading across the IHS (partially explaining the difference from the plexiglass "dry run")
I think a (thick) line down the middle would do very well though as shown in this video, the application method hardly matters. I would attribute that to AMD soldering the IHS rather than putting mayo underneath
if you made that piece of plastic a little longer and drilled holes through it where the mounting holes are you could have simulated the pressure more accurately with the cooler while still being able to see how it would spread and while keeping the cooler clean for that step.
Why not pinch the plexi between the CPU IHS and the cooler? I know the plexi adds depth but you could still use it to clamp down to roughly the pressure. Removing the cooler would then reveal the plexi and paste spread without much disturbance.
I really wish you tried a circle!!! I got into an argument recently with someone who insisted that it would completely spread through, as if air doesn't exist.
Nice Video. The heavy and medium blob method doesn't produce small air pockets as the others where thermal paste has spacing between application areas. How about the old method of making a thin layer on the copper plate of the heat sink covering the whole circle even filling the screw holes (we used to do it on the CPU but seeing that the CPU is bigger then the heat sink you put the layer on the heat sink)
Where can you get those tubs of thermal grease? I think I'm going to try a good sized bit in the center of the IHS and additional 4 smaller globs on the die itself.
1st Take a tour of CLC manufacturer.. 2nd Use your long gorgeous hair to distract the person giving the tour 3rd Swipe a few jars and hide them in your large cargo pants pockets. 4th Run like hell when they try to stop you and ask what that bulge in your pocket is!
Great video! I wonder if manufacturers will start building AIOs with bigger blocks/pumps. They've got all that contact space to use- why not take advantage of it by fitting a stronger pump or larger reservoir to the CPU socket assembly? :D
what you do is take those nzxt pucks, put paste in the magnetized connect points, put to top part of the puck on top of the cpu and the bottom puck under the mobo and let the magnetic pull squeeze and spread the paste with even pressure. Then take the closest nzxt tshirt and wipe of any overflow then remove the pucks and install your cpu cooler of choice to your now, perfectly spread and covered, threadripper and begin the overclocking glory!
Article, for folks who prefer a skimmable format: www.gamersnexus.net/guides/3013-amd-threadripper-thermalpaste-application-methods-benchmarked
You've gotta update that article author photo....
I use AMD's method on all CPU's (I call it "x-dot" :D).
It gets great coverage with small use of paste (5 dots isn't enough for this CPU).
I still would use more on mid dot, and smaller ammount on corner dots.
That is one HUGE IHS... is it 4-6 times as big as LGA 775 ?
An all IHS CPU's history live comparison would be great :)
have you seen latest linus video, he showed a heat sink the make a dull contact with the cpu that's what should be used as amd stock cooler.
i love the little barbs you are throwing out lately, so funny.. you're sharp lol
RIP, work network blocks gamersnexus
I always just make a smiley face when applying thermal paste. If the paste is happy, then the cpu is happy, and the cooler is happy too! They'll all be happy friends, unless you set the voltage too high then they all burn to death.
Hahahaha
aw that is cute. 😆
9:48 "Wait... That hand's okay".
9:56 *puts other hand in hair too*
😂 That legitimately made me laugh.
This is still handy almost 4 years later...I just Frankensteined a 1920x system together using spare used inventory I scraped together from the darkest corners of the interwebs. Your results were a perfect match to my field tests (as discovered when I put the cooler on backwards lol). Thanks for such a great collection of very useful info!!!
Thermal paste application always seem to get people divided for some reason. I've always just put a bit in the centre.
PhilsComputerLab this has four processor dies under the heat spreader
Looking at the tests though... Just put a bigger blob.
ColtaineCrows Yea that's it.
And it seems the blob method still won.
I have always spread a thin layer of paste over the whole CPU, but in my last 2 builds, i have used the small blob in the center method. I am not sold on the blob method as the temp differs a lot more between cpu cores when idle. Spreading a thin layer over the whole cpu gives me a similar temp on each cpu core when idle. I dont think i am going to use the blob method anymore. It would have been good if we say an even spread over the whole cpu here for comparison.
Personally, because there is 4 sectors to cool on the Threadripper, they should be putting a small diagram on the Threadrippers cooling plate to begin with.
They can have their current logo still visible, then also put the 4 rectangles on the cooling plate as an outline of the 4 sectors.
Doesn't make sense that we as customers have to use stencils of sorts. Something so simple, yet the end result would be magnificent, then we would surely, more accurately be able to put the thermal paste on the CPU for proper coverage.
Basically, it doesn't matter. You are supposed to have a thin sheet of "paste", to as much of the contact area between the cooling apparatus and the CPU, as possible, ideally the 100% of the supposed contact area.
Sure covering just the cores provides the greater efficiency between cooling and quantity of paste and actually is good enough but still having all the contact area (barely) covered will provide an improvement, even if less so if you take into consideration the paste wasted.
More paste is detrimental only if it translates to thickness of the sheet OR if it means that the sheet became larger than it should and dropped out of the CPU's surface (which might or might not short circuits depending on conductivity and what it came into contact with).
I can always count on Gamers Nexus to make a long, in depth , methodical and detailed video on the smallest of tasks.
Are you Montez from workaholics
This is why I subscribe!!! Where else will you see medium blob and heavy blob in a graph? Plus I'm sure tons of people out there are wondering what to do when it comes to this!
Alex Collins truly they are blob masters
While everyone else is still screwing around, Gamer Nexus is posting what theally matters.
Rexin Oridle Tyson is that you??? Lolz
I work in power electronics. The rep for the semiconductors we use recommends as thin a layer of thermal compound as possible, only a few molecules thick, spreading with a credit card. This is just to fill the microscopic gaps left in the machining of the heat sink. What he recommends even more than thermal paste is phase change thermal sheets. Just put one on, and when it gets to 52 centigrade the materials phase changes, and it becomes a perfect thermal interface. How about trying this on a cpu? I would be very interested to see the results.
Good technical comment - those sheets sound interesting (googling intensifies)
Oh boy this is going to be the new "blob vs spread" internet argument for the next coming decade.
spread is better and you know it
also i take issue with the paste used, this is not my favourite paste. I'm out, you're both fired and the weakest link, so triggered
Blob master race!
So what should people without a cock do, Filmer?
thermal grizzly kryonaut blob master race
From my optical contacting attempts, starting from a single blob introduces far fewer (insulating) bubbles which is a good reason for why the blob performs the best.
"Air bubbles are not a issue at all; the spread of the thermal paste, a solid, with pressure, and heat cycles, totally kicks out any air between the thermal paste and the surfaces. If that was a issue, placing those silicon gel protections on smartphone displays would be way harder because of the surfaces being drastically different, or doing a display change would be almost impossible to kick air bubbles.
Also, not the X method, but a + plus sign method, is the one that spreads the most without yourself spreading it."
@@calypsojordana3274 it is much easier to peristaltically pump the air bubbles out when you have fl3xible surfaces to work with and can see the bubbles you are trying to remove. Mounting a heatsink you have neither
@@EndersSaga ?
@@calypsojordana3274 It pushes the air out towards the edges if you start from the center. If you push on a bubble from all sides it doesn't go anywhere.
A bit wild to see that a X62 isn't really able to cool this CPU effectively with that power draw... Time for a 420mm AIO cooler with a larger mount designed for TR! Come on NZXT, impress us! :)
It’s the size of the tubing and the thickness of the plate. Threadripper probably could do with a super slim cold plate, and a larger pump block with slightly thicker tubing
Why not just use a thicker piece of acrylic/ plexi and drill some holes out for the retention screws to mount it just like a CLC?
Christopher Morgado it would probably crack. But I guess you could cnc a metal frame.
Dude, thank you SO much for the videos you upload, specifically for these threadripper ones. I run a small project studio for audio and video work, and I am pre-ordering the 1950x with the Gigabyte AORUS Gaming 7 board. I JUST bought my Corsair H115i cooler a couple weeks ago (before I knew I was going to able to swing the Threadripper setup) and your videos on the cooler/paste topic have helped a great deal! Keep up the great work!
ill just draw "amd" on mine and call it good.
i drew a swastica
Result:
Two straight long lines are the best for THREADRIPPER, *IF* you have a cooler with wide rectangular plate that fully covers it.
Worst: "X"-shape and "one single drop in the middle". Medium tier - "five small drops". Runner up - "two lines with mid drops".
P.S.
Enermax just now has announced LiqTech 240 AIO that has a FULL wide rectangular plate footing for TR4. So there's that.
I love this stuff, this channel is not nerdy, its HARD nerdy. I feel like I am taking a mini I.T. class when i watch this channel.
What about putting the thermal paste evenly over the baseplate of the AIO Pump then applying that to the CPU instead of applying to the IHS.
I like that you do this stuff methodically as humanly possible.
Though the webpage articles are more concise, your videos are more comprehensive, and yet don't have a bunch of gimicky fluff like many other techtubers.
This vid was extremely informative, and will keep a lot of this info in mind when I build my TR rig.
I'm not a ThreadRipper owner (at least, not yet) but my method: 1st-spread a very thin layer of TIM's on the heat spreader. The idea is to squeeze paste into the microscopic crevices & cracks in the heat spreader. So the best way to do that is to apply & spread the paste then scrape off the excess leaving a very thin layer. When done, then use your usual method of applying TIM's.
method i've always used is an x with four dots, one dot in the middle of each v part of the x. always had the paste spread perfectly across the entire cpu lid.
Have SC2, can confirm its baller. Also, glad you guys "got" the go ahead to show thermals!
I am guilty of over spreading, but I also get some high over clocks on air that run fairly cool. My method is a big glob in the middle and spread with a razor blade. I barely have an seepage, and I always run at about 26c/45c no matter the chip. people have always question my temps. I have never killed a cpu so far.
Explain why blob is stupid? The spreader will make it even with pressure. Sure it won't be absolutely perfect, but air bubbles will be less with the pressure method.
even though I have no desire to buy a Threadripper...you sir are freaking awesome to take the time and abuse for doing this especially considering there isn't a proper cooler at launch for these cpus..
can't wait for you to test those new Noctua coolers released for ThreadRipper
I love your videos for what they cover, but they're SO LONG for the information that they present, and often your conclusions could use more in depth analysis of the data you gathered. Keep it up!
Nice video guys! Taking the time for practical testing is much appreciated. It saves many of us a day of our own testing. Thanks again!
Do we know roughly how much paste we need for the most prefered "heavy blob" method...? i.e. 5g, 10g, 15g or 20g etc etc... Thanks in advanced 👍
Gamers Nexus - My first resource for *useful* tech information.™
Myself i personally put a small dot of paste in the middle then use plastic spreadder to spread through the entire CPU!
make sure to do it small bits at a time and not too much of a dot that will leak over, once the metal of the CPU is fully covered with no leakage of the paste, then i know its safe to apply the Heat Sync!
Thats my preferred method as this i know u get full coverage of the CPU!
If you want patterns with normal mounting pressure, put cling-wrap on the bottom of the heatsink/cold plate. Wouldn't recommend doing a burn-in with that setup though.
Steve is having way too much fun with this video.
Thanks for testing it! I don't even want to know how long you spent on it. But I've asked myself that question more than once. :-)
I'm guessing AMD's recommended method is so the load is even among each core. Depending on your compound, a giant blob in the middle could be enough to indent the IHS.
Meanwhile, the parallel lines likely create air pockets. I think concave lines would work better.
Excellent video! Love the plexiglass idea. Definitely learned me something new from this.
This is actually a very valuable piece of information.
Gamers Nexus thanks for testing without thermal paste, I was always wondering what it would produce for temperature.
Great vid. What if going forward for tests like these, you make the plexiglass piece more of the size of the IHS and you mount a cooler on top of the plexiglass for a more accurate spread? Then you just pull it off and boom, more accurate then just pushing down on it with your hands.
Call me crazy, but I combine a (thin) spread with a (small) blob in the middle for standard CPUs.
Best of both worlds! (Maybe?)
For the Threadripper I'd probably use four small blobs, though.
Doesn't the heat spreader do just that-- spread the heat? under continuous usage, the whole surface comes to an equilibrium depending on how much heat it's getting, so focusing on where the TIM should be, instead of trying to maximize surface area contact for a given heat sink, seems a little silly? If I disable all but one core on my ryzen 1700, after a few hours, temps reported for each core will be within 1-2C.
That's the point. People argue about method for application all the time, though.
Hey man I've watched your videos before and always found them useful and full of information. You're doing great work and I finally subscribed to you today. Only one thing annoyed me in this video, the table isn't exactly leveled and I can see everything on the table rocking. Easy fix! :-) Thanks! Keep up the good work!
The Spread Method has always worked the best for me.
If full coverage of the bottom of the cold plate is what is desired then it seems like using the spread method on the bottom of the cold plate would be the best and most optimum method.
What brand/type of tim was used? Perhaps you could add that info to the description. Good video.
Any time you want to more more specifics read the article..
"We used an unbranded thermal compound from a liquid cooling factory for all tests; it is the same compound used on Asetek CLCs"
That's why i love you guys. 100% up front about where inspiration comes insteam of stealing entire video scripts wholecloth and trying to be hush hush about it.
That photo of the 5 point method looks like the dots aren't placed very well compared to the dies. It seems like you could have gotten the rest of that edge coverage if they were more in the center of the dies.
As long as it squishes out and covers the whole spreader it hardly matters, more important is using good compound and how well the surfaces are produced, in fact using no compound at all is only a little worse at transfering heat as long as theres good contact it works
Why is your image/video quality so much clearer and crisper than everyone else?
AlienRenders 60fps
With those temps in mind from using that cooler, I think a Ryzen CPU will be the first CPU I buy that will actually benefit enough from a custom cooling loop where the cost is actually justified from performance more than aesthetics.
The covers on the tubes of the clc look awesome!
I suspect that using a center blob would allow all if not most of the air to be pushed out the sides where it can escape where some of the other methods could trap some when the paste fronts meet up during spreading.
2nd point about a single blob is there is a much higher concentration of force on the paste which would spread and thin out faster (of course slowing down once spread some), so while the others are still spreading under load and being tested the blob method is more than likely spread thinner by that point and probably is why the performance in the test is better. The thinner the paste layer the higher the thermal transfer rate.
over say a few days or weeks the others should catch up in performance and all should improve as well.
My method I found with testing to work best is to spread a good amount paste over each mating face, CPU and cooler and then using a lint free cloth I rub the paste thoroughly into the surfaces without removing it. Then I use a plastic card as a squeegee I scrape off everything I can. What this does is it forces the paste into the microscopic grooves and texture of the surface filling the voids so majority of gases are pushed out and replaces with paste. The surface looks clean but discoloured, darken using most pastes on the market, I use AS5. That is the paste embedded into the voids. Then from there I proceed the normal blob method.
Damn I have to sub.
You doing stuff no1 else seem to do. Thanks.
If you use the proper heat sink, water block, or AIO for AMD Ryzen Threadripper like Enermax Liqtech 240 or 360. I recommend use spreading method cover entire CPU lid, like what TechYesCity like to do usually, it spread like perfect evenly all corner.
How about messure where the heat spots on the chip are, to gain knowledge about the location of the dyes?
Man, I am such a nerd. It's Friday night and I am watching a two-year old video of someone testing thermal paste applications to a CPU I have no desire to own.
13:30 Thorough look at it, thank you.
Here's an off wall question, does anyone happen to know the torque setting for a third party torque driver when mounting a CPU for the first time, or ever, or just in general for different parts being attached on a DIY build?
Thanks!
community.amd.com/thread/223792
Is this method any good?
Use the thin spread layer method(possibly a little thinner than normal) on the cpu or heatsink surface(whichever is smaller). Press the cpu and heatsink together making sure it's around the point it will be when mounted. This will create an imprint showing the surface area shared by the two surfaces. Lift the heatsink up and use the paste that was already applied to spread a very thin even layer in the space shared between the two surfaces on both the cpu and heatsink. Make sure it's even and covers completely(the shared surface area). Then apply a very small pea on/over each die area. Use as small of a pea as possible. Then press the heatsink back on top of the area imprinted and rotate back and forth while pressing very gently on the heatsink. It should feel like it's floating on the paste. Without lifting the heatsink align and fasten the heatsink and try. If done right it should not leak anything out.
I do this on the thought it will fill in micro crevices. This method also scales with any level surfaced varied size die/heatsink and logically produces full, basically exact, coverage. I'm not sure if there is anything bad about it. Are there any problems with it that you know of? I don't really know anything about how thermal paste should work on a technical level.
You can do the gentle rotation part while the heatsink is partially attached if needed. It should help spread the pea sized dots around.
Also, Is it bad if small amounts of black substance are left on a die surface. Mainly left after cleaning with just a paper towel. I usually then clean with fresh paste itself and a further wiping down to remove as much as possible. But there is some small residue left afterwords normally.
Thanks for the exhaustive testing. But I have a suggestion:
What about trying without any thermal paste? This would clarify the issue that some people have with it: after all, the best heat condutor is direct metal to metal contact. If there are not enough air bubbles to justify applying the paste, it actually has a detrimental effect.
They did and he mentioned it towards the End.. They said TR did not like that at all.
Oops, probably skipped that part. Thanks for letting me know.
Yep, just found it at 18:30. Thanks again.
New camera? You guys shooting on a GH5?
How about evenly spreading with a plastic card over the cooler itself?
Do I something similar but use a piece of plastic food wrap over one finger to make a nice thin, even layer. The paste is only really meant to fill any tiny air gaps in imperfections on each surface. The smoother the surfaces the less you should need.
ive always blobbed it..nice to see steve PROVE its the superior method lol....nice work guys
@Gamers Nexus, do you think that OEMs will come up with updated cold plates that properly support Threadripper in the future?
Of course they will.. This way they can SELL more coolers..
Enermax
Also, is there a reference for the AMD TR die layout itself? Would be great to gauge TG application. Thanks again!
Great video, as always. I really liked seeing the plexi even if it isn't representative of real world spread; I've always worried about air bubbles being trapped when using anything but the blob in the centre method but seeing the spread with all the methods really put my mind at ease! Thanks!
^^^ Exactly This ^^^
So all the people saying you should have used this or that over the plexi - Or Full Spread - or any of the other nonsense..
This video gives you the real world use case example; not some super long drawn out fringe unneeded complexity.
@@tisjester As far as I can tell, from observing every "Full Spread" benchmark I could find, spreading all over basically always results in the lowest temperatures on the bench given enough compound is used. From a physics standpoint this makes sense because theoretically, if the goal of thermal paste is to fill in tiny air gaps between the cold plate and the IHS, you would want to cover every square millimeter of IHS coldplate contact area right? Then why would you ever use a method than only covers part of the contact area if done incorrectly? Just spread it all over, more is more when it comes to thermal paste. There is a video on YT of a gentleman taking some Arctic Silver 5 thermal paste (literally has silver in it) and putting it UNDER an AMD cpu, installing it and the computer booted up just fine. Full Spread with a blob on top of it if you're worried about "air bubbles'
so basically just make a huge splodge at the middle of the Z. Also DAMN, them temps when Overclocked leave a good bit to be desired, but given the number of cores I'm not surprised, will be interesting to see how the full coverage plates turn out, I suspect some more significant temp changes.
Did you fully burn-in and cool for these results? Otherwise, the gradual increase would be indicative of the progression of testing alongside ambient water temperature...
so that's what you use for your hair.
Only way to cool that hotness down. ;)
What are those white tubes you used on your corsair AIO?
they're from cable mod
Great video mate. Such simple tests for any cooler vs cpu thermal coverage. This us a test that can be done on any build to test coverage and temps and to me is a bit of peace of mind for the small cost of a syringe of TP. 10/10 👍😎
Thank you for another really interesting and thorough test! A bit of an offtopic question, is the reason you haven't done any benchmark/performance test videos for this yet because you're not allowed to until release date? I would love to know how this processor handles some very specific x264 presets before i consider getting one of these myself.
Yes it is still under NDA..
Is the EPYC processor able to be used in a ThreadRipper Motherboard? Isn't the EPYC essentially a ThreadRipper with all 4 dies active? So then wouldn't the best cooler solution be to have a heat transfer plate re-designed so as to be rectangular in shape and to coverall 4 die locations so that a cooler solution would be both ThreadRipper and EPYC compatible?
Just thinking why don't you mount a cooler on top of the clear plate or mount a retention kit to the clear plate.
great info however
5c is 41F 15c is 59f you can never get below ambient liquid or air, so were you in a fridge during your tests at idle??
how do you account for this?? or is it the temp difference between idle to load??
Delta means difference, in this case difference between ambient temp and cpu temp.
thank you, i am fairly new to this and learning as i go
What? We're not measuring in fahrenheit. Not sure why that's being mentioned.
The temperatures are delta T over ambient.
as i have said im new to the deeper technical stuff, and learning as i go, thanks
No problem, you always have new things to learn :)
Getting the most use out of their cool Plaque! I love it!
My old method is hand spread very thin with a bit of a peak, ridge, or pyramid near the middle. Conceptually like a thin spread for coverage plus dot in the middle to displace air.
i always just used card or something thin and manually spread it thin layer all over the heat spreader then let the cooler press out uneven spots always worked fine for me even gotten some what cooler results then using just blob in middle method since not insanely caked on
I think you should do this again actually bolting the cooler to the socket. You can make up for the thickness of the plastic sheet by using spacers of the same thickness with any bolt-on cooler that uses the 4 corners.
We did that later in the video.
so which dies are active? do some thermal reading between above every die on the ihs and a cooler.
I'm pretty impress about the fact of nobody talking about how much screws must be tighten. In my experience makes difference.
That was fun to watch. Thanks.
if you made a aluminum frame with a thick plexiglass inset you could tighten it like a cold plate while still getting a good window. If you wanted to do it the clear way with more accurate pressure.
I would use automotive copper grease and car radiator for that :D
*hides computers*
vabese XD. I actually probably wouldnt do the copper grease but car radiator would be good. P.S- using copper grease on core 2 duo(does pretty well 50C maximum with Scythe Ninja 4 passively cooled)
I've been considering using a car rad for a long time. Guess when I have my own house I might do it.
ColtaineCrows I knew a guy from Germany who used a car or truck radiator with his set up. Had it under his desk leant against the wall with a bunch of various fans attached with wire.
It performed really well ha.
I'd just stick it outside and let it cool passively, don't want to go too overboard in "sub ambient" cooling. But yeah, it would work well, it's meant to cool off lots of liquid with several kW's of heat getting pushed into it, having to keep that liquid at around 90*C even while the car isn't moving and using only a dinky little fan in many cases. Cooling a mere ~200W or even 500W would be a piece of cake.
Anyway, there's the whole mixed metals thing, but a G30 auto-coolant or similar should have the correct corrosion inhibitors to make that less of an issue as long as it's mixed correctly.
have you guys tried reverse application rather than on the cpu on the cooler before it goes on, yes it may be a little more messy but it would optimize contact for the cooler since the cpu's contact is the bigger than most coolers...
Very informative video. What case is that featured in the video?
David Monk It's a Thermaltake core p3. They have a review of it as well.
Am thinking about swapping my AIO with a noctua air cooler for my 1950x. Curious if youd be willing to revisit this with Threadripper 2 but instead of different patterns, play with the blob method and weigh the thermal paste tube to get an exact gram measure of how much is needed before the point of diminishing returns is reached.
Redo this test after you get a cooler with full TR4 wide rectangular plate footing on your hands. Not that small-ass, wrongly shaped AseTek. Fortunately for you - Enermax just now announced the availability of LiqTech 240 TR4 AIO with a plate that FULLY covers the entirety of THREADRIPPER's socket. This thing should (and highly likely will) work way better than AseTek.
In their (Enermax) video they put a tiny blob of thermal paste on center the IHS. That cant be right after watching this video.
Naturally that was just purely for example purposes.
It may have been an example but the problem is people may see that video and continue to think that putting a tiny little blob on paste is still the way to do it. From the size of the blob they put on in that video and comparing to what was shown in this video there will be problems.
I personally highly doubt that anyone who decides to buy an expensive, high quality, high-tier AIO water cooling solution, is so utterly inexperienced or downright stupid as not to do the thing the right way. People such as these won't bother and would just buy an air cooler, instead of "messing with water". Even if it's "just an AIO" and not a hardcore solution that is FCWCL, you still need to have at least some minimal experience accumulated of dealing with computer hardware, before you go and decide to buy yourself one. Air, on the other hand, is, mostly, absolutely "fool-proof" (ESPECIALLY if it's something like a Noctua or a Cryorig product) thing. At least that's how that usually works.
Not everybody who buys these high end systems builds them themselves. I new someone who bought a boutique HEDT system and there were no air intake or outflow fans on the thing - it had an AIO and the builders told them that was enough airflow. Sure enough when it was overclocked it overheated which i cured by installing more fans.
2:40 excellent demonstration of the dies on threadripper
it looks like what you need is the five dot pattern but spread out more. maybe a big blob in the middle with small dots towards the corners
Does it make sense to apply the compound to the heat sink so the entire surface is covered instead of guessing when applying to the chip?
Great use of their "gift". Basically a "Hey remember this gift when you review our product!"
I imagine heat is also a contributing factor to those pastes thinning out and spreading across the IHS (partially explaining the difference from the plexiglass "dry run")
I think a (thick) line down the middle would do very well though as shown in this video, the application method hardly matters. I would attribute that to AMD soldering the IHS rather than putting mayo underneath
if you made that piece of plastic a little longer and drilled holes through it where the mounting holes are you could have simulated the pressure more accurately with the cooler while still being able to see how it would spread and while keeping the cooler clean for that step.
Why not pinch the plexi between the CPU IHS and the cooler? I know the plexi adds depth but you could still use it to clamp down to roughly the pressure. Removing the cooler would then reveal the plexi and paste spread without much disturbance.
I really wish you tried a circle!!!
I got into an argument recently with someone who insisted that it would completely spread through, as if air doesn't exist.
Nice Video. The heavy and medium blob method doesn't produce small air pockets as the others where thermal paste has spacing between application areas. How about the old method of making a thin layer on the copper plate of the heat sink covering the whole circle even filling the screw holes (we used to do it on the CPU but seeing that the CPU is bigger then the heat sink you put the layer on the heat sink)
Where can you get those tubs of thermal grease?
I think I'm going to try a good sized bit in the center of the IHS and additional 4 smaller globs on the die itself.
1st Take a tour of CLC manufacturer..
2nd Use your long gorgeous hair to distract the person giving the tour
3rd Swipe a few jars and hide them in your large cargo pants pockets.
4th Run like hell when they try to stop you and ask what that bulge in your pocket is!
Great video! I wonder if manufacturers will start building AIOs with bigger blocks/pumps. They've got all that contact space to use- why not take advantage of it by fitting a stronger pump or larger reservoir to the CPU socket assembly? :D
They will, but remember they have to use Asetek designs for their CLC's; due to patents..
what you do is take those nzxt pucks, put paste in the magnetized connect points, put to top part of the puck on top of the cpu and the bottom puck under the mobo and let the magnetic pull squeeze and spread the paste with even pressure. Then take the closest nzxt tshirt and wipe of any overflow then remove the pucks and install your cpu cooler of choice to your now, perfectly spread and covered, threadripper and begin the overclocking glory!