This man is out here literally murdering entire product lines with the scientific method like its nothing. This is quickly becoming my favorite channel.
Only variable this man didnt isolate were the 15+ beers he crushed while doing this experiment. Been watching this channel for a number of years and still impressive as always. Hope you never stop making videos my friend!
There are very few presenters that capture full attention from first word to last. I had a physics instructor like that in 1962, and Michael Arbib had that effect on me five years later in a computability class. Only fifty years passed before you came along and reliably do it to me again. Great work.
I though that he has to endure this boring test so many times, I can stay with him the full length. From the sights after the experiment reset you can relate his boredom.
Controlling for variables while still being practical for experiments/test setups is a rare art form that shows just how serious an applied scientist you're dealing with. This apparatus is a remarkable demonstration of that. I'm in awe.
Thermal Paste cost $6 delivered to your home via Amazon. Buy it. You know it is correct. No guessing. It will last many many years. Simple and Effective.
The problem is it's more difficult to perform these tests over a period of months with power cycles. That's where issues with the cheaper pastes show up.
@@oldmech619 This isn't to save money. This is purely applied science. It's fun, you learn something. Can we make something better than the commercial guys, and can we make that at home?
@@crabmansteve6844 I totally agree with you. This is a demonstration is for applied science. It is not a recommendation to circumvent proper commercial and professional procedures.
I've listened carefully and I'm left with one question: how will the thermal compound perform after a year of use? I mean, it's great that your mixture works well in these tests but how will it hold up over time? Perhaps the next step could be some kind of endurance test involving a much higher temperature over a longer stretch of time. How does it look when you "bake" the compound, compared to commercial products?
I think he kind of adressed this by talking about the water content of the honey, as long as the compound doesn't lose its fixitive ( like with water evap ) its thermal properties should remain
@@crashoverride93637 Evaporation isn't the only concern. Does it slowly react with metals normally used in PC components? Does it oxidize? How quickly? How does temperature affect all of those factors? An endurance test shows you what reality's doing to undermine you. Sometimes it takes a while for those effects to become apparent.
Since the motivation for this project was the apparent lack of public data on the topic, I'm guessing the reason why he's sharing the formula free of charge is so that the time and competence of other people will scrutinise their compound and find its actual properties.
@@Teth47 much like Tesla and fsd the best way to do massive scale testing is to have users do the test by putting the formula out there anyone can do the testing for massive scale
@@ParadigmUnkn0wn ...your kind of proving the efficacy of the method, also as an aside you might be holding fsd to a impossible standard it doesn't have to be perfect to be useful it just has to be better then humans, humans hit things with alarming frequency
This has gotta be the most controlled thermal paste experiment I've ever seen on TH-cam. Thank you for providing both entertainment and education, especially in such a broad range of areas. If I could only watch one TH-cam channel, it'd be this one.
@@1MRSomeguyunless we're talking about personal lubricants I fail to see the relevance of sexual preferences. After saying that I now have a very uncomfortable mental image of Linus using thermal paste as lube. Someone pass me the eye bleach, please.
The question is: Is it really necessary to control it to that point? Measuring temperature difference between the two sides after it has settled should be perfectly sufficient. As long as all thermal pastes are tested the same way (i.e. same surfaces, same pressure, same input (heat) power etc) it should be easy to find the best (and get a good indication by how much).
@asdrubale bisanzio Steve does pretty well controlled tests. Nothing on this level, but pretty good for determining real-world performance in a typical use case. Linus on the other hand veers more into the entertainment side of things. Still fun to watch, and not entirely useless info, but a lot less controlled.
@asdrubale bisanzio But is it actually important to keep the temperature of one side constant if you are only interested in the temperature difference between both sides? Actually it would be much more accurate and repeatable to measure the temperature difference only and let the experiment run until the temperature difference has settled. Instead of introducing additional inaccuracies from the cooler, measuring the 12 minutes and so on.
Oh, and a monologue of 15min as intro is rare on youtube these days. A lot of speakers are just cut after every sentence. It shows how well you know your subject and that you're getting good at this game ! Best channel ever.
Yea, I find it incredibly impressive that he can speak so eloquently and detailed for 15 straight minutes off the cuff. No teleprompter, notes, hesitation. You'd think this was his 100th take, but I wouldn't be surprised if it were his first.
I just wanted to say that you’re a personal hero of mine. I always dreamed about being an inventor and just communicating so clearly these real projects and how we know their better. You’re an amazing influence on the platform and a true scientist. Thank you for making such honest and clarifying videos.
Honestly I don't know where this guy came from... But he came onto TH-cam and has top notch editing and presentation. You never feel dumb watching his videos, even on the most advanced topics. Nor do you feel like it's below you. Honestly this is one of my favorite channels.
@@girrrrrrr2 yeah, really good channel. they aren't selling themself out for outside forces. at least yet. fingers crossed. maybe they have enough capital to be independent. rare breed.
@@cubertmiso I honestly think that he, maybe his son is some form of engineer or something along those lines. And I honestly hope he keeps being independent or at least is very objective when sponsored. Such as this was given to us. Let's see how good it really is.
I really like this video and find it interesting as I'm something of a computer nerd myself. I only feel that there is one variable missing in this video that is considered a fairly important property of thermal paste which is longevity of effectiveness. Maybe an idea for a future video? Either way, your dedication to these projects and results that you get are pretty impressive. I'm glad I subscribed a few years ago. You guys have an amazing channel!
Correct, this is a time zero test. When validating thermal solution it is temperature cycled to accelerate life degradation. Another important aspect is measurement account part to part variability.
Thermal Paste cost $6 delivered to your home via Amazon. Buy it. You know it is correct. No guessing. It will last many many years. Simple and Effective.
This is very important. I've seen compounds degrade over time and some of the cheap ones even leaking out from under the heatsink which led to thermal shutdowns. I generally stick with NT-H1 / H2 or ArticSilver MX. The latter will retain its properties for years and makes taking off a heat sink easy.
As an electrical engineer (technically EIT rather than PE), I greatly appreciate the rigor and thoroughness of these videos. I've been subscribed for a while now. I love how these videos are extremely information dense, and lack the dramatization seen in other channels. I tend not to watch your videos immediately when they appear, but I promise you I don't miss a single video. Well done, and thank you and your crew very much for all the free* content... and the "open source" thermal compound. Cheers! 🍺🙂 *I have youtube red, so I guess it's technically not "free", but it's not paywalled, so it's close enough to deserve a hearty "thank you!"
Professors are employed for their research output and ability to attract funding - not their skills in teaching adults undergraduate courses. Generally professors make the worst teachers at any education level
It's because it's not his job... Every time I've had a hobby that I was great at (IT work, photography, RC, etc) that people started paying me to do, I quickly grew to hate the hobby. Maybe I am strange in that way, but I suspect this is a big reason a professor learns to stop caring. A bunch of 20-somethings that are aimlessly drifting through life would quickly wear on me as well.
@@jacobwcrosby Have you tried working for yourself where you have the freedom to accept or reject work/clients? It's a little bit like musicians that have given up creative control (sometime copyright) to the recording and music industry. This has happened to the biggest musical acts - even the Beatles. I recall George Michael's dispute with Sony. The company would not release George Micheal from his contract so that he can collaborate with other artists, so Michael stood down from any music for a few years until his contractual obligations with Sony lapsed. He hated the music industry at the time (and probably until his death). I tend to agree with you though - money and external non-creative demands often contaminate the passion and creativity that drives a hobby. There are some artists and scientists etc that have managed to avoid these pitfalls.
I have been following your channel for many years now, and seen everything you have posted. Thanks for the diversity of knowledge shown. I just saw the LTT quick response video, and you compound just in my eyes crushed the tech of a multibilliondollar company (they said on par, but that is crushing, given the time they have spend against your inventiveness). You always were a "hidden" pearl, glad your channel is growing, and beeing recogniced more.
This level of test thoroughness rivals the testing of Gamers Nexus. I really think you should collaborate with them for a future project. I'm sure they can properly appreciate the amount of effort you're putting in.
@@signalworks Gamer's Nexus might not, but Der8auer would probably be able to get his hands on the professional equipment to really test the TIM here. Since, iirc; he worked with one of the big TIM companies on the development of their extreme OC TIM used for some of the competition overclocking runs.
I never heard of this channel before this video. You are so good I subscribed in the first 20 seconds. Quality shows easily from the jargon. Im so glad to have found you. Keep up this kind of thoroughness and rigour it's what the world needs more than ever before.
I absolutely love how methodical you guys are. It is soothing to my OCD self. I appreciate the attention to detail and the fact that you leave no i un-dotted and no t un-crossed. Thanks for your videos and your stringent methodology.
..This man is brilliant, and his assistant.. .I'm so jelly.. ..i consider what they both do, LOGICAL.. ..i would kill, to be there.. ..props to both of you..
@@JustinWillis_actual it would show how good the surfaces already where and if the thermal compounds where acting as insulators. Thinner the better the more perfect the two surfaces are. It's probably a huge variable in how you should choose your compound for the job.
Love everything you do. Get so excited every time I see you upload. As a chemist, a scientist, a dad, a tinkerer, and a teacher to my children (homeschool during covid) - you’re amazing. Keep it up!
Use the props in the background, papers and things sitting on the table and anything else you can get into the shots to tell a story. plan it all out, something like you were messing with time machines and accidentally fractured the time line or something starting with little things like a one second jump that just looks like a mistake in the editing, having things zap in and out of existence and ending with all out clones/doppelgangers walking around behind you. Get in touch with Captain Disillusion to do the special effects. It will add a science fiction film to your science reality stuff told through a game of I Spy! It might even get your channel featured by some other mystery solving channels!
I'd agree - I saw the 40+minute length, watched anyway, and was filled with disbelief that the video was actually over - surely it wasn't longer than 15 minutes! Nay, it was in fact 40...
Never, in my life, have I seen such a high level of explanation. I came here to learn a little about alcohol distillation and kept watching your videos all sorts of topics. It's just amazing on any scale to learn from you about anything. I have no words to describe my impression of everything I learn here from you. The pleasure is immense. The quality of the videos is at the highest level. Perfect cuts. Perfect explanations. Perfect quality. I have never been so impressed with the quality of study and investment made here. Just a very high level. 100% Thank you very very very much for everything. appreciate it a lot. No words. Please continue to teach from now until forever :)
"You can believe me that Crest is one of the best thermal interface toothpastes out there" That had me laughing out loud. I can't help but wonder how many did you actually test for this video
I like all the new intros, nice talent going on. I noticed the mountain ale bottle appear. I burst out laughing when I saw all the cans start appearing on the steps. Great outro too, though the music was a bit loud. Thanks for all you do.
I am always amazed at the break down of the steps and how simple you make it out to follow but how in depth it is at the same time. Performing quantitative science with backyard materials on things that seem like they require TONS of money to get into.
You guys are freaking awesome. Not only do you put in massive effort doing some sweet experiments,but you make your research available to the community for free! This is what science is all about.
Am curious about the other properties that one might find desirable in this application; corrosion resistance, durability, etc... Am really enjoying your videos and the tedious ( in a positive sense ) nature of them, youtube algorithm be damned we need more content like this.
You make some amazing content, thanks for sharing this with us. I do have 2 suggestions, however. I've mentioned it in 2 videos now I think, but please consider sending a sample or 2 to Gamer's Nexus in addition to LTT. I'd love to see more data from independent testers. And he's definitely one of the best I know of. He also has a much narrower focus compared to LTT, and a major part of that is thermal benchmarking. My second suggestion is to present your data in a graph at the end.
Thank you for slugging through it, I could feel the "I've done this test so many times..." I love this channel and recommend it any chance I get! Also I knew the cans/bottles were in jest because I imagine you would be sipping your home-brew stuff!
I'd really like to see you send this out to Gamers Nexus as well, those guys are much bigger into careful testing of PC components than the folks at Linus Tech.
I'd have loved to see a summary of the results at the end, to make it a bit clearer how the different compounds compared. I'm also curious how the temperature vs. time chart would have looked. Was it more or less linear? And a suggested improvement to the setup: use an Arduino or ESP32 or something with some relays to automate the cooling/measuring cycle. That way you could leave the equipment unattended for a full test cycle.
AGreed there, I was just going to suggest a temp logger. But You can also chart the temperature inside the Arduino. (I wouldnt use ESP32, they don't tend to have very good built-in ADC's. And, if you use a digital temp sensor like DS18B20 , I have noticed that they tend to be +-3 F cyclically, which requires a ton of averaging. ) I found Arduino's ADC and a simple PTC thermistor to be the best of the bunch
@@gutrali I was thinking they could interface with their existing high-accuracy scientific equipment. It's pretty common for it to have some kind of digital interface. If not, I think the built-in ADCs on both boards are pretty garbage, better to grab an external ADC on a breakout board or something.
@@Z3DZ3R0 they are suitable but it depends on your requirements. Check this out. th-cam.com/video/UAJMLTzrM9Q/w-d-xo.html I had so many problems calibrating an RTC thermistor until I started to use an external ADC. I'm not sure but I think the Wifi onboard the Esp causes some voltage spikes that will affect your ADC readings. Plus it's 10 bit and only 3.3v
Cycles would also be interesting, how does the paste hold up? And: If you'd increase the power it would help with the resolution. You have 0.1K differences and you can resolute 0.1K -> If you'd pump more power in you'd get more dT thus better "signal to noise".
Yeah, I missed the temp level but if he was pushing 100C it would be interesting, seems my old CPU that plays TH-cam videos at a temp of 34C Plays other games at up to about 72 which is a bit over the recommended value of 66. But if I push it harder and it gets into the '90s my computer says that enough of that silliness and shutdown :( Now I mean I'm glad it shutdown because otherwise, I would have already had to buy a new CPU, but this did make me wonder about changing my factory Heatsink for a replacement with a bigger heatsink, pumps and grills and all that seems like too much for me, but a big heat sink and a quiet fan would make me happy, my fan case has actually blown all its fans but the CPU and Video card fans are still working :)
@@basildaoust2821 If your CPU is getting so hot that the system shuts down it's likely that one of three things isn't working properly anymore - either the fan isn't spinning as fast as it should, the fins of the cooler are filled with dust and debris, or the thermal compound between the cooler and the CPU die has dried out (could be all of the above). Also, if the case fans are not spinning, you are not getting enough fresh cool air to help with cooling your CPU, video card and other components inside the case. If you don't feel like changing the case fans, I would recommend removing a side panel and keeping it off so the hot air inside the case at least has someplace to go. Then take your CPU fan off (remember the orientation) and take a vacuum cleaner or compressed air to the fins to remove the dust buildup, then screw it back on and see how it goes (you don't have to take the cooler off the CPU in order to do that). If this doesn't help with your CPU temperature, then it's probably time to change the thermal compound, as it might be dried and cracked and effectively insulating the CPU instead of helping dissipate the heat. This is all assuming that your cooler isn't seriously underspecced for your CPU, in which case you should look for a better alternative, however, you might as well try the inexpensive solutions first.
@@benedictul Yeah I don't think the issue is any of that, it's just if I push it too hard it gets mad. I mean my case fans have died and I have a house fan blowing air through the case with the side panel off. The CPU fan and the video card fans are working just fine. I have a new CPU that is much better, not good compared to the new ones but more than twice as good as my old junk. So yeah the plan is to first run the one or two games that complain at lower graphic settings which does work quite well while I buy a few pieces to put the new system back together. The change does also let me switch to an m.2 drive so I expect some nice improvements. Thanks for the suggestions.
I think "Breaking taps" and his AFM (atomic Force Microscope) and " Applied science" with his SEM (Secondary Electron Microscope) would be good collaborations for looking at surfaces and interfaces plus these guys would be interested in the compounds themselves....great work as always.......cheers,
scanning* electron microscope, but yeah, I'm sure Ben would be a good third party to verify these test results. He might actually be able to produce a standard measurement as well as relative results.
Thank you so much for your dedication to the testing procedure, I would hazard a guess and say it produced much more reliable information than you can get off the box on a commercial product. And it's good to hear your humour survived intact as well! I'm really looking forward to the next video, take care and thank you again.
This is the best quality channel on TH-cam. You have excellent experiments with great explanations and background. I think it would be worth testing without any thermal paste, your surfaces are so smooth it might be better?
Loved the increasing amount of cans and bottles. This is one of the best channels on youtube so many reasons, but first and foremost it's actual science. I only wish you'd explained that it takes exponentially higher thermal conductivity to get closer and closer to the loop temp. I'd also love to see a no-paste run. Off to buy some paste for use on my cpu to test its endurance.
Man I wish you were my science teacher back in the day. I feel like I would have learned so much more and more importantly paid more attention. Excellent job!
I wish I had a science or engineering teacher like you when I was in college, I guarantee I wouldn't have fallen asleep in class. Great content! Suggestion: for all the effort you put into the thermal test, you could have used a thermometer with an extra digit of accuracy so there would be more than 1 count difference between all your samples.
an extra digit of accuracy would be telling as of the comparison between the different pastes, but would still be practically not-interesting... 0,01°C-difference isn´t gonna make your overclocked CPU feel any difference in longevity or performance. So, althought u r right, this change would only raise the "thrill-factor" and production-value of this test itself (raised production-value pays for enhanced thrill-factor), without any other practical significance.
@@klausbrinck2137 The thing is its not .01C difference really. The units arent really useful in this situation its a qualitative measurements that don't apply to cooling a cpu.Its just showing which is better than the next one. It could end up being 10C or not noticible. Like he said the actual temps are only applicable to his testing setup but will show a difference as to which performs better than the other
I like this series and hope you continue expanding on it. One thing you need to test/state is the electrical conductivity of your epoxy and paste. All of the commercial products you tested are considered electrically non-conductive. Also a control test with no TIM would be helpful and a high end test with liquid metal would be another great data point to add in.
I have to say, I have been casually watching your videos for years out of a shared interest in many of the technologies and processes you are knowledgeable about, but this particular project seems likely to push you to the million subscriber mark and above. I really appreciate the sentiment, intentional or not, that a small startup or even someone in their basement can create something that is better than companies with millions of dollars of R&D budget, with just a fresh set of eyes and solid knowledge of basic principles. It's ... motivating!
TE, I'm glad that you took the time for this video. I know it took a long time, and you very much displayed the tediousness of the test - I mean, the video itself was 41:51 - but I wanted to say that I greatly appreciated you going through it all, and as I also follow LTT, I'm hopeful that we'll see their review soon. Best wishes on your new product!
I'd love to see comparison to more thermal compounds. I switched to Grizzly Kryonaut not long ago and this stuff made noticeable temp difference compared to other brands I used before.
Great video and work as always. I feel like you've missed some important properties that should be measured: 1) How much does it 'dry out' / oxidise over time (sometimes can be simulated with elevated temperatures) 2) Some compounds need a couple days to reach peak performance, including AS5, how bout testing after a week? 3) Feel like you need to have sandwiches that hold the different compounds over time rather than always scrubbing them. The interfaces to the sandwich could be designed around them or just all ultra smoothed lapped so it's not performance limited there. 4) I haven't used AS5 as my premium compound in a decade, I use IC Diamond (tiny diamonds) for some installs and Thermal Grizzly's Conductonaut (liquid metal) if it doesn't eat into the heatsink. You're not exactly comparing against cutting edge? 5) The performance of a lot of these pastes will be somewhat muted in practice because many Intel chips have poor thermally conducting heatspreaders with middling thermal paste over their silicon. Would be interesting to do a die delidding or lapping in combination and see how that goes. 6) For ultimate performance, I'd have a delidded CPU with a Graphene horizontal thermal transfer sheet (can't transfer vertically but does great horizontally), then have lots of holes with a silver sheet overlaid including silver going into the holes to provide vertical transfer, and so that the heat distributes at a fantastic rate over let's say 6cm X 6cm, and the silicon die is already incredibly smooth (but not always flat, can be lapped), then the Graphene sheet has enough compliance not to need thermal paste (still good to test), then a heatsink bonded to the silver layer with a liquid metal thermal interface. If you do that, now that would be world record impressive and a heck of a video!
As a filler material, graphene is not good as I'll explain in the next video. If price is no object (and it always is) then we have some disk heat sinks with grown vertical arrays of carbon nanotubes that rival bulk diamond for heat transfer.
One concern I have is how long this will last on a CPU or GPU. I use Thermal Grizzly Kryonaut which is almost the same as PK-3 if not the tiniest bit better and it's guaranteed to last years. How long do you expect this to last?
This immediately entered my mind as well. How chemically stable is it, and if not how long before it begins to degrade as it undergoes thermal stress? How long before I'll need to remove and reapply it, and could it potentially begin to set and harden over time and require specific chemicals in order to remove it? If it does degrade over time, could it potentially harm the CPU lid?
Most thermal pastes use vegetable oil (ie the stuff you use to cook your eggs) to keep from drying out or polyethylene. You can literally take a couple drops of glycerin (vape juice) and mix it with toothpaste and you have the standard thermal paste that dell etc uses. He has obviously done the research and knows what is needed or he wouldn't sell it.
Exciting work! Considering your extensive work on thermal conductivity, cryogenics, and magnetism, I’m surprised I’ve never seen a video on superconductors. Do you have any upcoming projects with superconductors? Thanks
"Hello, this is Tech Ingredients and today we'll show you how to make the worlds first room temperature superconductor from common hosehold materials and some stuff you can order from ebay for a few bucks"
@@BlackHeartScyther Maybe, but those need to be cooled down with liquid Helium or Nitrogen. That's why making a "room temperature" superconductor would probably win you the nobel prize immediately (if it doesn't need a few thousand bars of pressure to be stable) And afaik the'yre all ceramics, so processing a brick of it isn't really an option.
Modern CPUs can even runnwithout any heatsink on them (throtling down so far that they're basically unusable, but still running), so I'd say the temps aren't an issiue, but the clocks the CPU will run at.
@@Donnerwamp True, but the default surface on CPU's and heatsinks are nowhere near as flat as his test rig. A few of the 'lapped' setups on TH-cam would probably be pretty close, but true lapping with wavelength precision like he was talking about is nearly never done on CPU or heatsinks at least on TH-cam. (Literally only 1 video with no numbers posted) His test rig is not true lapped, but that's fine.
*controls for every variable imaginable* "now I can't attest to the exact accuracy of this piece of equipment or how reliable it is, but I've got reasons to believe it's pretty good." I got a good belly laugh out of that.
I get that well-executed scientific methodologies are rare on youtube but I feel like he was actually making a really good point rather than an unknowing meme. It's a testament to his scientific rigor that he understands their test setup's merit is only as good as its ability to stay consistent with itself and when compared to other's test results rather than just assuming that he really did account for every variable imaginable or that the experiment happens to be designed soundly just because it seems that way.
I really wonder how this compares to "liquid metal" thermal interface material(s), wonder if you could have a go at testing that out in this contraption as well.
Those outperform normal pastes by a lot. And so they'd outperform this. They have like 6 to 30 times higher thermal conductivity, depending on what paste you compare them to. Also, they didn't make the best thermal paste until they beat Thermal Grizzly Kryonaut, the best paste I can think of.
Watchable and excellent as always. Would like to have seen a control with no interface material at all. Also a real metal interface such as a solder-joint or mercury pool, even if those are not usable in most real-world applications.
Distilation was what caught my eye on this channel since im brewing and distilling myself at a hobby level, but the other stuff you are filming and the accurate tests you are carrying out is quite interesting and satisfying to watch. Keep it up and greetings from eastern europe!
One more factor to ponder about: perhaps the optimal thermal paste is not the same for a pair of polished flat surfaces, as for two much lower grade (rough, non-flat) pair of surfaces. Thermal paste composition is usually a tradeoff between thermal conductivity, viscosity, etc. It is plausible that the optimal viscosity is different.
How reliably could you clean a rough surface between tests of different materials? Polished surfaces at least give you a fighting chance at removing the previous material. This is a test of thermal pastes and not of surfaces. That would be a separate video all together. He was careful to ensure the paste flowed out around the test cylinder under the spring load, displacing the air, which would minimize the effects of the different viscosities of the pastes one would think (I’m not an experimenter, so I may be overlooking something)
@@gordonborsboom7460 Hmm, cleaning would indeed be a challenge. My point is that in pactice, the surface of an IHS is neither flat nor mirror smooth, and depending on the cooler it might be even worse.
I was having a beer while watching this... then started to see the beer bottles and white claws piling up in your video shots... so I was happy to "have a beer" with you guys... only to find out at the end that no beers were had. Love all you do, keep up the good work!
These videos have been gradually but consistently improving and while continuing to stay focused on hard science. I cannot wait to see what innovations come out of this team
I'm subscribed since the first video i watched, love em and have binged the channel, as a TH-cam premium guy, i love that there aren't any huge sponsor ads that repeat on every other channel, the reson for getting premium in the first place, based in germany buying stuff is cumbersome, so i support with watch time
@@danielpope3579 or do a thin nickel coating as it should not interfere to much with the result and block the gallium from reacting with the aluminum. but you're correct, for precise result they would have to retest everything with a nickel coated material, as blank copper will also be attacked from gallium
Hey i was waiting for this episode, this topic is among my interests! However it took a few days for the video to pop up in my feed and it missed the notification. Will be very curious about the recipe next. It has been indeed noticed that for example Kerafol's paste Keratherm KP92 has 10W/mK but performs worse than MX-2 which is specified at 5.6, in fact it performs worse than just about anything north of 1W/mK. It would be weird if Kerafol wasn't being truthful, they are an industrial supplier rather than a customer facing brand. Different measurement method is indeed a possibility, but another is that thermal conductivity just isn't all there is to thermal paste's performance, as thermal conductance isn't really its primary purpose, it is to provide a high surface to surface thermal transfer while not actually bridging a substantial distance. There could be several reasons for surface to surface thermal impedance, like microscopic coverage, or just something about carrier material properties. Also the layer thickness that the paste ends up spreading out to under a given pressure for sure has to affect performance. Curiously, AS5 was estimated by NREL to only have less than 1W/mK thermal conductivity, 9 times less than specified; as opposed to Shin-Etsu X-23-7762-S and Dow Corning TC-5022, where the test result was similar to manufacturer's spec and several times better than AS5. Also regarding MX-2, it's Arctic Cooling's lowest cost offering, with the mainstream one being MX-4. Both have insane long term stability. It has long been known that in CPU tests, MX-4 performs better, this has been a result of pretty much every test to date regardless of publication. However Igor's Lab (former Tom's Hardware Germany, now independent publication) has noticed that when tested in direct die applications, such as on a GPU, MX-2 performs better! One fairly obvious difference is that MX-2 is harder to spread and potentially requires more contact pressure to spread thinly. MX-4 has lower density, lower viscosity, and higher nominal conductivity. It has presumably less solids content by volume, but the solids that are in there might be more thermally conductive. Up until recently, MX-4 was Arctic Cooling's high performance paste; just months ago, it was superseded by MX-5. Arctic Silver 5 is not the higher end offering, it's a competitor, and a very old one. These are two different companies. Arctic Cooling is from Switzerland and Arctic Silver is from the USA. Arctic Silver's lower cost product is Céramique 2, which doesn't contain silver. So your results have been that MX-4 performs worse than AS5. That's a discrepancy with tests performed by the PC enthusiast press, such as Planet3DNow (defunct), Guru3D and Tom's Hardware & Igor's Lab, that usually in their tests, MX-2 comes out equal to Arctic Silver's AS5 in performance, which was quite a shakeup when MX-2 first came out, because it just feels nicer, is safer not having any conductive material in it, and is a whole lot cheaper too. There was MX-3, but it didn't make a strong case for itself and was discontinued. MX-4 is a substantial improvement upon the former ones in tests. Ceramique never performed particularly well. There is a small number of people who still swear by AS5, so it survives on the market, but it fell out of most tests by now. I wonder if water based materials would perform better if you lower the surface tension. So add dish soap to toothpaste. This is not a practical idea, after all toothpaste is not a material that will last long term and soap isn't going to change much anything about that, but it would potentially allow to test your hypothesis that better surface contact is the reason honey performs better. It won't be a conclusive test, but it can be a data point and more tests can be devised later. I suspect that given you have practically mirror polished and matching contact surfaces, the solids contents of toothpaste could be doing it a disservice and increasing the actual layer thickness, since you don't need the material to actually be thermally conductive, you only the material to have a low surface contact impedance to the both surfaces, and low layer thickness. Real CPU heatspreaders and heatsinks aren't actually like that, it's not uncommon for the integrated heatspreader of the CPU to be a little domed in (concave) due to the manufacturing process, and many heatsink manufacturers deliberately make their heatsinks a tiny little convex, which is most noticeable if you look at ones made by Scythe, reasoning that increased centre contact pressure where the CPU die lives in the heatspreader package is going to make for a better heat transfer than corner contact, and also it can help the excess paste squeeze out. Additionally, different geometry and surface quality of your test rig can explain some results discrepancies vs. computer press. You have not specified whether Noctua paste is NT-H1 or NT-H2 and which Shin-Etsu paste. Also you made a typo in Noctua's name in the spreadsheet. :D I recall there are a lot of pastes with known practical issues - Gelid GC Extreme, IC Diamond, etc, which showed spectacular performance but were known (at least initially, they could have gotten better since) to not last very long, rapidly drying out. This occurs because the heat-up and cooldown incurs some material movement, and the carrier liquid and the solids can then separate, and the liquid is then free to leave the thermal interface area. Do you think you can devise a test, which can demonstrate whether the paste stays in suspension or will separate over time? With these pastes, separation took months or years, but obviously you don't really want to repaste your computer every 6 months, that's effort; do you think an accelerated test is possible which can demonstrate this issue within a week or two?
KEEP UP THE GREAT JOB GUYS!!! Just a sugestion: Is it possible to log the temp curve to see if there is any improvement in temperature oscilation and adaptation from the interface solution (AKA termal paste) ? simple arduino log is an easy go to and an excell sheet can draw a curve really easily... JMHO
I was also thinking the "curve" of the logged data. Once the general curve form is established, more performance information could be revealed e.g. how fast it reached equilibrium, without having to spend the whole 90 minutes and by just logging a few more "points" than only the final (12min, temp) one. BTW, Awesome video, as always!!!
Found the engineer. ♥❤♥❤♥ I agree, BTW! Plotting temperature against time might actually reveal something interesting. I mean... it probably won't give any additional insight... but it *_MIGHT!_*
It would have been awesome if (for fun & comparison) you also included a test of liquid metal, like Thermal Grizzly's Conductonaut. I use that on any interface I can because I get such low temps from it. Awesome video, and love your attention to detail and replicability!
Your production quality has matured nicely over the years. I know you know you are good, but I still think you don't know how good you both really are. Always a pleasure.
I'm surprised you didn't test the indium sheets. I'd love to seem some numbers since you keep recommending it. That and I'd like to see galistan, or conductonaut, tested.
I think he stated the indium sheets required a large amount of pressure to work effectively. I expect the springs in his testing apparatus don't apply 1/4 ton of clamping pressure which he stated is the minimum for indium. The liquid metals also have a secondary effect where they can permanently change aluminum (they are intended for use with copper). Using a liquid metal that flowed into the aluminum structure of the test rig would change/invalidate subsequent tests performed before since a surface wipe won't remove the liquid metal.
Just the sheer amount of educational knowledge on topics is enough to get me hooked. Add in actual testing and tangible lab results and oh my... Now add in the best product that I can purchase?? Take my money
His temperatures are all the same because the two surfaces are mirror flat with no gap. If he used no compound at all I bet it would still read 30 C. The setup has nothing to do with the real world.
@@liambergstrom8183 Tomshardware shows toothpaste at 42 C and most top compounds at 30 C, which is to expect because toothpaste has very bad conductivity. His setup is bad and is in reality more like a bad viscosity tester.
@@rrrlasse2 so it's in the same order...but higher temp would be nice so there's a bigger difference? Like I said? If his setup is ranking them the same order...then it clearly works. The actual temperature has no bearing between one setup to another, it only applies for reference to other compounds on the same setup. The overlap between the two different setups means nothing
@@liambergstrom8183 It's not even really the same order. Most PC publication tests show in CPU use the MX-2 to be equal to AS5, while MX-4 performing substantially better. Curiously Igor's Lab has found that MX2, while having nominally allegedly a lower thermal conductivity, performs better on a GPU in direct-die application than MX4. I have no clue what this means. Higher mounting pressure? Surface evenness for the GPU die vs. the usually warped CPU heatspreader? A lot of potential variables. But it stands to reason that surface contact on the microscopic level and layer thickness are more deciding than straight up conductivity. Indeed mirror like and exceptionally flat surfaces are not what we're dealing with, with PC components, we've got concave heatspreaders and occasionally lumpy direct contact heatpipes to contend with. Which Shin-Etsu paste was used here, was not specified, they have many. Nor the Noctua, they offer two, the H1 and H2. About Tom's Hardware and their toothpaste test. It's a test performed by their former German subsidiary, currently known as Igor's Lab. It may have been lost in translation, but the way Igor tests toothpaste, and he did it several times after as well, he feels it's unfair to test it immediately after installation, as thermal paste performance stabilises with burn-in over the first 20 or so hours, which is something he can work around by running several tests and discarding the worst results, so this is the expected performance long-term, but for water based products, the opposite happens. When just applied, they perform about as well as a standard thermal compound, but afterwards they degrade rapidly. So with toothpaste tests, he instead burns it in by operating the rig at specified power for 12 hours, and measures how they perform after the burn-in! This is a possible explanation for the disparity in the results, why in Tom's Hardware test the toothpaste looks a lot worse than standard thermal compounds, while here it looks similar enough. A lot of people have independently measured toothpaste against thermal compound, and just upon installation, it often delivers very similar performance. There are two other funny and extremely cheap materials i'm aware of that can serve as an impromptu thermal compound replacement. One is copper brake grease, it has been discovered and tried by various forum members and seems to perform about on par with thermal paste, it's ultra fine copper particulate suspended in grease. Another has been discovered by Igor's Lab, and it's endurance gearbox grease, Liqui Moly LM47 with MoS2. Both presumably use lithium soap stabilised silicone or mineral oil as a liquid base. Igor's Lab even used his finding in one of his lower-power systems long-term to see its endurance, and it has not degraded in performance over a better part of the year.
I wonder if the honey performed so well, because some of the heat energy was "used" changing the viscosity. I know it's not a full phase change but It did make me wonder if a substance that changes viscosity with temperature actually "uses" energy to do so. I had a brief (a couple of minutes) poke around on google to see if viscosity changes use energy or not but found nothing useful.
Viscosity is a property due to the intermolecular forces, as is the specific heat capacity of a material. For nonmetals, as temperature increases, some of that energy is stored in the intermolecular bonds which will also affect the viscosity. I would expect the specific heat capacity change over the temperature range in this video to be relatively small and insignificant. While this test did not reach steady state, it is not a short duration test, so changes in heat capacity should make little difference. The amount of thermal interface material vs the power going into the system is small. Another effect might be the viscosity directly affecting the film thickness for a given pressure, but I can't convince myself either way of its effects.
What a class act this gentleman is. We would love to see the results of your opensourced TIM. We all can prepare and test it in different heatsinks, CPUs, GPUs, APU's. Hats off Sir!
I'm actually a bit surprised by the materials used here, instead of just using copper. Because if there is another video in the thermal interface series that covers thermal films like liquid metal, they're going to have to revalidate all of their testing on the setup that can actually test the liquid metal. The tests are idiosyncratic to the specific machine after all, which would make any attempt of cross-validation of a singular test between both machines (say, toothpaste to toothpaste, as an example) would be analogous at best, meaning all pastes should be revalidated for accurate science. With how long these tests are, that'll be taking a long time to do, just to compare products of one thermal interface type to products of another thermal interface type. Though, I'd be more interested in thermal pads, not the 'reusable' thin graphite ones like what Innovation Cooling makes. Liquid metal is analogous to thermal pastes and thermal greases, it's a thin liquid film that aims to serve the same purpose, though I would argue that thin metal films are a different thermal interface entirely, just as thermal pastes are not thermal greases, and just like how thermal adhesives had their own video in this thermal interface series. Thermal pads on the other hand do perform differently by which direction you're measuring, measuring along the Z axis (as Z-up) is a different measurement than what you can measure planarly along the pad's surface. This difference in measurements is how some products have ridiculous numbers as marketing, but their actual performance is lackluster in comparison. While the graphite pads are technically pads, and function in basically the same way, I would rather see them separated into their own sub-category, as they're thin enough to be marketed as a replacement for thermal pastes, films, and greases; similarly, the same can be said about films, indium films and gallium-based films are similar but how they achieve an interface is different.
@@xaytana iirc i think LTT? prety sure it was LTT - it wasn't the full gambit GN would have thrown at it - tested some marketed for PC use specifically thermal pads a year or two ago, commercially available product - the results were middling at best - on par with cheap paste - I remember looking at them & thinking like ya i've disassembled a laptop before & these have been everywhere since as long as i can remember but they don't get used on the CPU for.... have to imagine some good reason... Would love to see sheets of the materials you proposed cut down & similarity tested - that's a really neat idea
@@phishphan49 Of course pads wouldn't be used for CPUs or GPUs, but pads are an important topic when talking about thermal interface materials. Especially in computing when they're used on nearly every other heat-generating chip that needs to be cooled for performance purposes. They're also typically used for gap filling on multi-component heat sinks that're planar, where a machined heat sink would drive costs up by a fairly substantial amount, which is why heat conductivity via the Z axis is an important subtopic of thermal pads. Plus with how this channel manufacturers nearly everything they display, it'd be interesting to see housemade thermal pads, especially if they're far better than what's available on the consumer market. I also wouldn't trust LTT for anything science related. Just look at their thermal imaging video, and the entire lack of knowledge of the IR spectrum displayed by the entire team behind that video. Honestly, it's even worse than Austin Evans' recent PS5 cooler video where he just points a thermal imager at the exhaust of the older and newer PS5s to compare the 'performance' of the heatsink design. This is why everyone wants GN to test this channel's housemade thermal paste, because LTT just outright cannot be trusted with scientific evidence of anything, LTT is an entertainment tech channel, GN actually has testing methodology, testing equipment, and values hard science and fact over revenue. Even the research that goes into LTT's videos is rarely anything more than what's listed on a Wikipedia article, if even that much at times; more reason to not trust an _entertainment_ channel with actual science and fact.
@@jakesnyder2667 thanks - but yeah i get that - i saw this after making a separate comment asking them to reach out to GN & der8aur for this and a number of other reasons - this channel has the tools & knowledge to serve as an independent tarty steve could seen - like with the recent power supply fiasco - helps GN out by we sent our testing methodology and steps to other independent parties & they were able to reproduce out findings - independent confirmation helps - i mena steve was testing more extensivly something someone else had previously discovered there to try to see if he could reproduce it - good for the channels and good to the industry. as to the LTT stuff - yes, i agree with you, i guess i should have stated it differently in my first reply - LTT serves are a gateway into PC building, it's not a serious resource and yes it is entertainment - was simply stating that i thought I recalled them testing an item & pretty sure it was LTT because i don't recall it being very rigorous.
Send your compound to Steve at Gamers Nexus as well! His methodologies are generally much more stringent, no offense to the guys at LTT.
True, but still would love to see it there too. (unless they already did/do)
@@DoubsGaming they reviewed the original at LTT.
Edit: I would be much more excited to see Steve have a look at this as well.
@broton I second this
Omg yes
I don't trust Steve's bullshit opinion - he refuses stringently to add taste as a parameter of his test procedure.
This man is out here literally murdering entire product lines with the scientific method like its nothing. This is quickly becoming my favorite channel.
If I am building an expensive computer, I will stay with the manufacturer’s recommendation. Thermal paste is cheap insurance.
@@oldmech619 this response doesnt even make sense.
i know this guy knows so much everything from distilling to making homemade rockets. this is my favorite channel on yt
If only he could get a proper camera man.
@@ChunkyWaterisReal You obviously are not a professional builder of anything electronics in a commercial setting
Only variable this man didnt isolate were the 15+ beers he crushed while doing this experiment.
Been watching this channel for a number of years and still impressive as always. Hope you never stop making videos my friend!
There are very few presenters that capture full attention from first word to last. I had a physics instructor like that in 1962, and Michael Arbib had that effect on me five years later in a computability class. Only fifty years passed before you came along and reliably do it to me again. Great work.
I too appreciate the eloquent calculated speech, it is scientific in nature and delivery.
you're right .i didn't even notice it was 41min long..
You need a second chan with longer content in it and skipping the repeated tests over and over
I though that he has to endure this boring test so many times, I can stay with him the full length. From the sights after the experiment reset you can relate his boredom.
> Methodology
> Testing:
19:43 tooth paste
24:45 honey
28:10 arctic mx4
29:29 arctic silver 5
30:38 low cost DIY paste
32:22 noctua NH-1
33:25 Prolimatech PK3
34:24 Dowsil 56-22
35:20 Shin Etsu X23-7921
36:05 DIY high performance paste
Controlling for variables while still being practical for experiments/test setups is a rare art form that shows just how serious an applied scientist you're dealing with. This apparatus is a remarkable demonstration of that. I'm in awe.
Thermal Paste cost $6 delivered to your home via Amazon. Buy it. You know it is correct. No guessing. It will last many many years. Simple and Effective.
The problem is it's more difficult to perform these tests over a period of months with power cycles. That's where issues with the cheaper pastes show up.
@@oldmech619 This isn't to save money. This is purely applied science. It's fun, you learn something.
Can we make something better than the commercial guys, and can we make that at home?
@@crabmansteve6844 I totally agree with you. This is a demonstration is for applied science. It is not a recommendation to circumvent proper commercial and professional procedures.
I've listened carefully and I'm left with one question: how will the thermal compound perform after a year of use? I mean, it's great that your mixture works well in these tests but how will it hold up over time? Perhaps the next step could be some kind of endurance test involving a much higher temperature over a longer stretch of time. How does it look when you "bake" the compound, compared to commercial products?
I think he kind of adressed this by talking about the water content of the honey, as long as the compound doesn't lose its fixitive ( like with water evap ) its thermal properties should remain
@@crashoverride93637 Evaporation isn't the only concern. Does it slowly react with metals normally used in PC components? Does it oxidize? How quickly? How does temperature affect all of those factors?
An endurance test shows you what reality's doing to undermine you. Sometimes it takes a while for those effects to become apparent.
Since the motivation for this project was the apparent lack of public data on the topic, I'm guessing the reason why he's sharing the formula free of charge is so that the time and competence of other people will scrutinise their compound and find its actual properties.
@@Teth47 much like Tesla and fsd the best way to do massive scale testing is to have users do the test by putting the formula out there anyone can do the testing for massive scale
@@ParadigmUnkn0wn ...your kind of proving the efficacy of the method, also as an aside you might be holding fsd to a impossible standard it doesn't have to be perfect to be useful it just has to be better then humans, humans hit things with alarming frequency
This has gotta be the most controlled thermal paste experiment I've ever seen on TH-cam. Thank you for providing both entertainment and education, especially in such a broad range of areas. If I could only watch one TH-cam channel, it'd be this one.
Agreed. That one gay Canadian needs to delete all of his thermal paste vids
@@1MRSomeguyunless we're talking about personal lubricants I fail to see the relevance of sexual preferences.
After saying that I now have a very uncomfortable mental image of Linus using thermal paste as lube. Someone pass me the eye bleach, please.
The question is: Is it really necessary to control it to that point? Measuring temperature difference between the two sides after it has settled should be perfectly sufficient. As long as all thermal pastes are tested the same way (i.e. same surfaces, same pressure, same input (heat) power etc) it should be easy to find the best (and get a good indication by how much).
@asdrubale bisanzio Steve does pretty well controlled tests. Nothing on this level, but pretty good for determining real-world performance in a typical use case. Linus on the other hand veers more into the entertainment side of things. Still fun to watch, and not entirely useless info, but a lot less controlled.
@asdrubale bisanzio But is it actually important to keep the temperature of one side constant if you are only interested in the temperature difference between both sides? Actually it would be much more accurate and repeatable to measure the temperature difference only and let the experiment run until the temperature difference has settled.
Instead of introducing additional inaccuracies from the cooler, measuring the 12 minutes and so on.
"You can believe me that Crest is one of the best thermal interface toothpastes out there"
I love this channel
Okay, Colgate is of the worst thermal interfaces compounds, but, what about for polishing car lights? uh, uh?? (wink, wink).
well i wont trust tooth paste on my Ryzen 7 processor because after 15 mins that tooth paste dries up and over heating cpu's aren't cheap
Always start your test with crest..
...you know, to prep the surface for the other tests LOL
With so much fluff out on TH-cam, this channel is time well spent. Love your demeanor, explanations, and breadth of knowledge.
Oh, and a monologue of 15min as intro is rare on youtube these days. A lot of speakers are just cut after every sentence. It shows how well you know your subject and that you're getting good at this game ! Best channel ever.
Yea, I find it incredibly impressive that he can speak so eloquently and detailed for 15 straight minutes off the cuff. No teleprompter, notes, hesitation. You'd think this was his 100th take, but I wouldn't be surprised if it were his first.
I just wanted to say that you’re a personal hero of mine. I always dreamed about being an inventor and just communicating so clearly these real projects and how we know their better. You’re an amazing influence on the platform and a true scientist. Thank you for making such honest and clarifying videos.
Love the beer cans and bottles progressively appearing during these exciting tests !
Ain't no laws when you're drinking claws
No food or drink in the lab!!!!1 >:(
@@woosix7735 good thing it’s a shed not a lab and it’s owed by him so he can do what he wants.
@@carsonmcnamer5321 andddd it's the one kid who ruins the joke.
@@manavsharma2232 who?
I dont think I've noticed a single cut, you're so good at speaking and presenting your projects!
Honestly I don't know where this guy came from... But he came onto TH-cam and has top notch editing and presentation. You never feel dumb watching his videos, even on the most advanced topics. Nor do you feel like it's below you. Honestly this is one of my favorite channels.
@@girrrrrrr2 yeah, really good channel. they aren't selling themself out for outside forces. at least yet. fingers crossed. maybe they have enough capital to be independent. rare breed.
@@cubertmiso I honestly think that he, maybe his son is some form of engineer or something along those lines. And I honestly hope he keeps being independent or at least is very objective when sponsored. Such as this was given to us. Let's see how good it really is.
You didnt notice the sudden appearance of the beer bottle, then the whiteclaw can on the table. Then teh pile of empties on the stairs behind him?
@@liveepically well, I meant a cut mid-sentence, not just any
This right here is a very good demo of the scientific method. I can't imagine the time and effort given for just one video. Very High Quality Content
I like the progressively increasing amount of empty bottles/cans on the steps as the tests are waited for.
This was also my favourite easter egg. Gotta beat the monotony of testing somehow eh
Doin' science stuff, and killin' a 12 pack. Very efficient...
How well does white claw work as a thermal conductor? Moro brew beer…?? Lol
Noticed it too ... LOL
Just saw Linus recommend your paste on a silent passively cooled gaming rig. 2nd best loudspeakers and now best thermal compound.
Its liquid Metal not this what Linus used in the no moving parts PC
@@Comedy4noobs he did use LM for that video but linus did a shout-out referencing this video.
Which LTT video? can you give me the link with timestamp?
@@tafsirnahian669 Recent one with the passive cooling.
Link?
I really like this video and find it interesting as I'm something of a computer nerd myself. I only feel that there is one variable missing in this video that is considered a fairly important property of thermal paste which is longevity of effectiveness. Maybe an idea for a future video? Either way, your dedication to these projects and results that you get are pretty impressive. I'm glad I subscribed a few years ago. You guys have an amazing channel!
For Overclockers there is only the peak performance. so honey will do.
Correct, this is a time zero test. When validating thermal solution it is temperature cycled to accelerate life degradation. Another important aspect is measurement account part to part variability.
Thermal Paste cost $6 delivered to your home via Amazon. Buy it. You know it is correct. No guessing. It will last many many years. Simple and Effective.
Took the words right out of my head!
This is very important. I've seen compounds degrade over time and some of the cheap ones even leaking out from under the heatsink which led to thermal shutdowns. I generally stick with NT-H1 / H2 or ArticSilver MX. The latter will retain its properties for years and makes taking off a heat sink easy.
As an electrical engineer (technically EIT rather than PE), I greatly appreciate the rigor and thoroughness of these videos.
I've been subscribed for a while now. I love how these videos are extremely information dense, and lack the dramatization seen in other channels.
I tend not to watch your videos immediately when they appear, but I promise you I don't miss a single video.
Well done, and thank you and your crew very much for all the free* content... and the "open source" thermal compound. Cheers! 🍺🙂
*I have youtube red, so I guess it's technically not "free", but it's not paywalled, so it's close enough to deserve a hearty "thank you!"
This man teaches better than most Professor's. Love the videos, keep up the great work!
If Professors = Professor's
then videos = video's
Professors are employed for their research output and ability to attract funding - not their skills in teaching adults undergraduate courses. Generally professors make the worst teachers at any education level
It's because it's not his job... Every time I've had a hobby that I was great at (IT work, photography, RC, etc) that people started paying me to do, I quickly grew to hate the hobby. Maybe I am strange in that way, but I suspect this is a big reason a professor learns to stop caring. A bunch of 20-somethings that are aimlessly drifting through life would quickly wear on me as well.
@@jacobwcrosby Have you tried working for yourself where you have the freedom to accept or reject work/clients?
It's a little bit like musicians that have given up creative control (sometime copyright) to the recording and music industry.
This has happened to the biggest musical acts - even the Beatles.
I recall George Michael's dispute with Sony. The company would not release George Micheal from his contract so that he can collaborate with other artists, so Michael stood down from any music for a few years until his contractual obligations with Sony lapsed. He hated the music industry at the time (and probably until his death).
I tend to agree with you though - money and external non-creative demands often contaminate the passion and creativity that drives a hobby.
There are some artists and scientists etc that have managed to avoid these pitfalls.
I have been following your channel for many years now, and seen everything you have posted. Thanks for the diversity of knowledge shown. I just saw the LTT quick response video, and you compound just in my eyes crushed the tech of a multibilliondollar company (they said on par, but that is crushing, given the time they have spend against your inventiveness). You always were a "hidden" pearl, glad your channel is growing, and beeing recogniced more.
And here I was thinking as the cans/bottles piled up: "My man! Looks like it was a good night of testing!"
Yep. I noticed the bottles and cans too. :D Nice little joke there about the time you have at your hands when you wait for the test to run. :D
- for some reason the results got more and more inconsistant as the experiment wore on. 🤣
Remember kids, the only difference between science and screwing around is taking notes.
This level of test thoroughness rivals the testing of Gamers Nexus. I really think you should collaborate with them for a future project. I'm sure they can properly appreciate the amount of effort you're putting in.
Thanks!
@@TechIngredients someone also said you should send Gamers Nexus a sample to test as well, as they do a better job
@@Marksman123771 With all due respect to Gamers Nexus, they do not do a better job than Main Presenter here.
@@signalworks Gamer's Nexus might not, but Der8auer would probably be able to get his hands on the professional equipment to really test the TIM here. Since, iirc; he worked with one of the big TIM companies on the development of their extreme OC TIM used for some of the competition overclocking runs.
@@TechIngredients gamer's nexus and major hardware(fan showdown show)
I never heard of this channel before this video. You are so good I subscribed in the first 20 seconds. Quality shows easily from the jargon. Im so glad to have found you. Keep up this kind of thoroughness and rigour it's what the world needs more than ever before.
Thanks and welcome!😀
the number of white claw and beers increasing in the background is the best part. thanks for a great scientific video!
I absolutely love how methodical you guys are. It is soothing to my OCD self. I appreciate the attention to detail and the fact that you leave no i un-dotted and no t un-crossed. Thanks for your videos and your stringent methodology.
..This man is brilliant, and his assistant.. .I'm so jelly..
..i consider what they both do, LOGICAL..
..i would kill, to be there.. ..props to both of you..
It would have been nice to see a control, without any thermal interface material at all.
And one with some kind of thermal insulator (sort of a "worse than nothing" test), so we can get a better sense of the scale of the results.
I agree, but how accurate of a test would it be given the bit about surface imperfections and small contact surface area?
Yes! Negative controls are super informative and basically essential. Surprised it wasn’t shown. I bet they did this at some point.
@@JustinWillis_actual it would show how good the surfaces already where and if the thermal compounds where acting as insulators. Thinner the better the more perfect the two surfaces are. It's probably a huge variable in how you should choose your compound for the job.
definitely want to see a dry run!
”Crest is one of the best thermal interface toothpastes out there.” Lmao.
But I like Crest GEL....damn it all to pieces
9/10 dentists agree
Things you never knew you would know... Now if this ever comes in handy I'll be surprised
A true man of science tests everything
@@jamesfrantz9075 who knew honey was so great?
Please run 3 more tests for the benchmark:
- Clean surface (no paste)
- Drop of motor oil (10w-40)
- Drop of water
Would also like to see how the indium film does.
@@tegra5971 That would have to kick off a whole series on how to generate enough pressure...
Drop of Mercury?
@@Spiralem Likely too unsafe
@@Spiralem That'd be pointless, since mercury is actually a very poor thermal conductor.
Love everything you do. Get so excited every time I see you upload. As a chemist, a scientist, a dad, a tinkerer, and a teacher to my children (homeschool during covid) - you’re amazing. Keep it up!
I love the accumulating empty beverages.
who said science wasn't fun
Use the props in the background, papers and things sitting on the table and anything else you can get into the shots to tell a story. plan it all out, something like you were messing with time machines and accidentally fractured the time line or something starting with little things like a one second jump that just looks like a mistake in the editing, having things zap in and out of existence and ending with all out clones/doppelgangers walking around behind you. Get in touch with Captain Disillusion to do the special effects. It will add a science fiction film to your science reality stuff told through a game of I Spy! It might even get your channel featured by some other mystery solving channels!
39:27 he talks about this.
@@marianotombetta4149 I will MAKE it -legal- _fun._
A real *thirst* for knowledge.
This is legit my favorite channel on TH-cam -- one of the very few that i drop whatever i'm doing whenever i see a new video uploaded to watch.
keep up the good work and godspeed on the nuclear fusion stuff, i've been sad there hasn't been a video on that in so long!
You need to send some to Gamers Nexus for testing!
tech ingredients testing is far superior than anything gamer nexus can come up with.
Wow, that was the quickest 40mins of my day. Love the deep analysis, theory, and strict testing procedure.
I'd agree - I saw the 40+minute length, watched anyway, and was filled with disbelief that the video was actually over - surely it wasn't longer than 15 minutes! Nay, it was in fact 40...
Never, in my life, have I seen such a high level of explanation. I came here to learn a little about alcohol distillation and kept watching your videos all sorts of topics.
It's just amazing on any scale to learn from you about anything.
I have no words to describe my impression of everything I learn here from you. The pleasure is immense.
The quality of the videos is at the highest level. Perfect cuts. Perfect explanations. Perfect quality.
I have never been so impressed with the quality of study and investment made here. Just a very high level. 100%
Thank you very very very much for everything. appreciate it a lot. No words.
Please continue to teach from now until forever :)
Thanks and welcome!
"You can believe me that Crest is one of the best thermal interface toothpastes out there"
That had me laughing out loud. I can't help but wonder how many did you actually test for this video
Colgate is better.
He search for tooth paste with the highest water content
Does it matter if it's paste for under 6? It should be less abrasive too
@@garrysekelli6776 definitely Colgate 😂
No, is Crest. You can trust me.
I like all the new intros, nice talent going on. I noticed the mountain ale bottle appear. I burst out laughing when I saw all the cans start appearing on the steps. Great outro too, though the music was a bit loud. Thanks for all you do.
I am always amazed at the break down of the steps and how simple you make it out to follow but how in depth it is at the same time. Performing quantitative science with backyard materials on things that seem like they require TONS of money to get into.
Thanks.
They don't if you really understand the principles.
Each time it feels like I'm on the best lecture in my life
Science presented in a way that everyone can follow and with an excellent sense of humour (or humor). I love it 👍
You guys are freaking awesome. Not only do you put in massive effort doing some sweet experiments,but you make your research available to the community for free! This is what science is all about.
After the last few high octane rocket videos, Im super glad we get a little "Thermal interface material" chill. It's been wild ride.
"Chill" - LOL. I see what you did there! :D
Am curious about the other properties that one might find desirable in this application; corrosion resistance, durability, etc...
Am really enjoying your videos and the tedious ( in a positive sense ) nature of them, youtube algorithm be damned we need more content like this.
I think the tedium you are referring to is precise attention to detail.
Synonyms for "tedious" in this context: Rigorous, thorough, precise, meticulous, scrupulous...
I love how you pay attention to your words so you can make things as understandable as possible. I know the pain
You make some amazing content, thanks for sharing this with us.
I do have 2 suggestions, however.
I've mentioned it in 2 videos now I think, but please consider sending a sample or 2 to Gamer's Nexus in addition to LTT. I'd love to see more data from independent testers. And he's definitely one of the best I know of. He also has a much narrower focus compared to LTT, and a major part of that is thermal benchmarking.
My second suggestion is to present your data in a graph at the end.
Yes! The intro was so good that i didn't even notice it. Very on board with the importance of thermal interfaces.
Thank you for slugging through it, I could feel the "I've done this test so many times..."
I love this channel and recommend it any chance I get!
Also I knew the cans/bottles were in jest because I imagine you would be sipping your home-brew stuff!
I love seeing the bottles build up as an indicator of what test you're on XD
I'd really like to see you send this out to Gamers Nexus as well, those guys are much bigger into careful testing of PC components than the folks at Linus Tech.
I haven't seen a video from Tech Ingredients in a long time and i must say you have really improved your presentation skills significantly! Kudos
I'd have loved to see a summary of the results at the end, to make it a bit clearer how the different compounds compared.
I'm also curious how the temperature vs. time chart would have looked. Was it more or less linear?
And a suggested improvement to the setup: use an Arduino or ESP32 or something with some relays to automate the cooling/measuring cycle. That way you could leave the equipment unattended for a full test cycle.
AGreed there, I was just going to suggest a temp logger. But You can also chart the temperature inside the Arduino. (I wouldnt use ESP32, they don't tend to have very good built-in ADC's. And, if you use a digital temp sensor like DS18B20 , I have noticed that they tend to be +-3 F cyclically, which requires a ton of averaging. ) I found Arduino's ADC and a simple PTC thermistor to be the best of the bunch
@@gutrali I was thinking they could interface with their existing high-accuracy scientific equipment. It's pretty common for it to have some kind of digital interface. If not, I think the built-in ADCs on both boards are pretty garbage, better to grab an external ADC on a breakout board or something.
There's a link to the results in the description.
@@gutrali What is the effect of the ADC on the ESP32 not being good? Also, aren't they suitable for temp dataloggers?
@@Z3DZ3R0 they are suitable but it depends on your requirements. Check this out. th-cam.com/video/UAJMLTzrM9Q/w-d-xo.html
I had so many problems calibrating an RTC thermistor until I started to use an external ADC. I'm not sure but I think the Wifi onboard the Esp causes some voltage spikes that will affect your ADC readings. Plus it's 10 bit and only 3.3v
Cycles would also be interesting, how does the paste hold up?
And: If you'd increase the power it would help with the resolution. You have 0.1K differences and you can resolute 0.1K -> If you'd pump more power in you'd get more dT thus better "signal to noise".
Yeah, I missed the temp level but if he was pushing 100C it would be interesting, seems my old CPU that plays TH-cam videos at a temp of 34C Plays other games at up to about 72 which is a bit over the recommended value of 66. But if I push it harder and it gets into the '90s my computer says that enough of that silliness and shutdown :( Now I mean I'm glad it shutdown because otherwise, I would have already had to buy a new CPU, but this did make me wonder about changing my factory Heatsink for a replacement with a bigger heatsink, pumps and grills and all that seems like too much for me, but a big heat sink and a quiet fan would make me happy, my fan case has actually blown all its fans but the CPU and Video card fans are still working :)
@@basildaoust2821 If your CPU is getting so hot that the system shuts down it's likely that one of three things isn't working properly anymore - either the fan isn't spinning as fast as it should, the fins of the cooler are filled with dust and debris, or the thermal compound between the cooler and the CPU die has dried out (could be all of the above). Also, if the case fans are not spinning, you are not getting enough fresh cool air to help with cooling your CPU, video card and other components inside the case. If you don't feel like changing the case fans, I would recommend removing a side panel and keeping it off so the hot air inside the case at least has someplace to go. Then take your CPU fan off (remember the orientation) and take a vacuum cleaner or compressed air to the fins to remove the dust buildup, then screw it back on and see how it goes (you don't have to take the cooler off the CPU in order to do that). If this doesn't help with your CPU temperature, then it's probably time to change the thermal compound, as it might be dried and cracked and effectively insulating the CPU instead of helping dissipate the heat. This is all assuming that your cooler isn't seriously underspecced for your CPU, in which case you should look for a better alternative, however, you might as well try the inexpensive solutions first.
@@benedictul Yeah I don't think the issue is any of that, it's just if I push it too hard it gets mad. I mean my case fans have died and I have a house fan blowing air through the case with the side panel off. The CPU fan and the video card fans are working just fine. I have a new CPU that is much better, not good compared to the new ones but more than twice as good as my old junk. So yeah the plan is to first run the one or two games that complain at lower graphic settings which does work quite well while I buy a few pieces to put the new system back together. The change does also let me switch to an m.2 drive so I expect some nice improvements. Thanks for the suggestions.
I like the build up of cans on the stairs as the test go on haha
Edit: I just saw you explain it lol
I think "Breaking taps" and his AFM (atomic Force Microscope) and " Applied science" with his SEM (Secondary Electron Microscope) would be good collaborations for looking at surfaces and interfaces plus these guys would be interested in the compounds themselves....great work as always.......cheers,
scanning* electron microscope, but yeah, I'm sure Ben would be a good third party to verify these test results. He might actually be able to produce a standard measurement as well as relative results.
@@DFPercush "Secondary" yeah I had mass on my brain....cheers.
Alpha Phoenix would be great too.
I just discovered Breaking Taps a few days ago, and WOW. That channel is going to explode. Also, Huygens Optics is good.
Thank you so much for your dedication to the testing procedure, I would hazard a guess and say it produced much more reliable information than you can get off the box on a commercial product. And it's good to hear your humour survived intact as well! I'm really looking forward to the next video, take care and thank you again.
This is the best quality channel on TH-cam. You have excellent experiments with great explanations and background. I think it would be worth testing without any thermal paste, your surfaces are so smooth it might be better?
Thanks!
46 C
"you could read an audiobook..."
oh no here comes the ad
...nevermind
It actually got my hopes up that they had gotten a solid sponsorship. Was a bit sad they didn't.
as someone who wants to see the channel grow, where's the audio book sponsorship my guy? It's free real estate!
I was totally expecting segways to Audible and Honey.
@@P8qzxnxfP85xZ2H3wDRV it's viscosity versus residual effect, so for long term performance honey may Win out as long as the pressure holds out...
Yeah also the most sincere press subscribe bit I've ever seen on TH-cam
Loved the increasing amount of cans and bottles. This is one of the best channels on youtube so many reasons, but first and foremost it's actual science. I only wish you'd explained that it takes exponentially higher thermal conductivity to get closer and closer to the loop temp. I'd also love to see a no-paste run. Off to buy some paste for use on my cpu to test its endurance.
Must have been a good day, at least 2 people drinking and doing science [and who knows what else...].
Man I wish you were my science teacher back in the day. I feel like I would have learned so much more and more importantly paid more attention. Excellent job!
My girlfriend and I were laughing at the staircase gag.
By the end there are SO MANY haha
Great video, keep up the good work!
Exactly! :-)
So much attention to detail, the gags, and mojoesque sense about him had me subscribed a long time ago👍
Sponsored by Mountain Ale and White Claw :)
@@pavelperina7629Don't forget the Harp's Irish Lager! :)
I wish I had a science or engineering teacher like you when I was in college, I guarantee I wouldn't have fallen asleep in class. Great content! Suggestion: for all the effort you put into the thermal test, you could have used a thermometer with an extra digit of accuracy so there would be more than 1 count difference between all your samples.
You probably would. Thats the power of editing video. But the alcohol they didn’t drink, but you can, would make it more endurable ;-)
an extra digit of accuracy would be telling as of the comparison between the different pastes, but would still be practically not-interesting... 0,01°C-difference isn´t gonna make your overclocked CPU feel any difference in longevity or performance. So, althought u r right, this change would only raise the "thrill-factor" and production-value of this test itself (raised production-value pays for enhanced thrill-factor), without any other practical significance.
@@klausbrinck2137 The thing is its not .01C difference really. The units arent really useful in this situation its a qualitative measurements that don't apply to cooling a cpu.Its just showing which is better than the next one. It could end up being 10C or not noticible. Like he said the actual temps are only applicable to his testing setup but will show a difference as to which performs better than the other
Amazing video, would have been interesting to test the thermal conductivity without any paste, just the bare metals
Love the beers stacking up ^^
I like this series and hope you continue expanding on it. One thing you need to test/state is the electrical conductivity of your epoxy and paste. All of the commercial products you tested are considered electrically non-conductive. Also a control test with no TIM would be helpful and a high end test with liquid metal would be another great data point to add in.
One more vote for a control test with no TIM
every single one of ur videos is a treat. Very excited for the LTT collab coming up.
I have to say, I have been casually watching your videos for years out of a shared interest in many of the technologies and processes you are knowledgeable about, but this particular project seems likely to push you to the million subscriber mark and above. I really appreciate the sentiment, intentional or not, that a small startup or even someone in their basement can create something that is better than companies with millions of dollars of R&D budget, with just a fresh set of eyes and solid knowledge of basic principles. It's ... motivating!
Thanks.
Motivation is what we're all about. That and a good laugh now and then!😆
TE, I'm glad that you took the time for this video. I know it took a long time, and you very much displayed the tediousness of the test - I mean, the video itself was 41:51 - but I wanted to say that I greatly appreciated you going through it all, and as I also follow LTT, I'm hopeful that we'll see their review soon. Best wishes on your new product!
I'd love to see comparison to more thermal compounds. I switched to Grizzly Kryonaut not long ago and this stuff made noticeable temp difference compared to other brands I used before.
This is masterfully presented, thank you. Also, You explained the TH-cam algorithm better than anyone I've heard.
I just love your videos and am always surprised when I watch all the way through every time. Your content is just top notch.
Great video and work as always. I feel like you've missed some important properties that should be measured:
1) How much does it 'dry out' / oxidise over time (sometimes can be simulated with elevated temperatures)
2) Some compounds need a couple days to reach peak performance, including AS5, how bout testing after a week?
3) Feel like you need to have sandwiches that hold the different compounds over time rather than always scrubbing them. The interfaces to the sandwich could be designed around them or just all ultra smoothed lapped so it's not performance limited there.
4) I haven't used AS5 as my premium compound in a decade, I use IC Diamond (tiny diamonds) for some installs and Thermal Grizzly's Conductonaut (liquid metal) if it doesn't eat into the heatsink. You're not exactly comparing against cutting edge?
5) The performance of a lot of these pastes will be somewhat muted in practice because many Intel chips have poor thermally conducting heatspreaders with middling thermal paste over their silicon. Would be interesting to do a die delidding or lapping in combination and see how that goes.
6) For ultimate performance, I'd have a delidded CPU with a Graphene horizontal thermal transfer sheet (can't transfer vertically but does great horizontally), then have lots of holes with a silver sheet overlaid including silver going into the holes to provide vertical transfer, and so that the heat distributes at a fantastic rate over let's say 6cm X 6cm, and the silicon die is already incredibly smooth (but not always flat, can be lapped), then the Graphene sheet has enough compliance not to need thermal paste (still good to test), then a heatsink bonded to the silver layer with a liquid metal thermal interface.
If you do that, now that would be world record impressive and a heck of a video!
As a filler material, graphene is not good as I'll explain in the next video. If price is no object (and it always is) then we have some disk heat sinks with grown vertical arrays of carbon nanotubes that rival bulk diamond for heat transfer.
Could you imagine this guy being your Dad? Super smart. Always working on cool projects. It would be pretty awesome.
One concern I have is how long this will last on a CPU or GPU. I use Thermal Grizzly Kryonaut which is almost the same as PK-3 if not the tiniest bit better and it's guaranteed to last years. How long do you expect this to last?
True PCMR replaces Thermal paste each times he boots up the PC
@V0LDY So ... once every two years ?
This immediately entered my mind as well. How chemically stable is it, and if not how long before it begins to degrade as it undergoes thermal stress?
How long before I'll need to remove and reapply it, and could it potentially begin to set and harden over time and require specific chemicals in order to remove it? If it does degrade over time, could it potentially harm the CPU lid?
@@HallsteinI guess we will know in the next video where he teaches how to make it
Most thermal pastes use vegetable oil (ie the stuff you use to cook your eggs) to keep from drying out or polyethylene. You can literally take a couple drops of glycerin (vape juice) and mix it with toothpaste and you have the standard thermal paste that dell etc uses. He has obviously done the research and knows what is needed or he wouldn't sell it.
Exciting work! Considering your extensive work on thermal conductivity, cryogenics, and magnetism, I’m surprised I’ve never seen a video on superconductors. Do you have any upcoming projects with superconductors?
Thanks
A superconductor video would be dope
"Hello, this is Tech Ingredients and today we'll show you how to make the worlds first room temperature superconductor from common hosehold materials and some stuff you can order from ebay for a few bucks"
@@GerManBearPig You can buy superconductor pucks on ebay, not sure how you would reprocess them to make wires, but could be done probably.
@@BlackHeartScyther Maybe, but those need to be cooled down with liquid Helium or Nitrogen. That's why making a "room temperature" superconductor would probably win you the nobel prize immediately (if it doesn't need a few thousand bars of pressure to be stable)
And afaik the'yre all ceramics, so processing a brick of it isn't really an option.
This is easily one of the best channels on TH-cam. Keep up the good work! You're providing an essential public service!!!
Just for reference, what temperature do you end up with if you have no thermal paste at all? Just the 2 smoothed surfaces against each other.
I was wondering this too. Would be nice to have the control temp to see how it compared.
Modern CPUs can even runnwithout any heatsink on them (throtling down so far that they're basically unusable, but still running), so I'd say the temps aren't an issiue, but the clocks the CPU will run at.
@@Donnerwamp True, but the default surface on CPU's and heatsinks are nowhere near as flat as his test rig. A few of the 'lapped' setups on TH-cam would probably be pretty close, but true lapping with wavelength precision like he was talking about is nearly never done on CPU or heatsinks at least on TH-cam. (Literally only 1 video with no numbers posted) His test rig is not true lapped, but that's fine.
For reference, I agree this would be interesting. Also, just for reference, why not use a drop of water.
*controls for every variable imaginable*
"now I can't attest to the exact accuracy of this piece of equipment or how reliable it is, but I've got reasons to believe it's pretty good."
I got a good belly laugh out of that.
I get that well-executed scientific methodologies are rare on youtube but I feel like he was actually making a really good point rather than an unknowing meme. It's a testament to his scientific rigor that he understands their test setup's merit is only as good as its ability to stay consistent with itself and when compared to other's test results rather than just assuming that he really did account for every variable imaginable or that the experiment happens to be designed soundly just because it seems that way.
@@Chris-io2cs of course - I just found the intonation on 'pretty good' hilarious.
Just finished the video with the thought that you were Bob Ross doing science.
Thank you for your amazing passion
I really wonder how this compares to "liquid metal" thermal interface material(s), wonder if you could have a go at testing that out in this contraption as well.
Yeah, both the gallium-based ones, and the "liquid copper" ones would be nice to compare to this.
Well for one it would require nickel-plated copper test blocks, gallium destroys aluminium and modifies copper.
Those outperform normal pastes by a lot. And so they'd outperform this. They have like 6 to 30 times higher thermal conductivity, depending on what paste you compare them to.
Also, they didn't make the best thermal paste until they beat Thermal Grizzly Kryonaut, the best paste I can think of.
@@brei2670 Liquid metal is great for thermal transfer, but it has unique downsides that put it into a category of its own
Doesn't gallium react with aluminum . His test rod device is aluminum , if I heard right . Doesn't liquid metal have gallium in it ?
Watchable and excellent as always. Would like to have seen a control with no interface material at all. Also a real metal interface such as a solder-joint or mercury pool, even if those are not usable in most real-world applications.
Distilation was what caught my eye on this channel since im brewing and distilling myself at a hobby level, but the other stuff you are filming and the accurate tests you are carrying out is quite interesting and satisfying to watch. Keep it up and greetings from eastern europe!
That was really "cool"
Phenomenal work as always.
As a
One more factor to ponder about: perhaps the optimal thermal paste is not the same for a pair of polished flat surfaces, as for two much lower grade (rough, non-flat) pair of surfaces.
Thermal paste composition is usually a tradeoff between thermal conductivity, viscosity, etc. It is plausible that the optimal viscosity is different.
The same thing that was going through my mind and why a control run with no compound would have been helpful.
How reliably could you clean a rough surface between tests of different materials? Polished surfaces at least give you a fighting chance at removing the previous material.
This is a test of thermal pastes and not of surfaces. That would be a separate video all together.
He was careful to ensure the paste flowed out around the test cylinder under the spring load, displacing the air, which would minimize the effects of the different viscosities of the pastes one would think (I’m not an experimenter, so I may be overlooking something)
@@gordonborsboom7460 Hmm, cleaning would indeed be a challenge. My point is that in pactice, the surface of an IHS is neither flat nor mirror smooth, and depending on the cooler it might be even worse.
I was having a beer while watching this... then started to see the beer bottles and white claws piling up in your video shots... so I was happy to "have a beer" with you guys... only to find out at the end that no beers were had. Love all you do, keep up the good work!
I LOVE how as the testing goes on more and more empty beer bottles show up in the background 🤣
These videos have been gradually but consistently improving and while continuing to stay focused on hard science. I cannot wait to see what innovations come out of this team
I'm subscribed since the first video i watched, love em and have binged the channel, as a TH-cam premium guy, i love that there aren't any huge sponsor ads that repeat on every other channel, the reson for getting premium in the first place, based in germany buying stuff is cumbersome, so i support with watch time
Thanks!
it would be awesome if you could test liquid metal like Thermal Grizzly's Conductonaut as well
and maybe some graphite thermal pads
@@danielpope3579 or do a thin nickel coating as it should not interfere to much with the result and block the gallium from reacting with the aluminum. but you're correct, for precise result they would have to retest everything with a nickel coated material, as blank copper will also be attacked from gallium
Hey i was waiting for this episode, this topic is among my interests! However it took a few days for the video to pop up in my feed and it missed the notification. Will be very curious about the recipe next.
It has been indeed noticed that for example Kerafol's paste Keratherm KP92 has 10W/mK but performs worse than MX-2 which is specified at 5.6, in fact it performs worse than just about anything north of 1W/mK. It would be weird if Kerafol wasn't being truthful, they are an industrial supplier rather than a customer facing brand. Different measurement method is indeed a possibility, but another is that thermal conductivity just isn't all there is to thermal paste's performance, as thermal conductance isn't really its primary purpose, it is to provide a high surface to surface thermal transfer while not actually bridging a substantial distance. There could be several reasons for surface to surface thermal impedance, like microscopic coverage, or just something about carrier material properties. Also the layer thickness that the paste ends up spreading out to under a given pressure for sure has to affect performance.
Curiously, AS5 was estimated by NREL to only have less than 1W/mK thermal conductivity, 9 times less than specified; as opposed to Shin-Etsu X-23-7762-S and Dow Corning TC-5022, where the test result was similar to manufacturer's spec and several times better than AS5.
Also regarding MX-2, it's Arctic Cooling's lowest cost offering, with the mainstream one being MX-4. Both have insane long term stability. It has long been known that in CPU tests, MX-4 performs better, this has been a result of pretty much every test to date regardless of publication. However Igor's Lab (former Tom's Hardware Germany, now independent publication) has noticed that when tested in direct die applications, such as on a GPU, MX-2 performs better! One fairly obvious difference is that MX-2 is harder to spread and potentially requires more contact pressure to spread thinly. MX-4 has lower density, lower viscosity, and higher nominal conductivity. It has presumably less solids content by volume, but the solids that are in there might be more thermally conductive.
Up until recently, MX-4 was Arctic Cooling's high performance paste; just months ago, it was superseded by MX-5. Arctic Silver 5 is not the higher end offering, it's a competitor, and a very old one. These are two different companies. Arctic Cooling is from Switzerland and Arctic Silver is from the USA. Arctic Silver's lower cost product is Céramique 2, which doesn't contain silver.
So your results have been that MX-4 performs worse than AS5. That's a discrepancy with tests performed by the PC enthusiast press, such as Planet3DNow (defunct), Guru3D and Tom's Hardware & Igor's Lab, that usually in their tests, MX-2 comes out equal to Arctic Silver's AS5 in performance, which was quite a shakeup when MX-2 first came out, because it just feels nicer, is safer not having any conductive material in it, and is a whole lot cheaper too. There was MX-3, but it didn't make a strong case for itself and was discontinued. MX-4 is a substantial improvement upon the former ones in tests. Ceramique never performed particularly well. There is a small number of people who still swear by AS5, so it survives on the market, but it fell out of most tests by now.
I wonder if water based materials would perform better if you lower the surface tension. So add dish soap to toothpaste. This is not a practical idea, after all toothpaste is not a material that will last long term and soap isn't going to change much anything about that, but it would potentially allow to test your hypothesis that better surface contact is the reason honey performs better. It won't be a conclusive test, but it can be a data point and more tests can be devised later.
I suspect that given you have practically mirror polished and matching contact surfaces, the solids contents of toothpaste could be doing it a disservice and increasing the actual layer thickness, since you don't need the material to actually be thermally conductive, you only the material to have a low surface contact impedance to the both surfaces, and low layer thickness. Real CPU heatspreaders and heatsinks aren't actually like that, it's not uncommon for the integrated heatspreader of the CPU to be a little domed in (concave) due to the manufacturing process, and many heatsink manufacturers deliberately make their heatsinks a tiny little convex, which is most noticeable if you look at ones made by Scythe, reasoning that increased centre contact pressure where the CPU die lives in the heatspreader package is going to make for a better heat transfer than corner contact, and also it can help the excess paste squeeze out.
Additionally, different geometry and surface quality of your test rig can explain some results discrepancies vs. computer press.
You have not specified whether Noctua paste is NT-H1 or NT-H2 and which Shin-Etsu paste. Also you made a typo in Noctua's name in the spreadsheet. :D
I recall there are a lot of pastes with known practical issues - Gelid GC Extreme, IC Diamond, etc, which showed spectacular performance but were known (at least initially, they could have gotten better since) to not last very long, rapidly drying out. This occurs because the heat-up and cooldown incurs some material movement, and the carrier liquid and the solids can then separate, and the liquid is then free to leave the thermal interface area. Do you think you can devise a test, which can demonstrate whether the paste stays in suspension or will separate over time? With these pastes, separation took months or years, but obviously you don't really want to repaste your computer every 6 months, that's effort; do you think an accelerated test is possible which can demonstrate this issue within a week or two?
I will never get tired watching smart and interesting people putting science and ingenuity into action. Thanks for existing and sharing with us.
KEEP UP THE GREAT JOB GUYS!!!
Just a sugestion: Is it possible to log the temp curve to see if there is any improvement in temperature oscilation and adaptation from the interface solution (AKA termal paste) ? simple arduino log is an easy go to and an excell sheet can draw a curve really easily... JMHO
I was also thinking the "curve" of the logged data. Once the general curve form is established, more performance information could be revealed e.g. how fast it reached equilibrium, without having to spend the whole 90 minutes and by just logging a few more "points" than only the final (12min, temp) one.
BTW, Awesome video, as always!!!
Found the engineer. ♥❤♥❤♥
I agree, BTW! Plotting temperature against time might actually reveal something interesting.
I mean... it probably won't give any additional insight... but it *_MIGHT!_*
It would have been awesome if (for fun & comparison) you also included a test of liquid metal, like Thermal Grizzly's Conductonaut. I use that on any interface I can because I get such low temps from it. Awesome video, and love your attention to detail and replicability!
Thanks so much for going through all the boredom that must have been involved in the repetitions!!!
Your production quality has matured nicely over the years. I know you know you are good, but I still think you don't know how good you both really are.
Always a pleasure.
I'm surprised you didn't test the indium sheets. I'd love to seem some numbers since you keep recommending it. That and I'd like to see galistan, or conductonaut, tested.
And graphene and other "exotic" materials
I just want a control test with air on those surfaces
I think he stated the indium sheets required a large amount of pressure to work effectively. I expect the springs in his testing apparatus don't apply 1/4 ton of clamping pressure which he stated is the minimum for indium. The liquid metals also have a secondary effect where they can permanently change aluminum (they are intended for use with copper). Using a liquid metal that flowed into the aluminum structure of the test rig would change/invalidate subsequent tests performed before since a surface wipe won't remove the liquid metal.
Just the sheer amount of educational knowledge on topics is enough to get me hooked. Add in actual testing and tangible lab results and oh my... Now add in the best product that I can purchase?? Take my money
I would've liked to see higher temps where there would be a more decisive gap between the results for each material
His temperatures are all the same because the two surfaces are mirror flat with no gap. If he used no compound at all I bet it would still read 30 C. The setup has nothing to do with the real world.
@@rrrlasse2 then why did it show the differences you would logically expect?
@@liambergstrom8183 Tomshardware shows toothpaste at 42 C and most top compounds at 30 C, which is to expect because toothpaste has very bad conductivity. His setup is bad and is in reality more like a bad viscosity tester.
@@rrrlasse2 so it's in the same order...but higher temp would be nice so there's a bigger difference? Like I said? If his setup is ranking them the same order...then it clearly works. The actual temperature has no bearing between one setup to another, it only applies for reference to other compounds on the same setup. The overlap between the two different setups means nothing
@@liambergstrom8183 It's not even really the same order. Most PC publication tests show in CPU use the MX-2 to be equal to AS5, while MX-4 performing substantially better.
Curiously Igor's Lab has found that MX2, while having nominally allegedly a lower thermal conductivity, performs better on a GPU in direct-die application than MX4. I have no clue what this means. Higher mounting pressure? Surface evenness for the GPU die vs. the usually warped CPU heatspreader? A lot of potential variables. But it stands to reason that surface contact on the microscopic level and layer thickness are more deciding than straight up conductivity.
Indeed mirror like and exceptionally flat surfaces are not what we're dealing with, with PC components, we've got concave heatspreaders and occasionally lumpy direct contact heatpipes to contend with.
Which Shin-Etsu paste was used here, was not specified, they have many. Nor the Noctua, they offer two, the H1 and H2.
About Tom's Hardware and their toothpaste test. It's a test performed by their former German subsidiary, currently known as Igor's Lab. It may have been lost in translation, but the way Igor tests toothpaste, and he did it several times after as well, he feels it's unfair to test it immediately after installation, as thermal paste performance stabilises with burn-in over the first 20 or so hours, which is something he can work around by running several tests and discarding the worst results, so this is the expected performance long-term, but for water based products, the opposite happens. When just applied, they perform about as well as a standard thermal compound, but afterwards they degrade rapidly. So with toothpaste tests, he instead burns it in by operating the rig at specified power for 12 hours, and measures how they perform after the burn-in! This is a possible explanation for the disparity in the results, why in Tom's Hardware test the toothpaste looks a lot worse than standard thermal compounds, while here it looks similar enough. A lot of people have independently measured toothpaste against thermal compound, and just upon installation, it often delivers very similar performance.
There are two other funny and extremely cheap materials i'm aware of that can serve as an impromptu thermal compound replacement. One is copper brake grease, it has been discovered and tried by various forum members and seems to perform about on par with thermal paste, it's ultra fine copper particulate suspended in grease. Another has been discovered by Igor's Lab, and it's endurance gearbox grease, Liqui Moly LM47 with MoS2. Both presumably use lithium soap stabilised silicone or mineral oil as a liquid base. Igor's Lab even used his finding in one of his lower-power systems long-term to see its endurance, and it has not degraded in performance over a better part of the year.
I wonder if the honey performed so well, because some of the heat energy was "used" changing the viscosity. I know it's not a full phase change but It did make me wonder if a substance that changes viscosity with temperature actually "uses" energy to do so. I had a brief (a couple of minutes) poke around on google to see if viscosity changes use energy or not but found nothing useful.
Viscosity is a property due to the intermolecular forces, as is the specific heat capacity of a material. For nonmetals, as temperature increases, some of that energy is stored in the intermolecular bonds which will also affect the viscosity. I would expect the specific heat capacity change over the temperature range in this video to be relatively small and insignificant. While this test did not reach steady state, it is not a short duration test, so changes in heat capacity should make little difference. The amount of thermal interface material vs the power going into the system is small.
Another effect might be the viscosity directly affecting the film thickness for a given pressure, but I can't convince myself either way of its effects.
honey is thixotropic substance and yes they do take in energy when they change their viscosity state.
What a class act this gentleman is. We would love to see the results of your opensourced TIM. We all can prepare and test it in different heatsinks, CPUs, GPUs, APU's. Hats off Sir!
I'd really like to see liquid metal and other gallium alloys tested in this setup, but the gallium would destroy the aluminum.
I'm actually a bit surprised by the materials used here, instead of just using copper. Because if there is another video in the thermal interface series that covers thermal films like liquid metal, they're going to have to revalidate all of their testing on the setup that can actually test the liquid metal. The tests are idiosyncratic to the specific machine after all, which would make any attempt of cross-validation of a singular test between both machines (say, toothpaste to toothpaste, as an example) would be analogous at best, meaning all pastes should be revalidated for accurate science. With how long these tests are, that'll be taking a long time to do, just to compare products of one thermal interface type to products of another thermal interface type.
Though, I'd be more interested in thermal pads, not the 'reusable' thin graphite ones like what Innovation Cooling makes. Liquid metal is analogous to thermal pastes and thermal greases, it's a thin liquid film that aims to serve the same purpose, though I would argue that thin metal films are a different thermal interface entirely, just as thermal pastes are not thermal greases, and just like how thermal adhesives had their own video in this thermal interface series. Thermal pads on the other hand do perform differently by which direction you're measuring, measuring along the Z axis (as Z-up) is a different measurement than what you can measure planarly along the pad's surface. This difference in measurements is how some products have ridiculous numbers as marketing, but their actual performance is lackluster in comparison. While the graphite pads are technically pads, and function in basically the same way, I would rather see them separated into their own sub-category, as they're thin enough to be marketed as a replacement for thermal pastes, films, and greases; similarly, the same can be said about films, indium films and gallium-based films are similar but how they achieve an interface is different.
@@xaytana iirc i think LTT? prety sure it was LTT - it wasn't the full gambit GN would have thrown at it - tested some marketed for PC use specifically thermal pads a year or two ago, commercially available product - the results were middling at best - on par with cheap paste - I remember looking at them & thinking like ya i've disassembled a laptop before & these have been everywhere since as long as i can remember but they don't get used on the CPU for.... have to imagine some good reason...
Would love to see sheets of the materials you proposed cut down & similarity tested - that's a really neat idea
@@phishphan49 Of course pads wouldn't be used for CPUs or GPUs, but pads are an important topic when talking about thermal interface materials. Especially in computing when they're used on nearly every other heat-generating chip that needs to be cooled for performance purposes. They're also typically used for gap filling on multi-component heat sinks that're planar, where a machined heat sink would drive costs up by a fairly substantial amount, which is why heat conductivity via the Z axis is an important subtopic of thermal pads. Plus with how this channel manufacturers nearly everything they display, it'd be interesting to see housemade thermal pads, especially if they're far better than what's available on the consumer market.
I also wouldn't trust LTT for anything science related. Just look at their thermal imaging video, and the entire lack of knowledge of the IR spectrum displayed by the entire team behind that video. Honestly, it's even worse than Austin Evans' recent PS5 cooler video where he just points a thermal imager at the exhaust of the older and newer PS5s to compare the 'performance' of the heatsink design. This is why everyone wants GN to test this channel's housemade thermal paste, because LTT just outright cannot be trusted with scientific evidence of anything, LTT is an entertainment tech channel, GN actually has testing methodology, testing equipment, and values hard science and fact over revenue. Even the research that goes into LTT's videos is rarely anything more than what's listed on a Wikipedia article, if even that much at times; more reason to not trust an _entertainment_ channel with actual science and fact.
@@jakesnyder2667 thanks - but yeah i get that - i saw this after making a separate comment asking them to reach out to GN & der8aur for this and a number of other reasons - this channel has the tools & knowledge to serve as an independent tarty steve could seen - like with the recent power supply fiasco - helps GN out by we sent our testing methodology and steps to other independent parties & they were able to reproduce out findings - independent confirmation helps - i mena steve was testing more extensivly something someone else had previously discovered there to try to see if he could reproduce it - good for the channels and good to the industry.
as to the LTT stuff - yes, i agree with you, i guess i should have stated it differently in my first reply - LTT serves are a gateway into PC building, it's not a serious resource and yes it is entertainment - was simply stating that i thought I recalled them testing an item & pretty sure it was LTT because i don't recall it being very rigorous.
It makes me so sad it wasn't made of copper for liquid metal testing.