(And because we didn't expand on it here: Under identical test conditions or for base material properties, thermal conductivity is of course a real thing and is useful -- say comparing aluminum to copper -- but it's when you start making pastes and tuning the properties just for a marketing number that Der8auer needs an anger translator!) Check out our previous video with Der8auer speaking on this topic here: ua-cam.com/video/CCqxE-5Ct3w/v-deo.html And check out Der8auer on UA-cam (auf Deutsch): ua-cam.com/users/der8auer Und auf Englisch: www.youtube.com/@der8auer-en
Well its not something JDEC can step in and solve We just need to hope that manufacturers start to do the test at 95c and base the specs sheet off of that Because we have moved from CPU’s ramping up over time to them ramping down instead Hence pegging at 95c and so the paste needs to be spec’ed for 95c
No mentions of thermal resistance. Still remember - Kryo33 made a point (they probably have a lab you talked about) that paste does not contain silicone, or any additives, and viscosity comes from finesse of the material, and, I speculate, that smaller the grains > closer to each other > better thermal transfer, because role of paste is not to conduct, but to transfer the heat. Often parameters come in two values conductivity and resistance - such as conductivity 13W/mK, thermal resistance
Had my sister use household lubrication mineral oil as "paste" on a CPU for 3+ years, because she is living in annother country and was unable to get her hands on real paste. So, I talked her trough all steps of building a PC and she did not have thermal paste. It was a custom "as cheap as we can make it, but new parts" office PC and I could not visit her. It is still running, albeit she did buy some paste after a few years and applied it. I am a big fan of liquid metal. Used it on my AMD K6-2 the first time. Loved it. But yeah, there is a lot of fancy talk about heat paste, which a lot of people do not need. Even for overclocking gamers it is better to get a good paste and refresh it after a while, instead of getting the "best of the best" and never touch it again. I miss the nerdy OC forums, which tested all the waterblocks, heat pastes and cascading compressor cooling solutions, with zero bias.
@@Niko-qk3su smaller particle size can only do so much as its fully bottlenecked by the carrier material and how many of those particles even actually touch each other Its why copper shim modding Nvidia 3000 series is as good as it is
Roman: "If I turn this into dust, and add a little bit of silicon oil..." Steve: "What're you, a f**in' alchemist?" You guys have the best conversations, always such a great blend of tech information and relaxed humor. Thanks GN & der8auer!
Jesus loves you so much and He is waiting for you with open arms, please accept Him in your heart and life while He is near Jesus loves you so much and He is waiting for you with open arms, please accept Him in your heart and life while He is near
Der8auer's channel has some fantastic content, especially for higher-level enthusiasts. He's also an actual engineer, so some of his commentary can get pretty involved and nuanced (as you see here) so prepare to have a second window open for WIkipedia~
I love how he talks a lot about engineering stuff in some of his videos, it really gives an unique touch to them. One of the tools he used to test a coldplate costs 7000 euros, lol, true engineer stuff.
@@TheAechBomb kryonaut was designed for sub zero cooling, and it excels at that, but people usually don't daily drive sub zero, and instead frequently change the heat sink and repaste, meaning that they didn't need to design kryonaut for longevity. I don't know exactly how long it remains in an uncracked state, but it's very short compared to pastes that are designed for long term use, such as hydronaut.
Wow that ended so fast, I feel like I need the uncut 50 minutes. Good to know that thermal conductivity numbers are about as useful as TDP numbers. Just make everything up to get numbers you like!
Thermal conductivity has a very common standard using astm D5470 which is similar to the use case. The supplier should note the standard used. Laser flash and other metroligies are basically random number generators. Also, the number that matters is thermal impedance. Thermal conductivity is better number for heat spreaders while impedence tells you how good of a tim the material is.
These things are true but lab results communicated ineffectively to end users who are not sophisticated and only want to know "does it work?" "How long?" "Is it easy to use?" and of course "price?" just do not help overmuch. Standards can be misused as well as correctly used. Sometimes the guy who wants to build a computer just wants IT TO WORK OKAY. That I think is probably the largest segment of the market.
@@CutaneousSensationsI would argue that the same points that matter to consumers matter even more to OEMs and ODMs, longevity, ease of application, dangers like with Liquid Metal and how it’s conductive electrically, and price. For average Joe those things matter but less, I’d spend 30$ more to get 2deg C and have to spend an extra 10min applying it, for an OEM that just won’t do when they are looking at fractions of Pennies in their build. If I was an OEM I would look #1 at longevity, needs to be at least 5 years, next pick the top 5 lowest priced pastes that fit that criteria and trial a couple units of each in my existing assembly process and evaluate the results. Then I would go with the cheapest finest paste that performs well enough to stay within the cpu manufacturer specs for temperature.
@@radugrigoras all suppliers worth their salt do provide basics understanding on how their materials survive for 1008 h bake at 125c, temperature cycle B 1000 cycles, and uhast 85/85 1008h. Sadly this will not translate directly and the OEM/ODMs will have to verify it in their test vehicles and products since the use conditions will be unque. So basically they already do that. Only small material suppliers don't provide this data.
@@radugrigoras also, you are crazy to think they don't try to find the cheapest paste that will meet their minimum requirements. That's their goal. They work with the material suppliers to develop this.
Met Der8auer few years ago at Gamescon and he gave me a signed delidded 7700k chain, free ifixit kit and a hug, super nice guy. After I got a new CPU I tried to put that IHS on my old 7600k because reasons, didn't go to well.
What a great dude! Roman always seemed like such a cool guy to me, so I’m happy to hear he’s like that IRL. What a shame the IHS transplant didn’t work though.
I bought Thermal Grizzly paste with the thought of “If it’s good enough for overclocking it’s good enough for me”It was also part of the preparations for a planned Pc upgrade that never ended up happening.
If you're not overclocking or running a beast, any of the aftermarket pastes will be far beyond "good enough". At that point, it's worth getting stuff for alternate specs, like longevity or ease of use or cheaper price.
@@ashkebora7262 I was planning on upgrading to a 12900k (the 13th gen wasn’t released yet) at the time I bought it, but that never ended up happening. And so I now have an I7 8700 with an overkill cooling solution. Better than letting the money go to waste. Maybe when my financial situation gets better again I’ll get a 9900k and overclock that. At least my CPU stays cool even with my room being nearly 40*C during summer 😂
@@Anonymus-ih7yb ahhhhh, I see. Yeah, running a 12900k definitely counts as planning on running a beast! Ahh well, it'll hopefully at least last longer. I also try to overkill cooling, as unnecessary as it is for a lot of CPUs. Though with how power-dense the tiny cores are getting these days, it's arguably _always_ a good idea to use as good of a paste as you care to afford, even if it won't need 200W+ of total cooling.
Not always. Some of the most effective marketing in this industry is based on just actual fact. Noctua, as an easy example, is famous for its focus on performance and letting performance and engineering sell the brand. It works. It's just not common because it requires more difficult-to-acquire skills.
@Gamers Nexus I've taken a more jaded view of this because for every good example (Nocuta) you have 10 more bad ones. Not discounting the good but acknowledging the majority. Cheers!
I agree with Gamersnexus here. In big industries, people would only do very conservative figures as it can go south very quickly. The company I worked for would always understate the machine capacity (they would be like around 10% below with additional cuts based on the expected down time per day)
@@RogueWraith909 I feel like they sell at all BECAUSE they're "baby poo brown", tbh. It's not like Noctua hasn't done black fans before. Turns out few people cared about those because everyone's already associated Noctua with the "baby poo brown", and it's unique compared to the millions of "more marketable" black and RGB fans out three.
Two of my favourite tech channels right there! The amount of amazing knowledge that you two (and your teams) have shared with the community is just incredible. Thank you for helping to make me an informed consumer. Just 3 years ago I had never built my own PC, but then I started watching tech channels like GN, Der8auer, LTT, and Level One Techs, and now I’ve built two with my own hands! Learned so much about form factors, platforms and their chipsets, CPU and GPU architecture, PSUs and their safety features/certifications, basic wiring of connectors, PCIE generations and specs, and even how TIMs work. I can’t overstate just how much I’ve learned from you two. Thanks so much Steve and Roman! Back to you Steve!
-Thermal conductivity values will tend to decrease with increasing operational temperature. Though in ranges of 20-100 C they may increase before decreasing at higher temperatures (this can be somewhat material dependent however) -High thermal conductivity values are reported for solid sheet based TIMs usually because they are composed of graphene or graphene-based materials. The issue with this value is that graphene has exceptional thermal conductivity IN THE PLANE of the 2D graphene lattice. But thermal conductivity NORMAL to the graphene plane is significantly lower. -The reason that the solid piece of copper interfaced with the heat source has approximately the same thermal conductivity as the heat source is based on the heat transport. You can think of the system as a multi-compartment system with a rate of heat flux for the thermal paste and another rate of heat flux for the copper sink, the reason the SYSTEM has the same heat flux rate as the paste is that it is the rate determining portion of the heat transport process. Like comparing flow rates of water through chicken wire fence versus a fine mesh. -The thermal conductivity of the copper in silicone oil paste being low is based on, once again the rate-limiting process, which is heat transport through the silicone oil (silicone oil has a much lower thermal conductivity than copper). Increasing particle size of the dispersed metal increases thermal conductivity by increasing interfacial contact area between the heat source and metal, and sink and metal, and will be dominated by the high thermal conductivity of copper. Past a certain particle size however, you will see a decrease in interfacial contact area between the TIM and the surfaces of the parts (since the larger particulate will not easily fill the micro voids and imperfections in each surface). -I'm not sure what they meant by the delta between the cooler and die, if its the separation distance though, thermal conductivity is a measure of Watts per Meter * Kelvin. The meter refers to the thickness of the material through which heat transports. So it follows that if you are transporting heat through a greater thickness of the same material, that the thermal conductivity will decrease.
This is a great video and shows the real issue with current marketing and listed performance specs used by manufacturers and resellers. Even within a manufacturer's product line, they are all over the place. Other times there are apparent manufacturers which are actually just resellers and rebrand products as their own, sometimes claiming a different w/mk rating than that actual manufacturer without using any more additives. I no longer trust w/mk ratings at all and rely on testing to tell the truth. A good example of ratings not being accurate within a manufacturers product line would be T-Globals TG-PP10 thermal Putty (rated at 10w/mk) vs *TG-NSP80 (Rated at 8.3w/mk, *NOT SAFE for electronics as it conducts electricity). In this comparison the TG-PP10 conducts heat almost as well as Gelid GP Extreme pads, where as *TG-NSP80 would theoretically beat all currently available thermal pads (with the exception of 3D Graphite pads). Luckily we now have some extremely high performing Thermal Putties that don't conduct electricity and can effectively be used to replace Thermal pads on VRAM, VRM, etc. For anyone interested in Thermal Putties I currently recommend Upsiren UX Pro as the best performing, although that may change as we discover/test more putties in the future.
I still have a Prolimatech PK-1 Thermal Compound. I once build a Pc for my aunt it was an AMD Athlon 4850e (Dual core). When I bought it, it was relatively new. So could have been mid 2009. Summer 2022 I took the Pc apart that was used till then. I demounted the cooler and the paste was still like day one. It was not dryed out, it was still thick and sticky. I could have remount the cooler directly back on and it would have worked flawlessly I'm sure. Really impressed with the longevity. Just a little shutout to Prolimatech.
Computex 2023 was what was needed! So glad to see many of my tech reviewers able to share space and ideas. Especially after the last few years on this planet.
I like Der8auer, a smart young fellow, I am subscribed to his channel. He is spot on about manufacturers picking a sweet spot for their products out of the minimum to maximum 'Curve' , to make it look like it's the best. Cool discussion / video . Cheers
It stands to reason that they want to cast their product in the best light possible. No one's out there saying buy our stuff because it sucks unless they're selling vacuum cleaners.
I feel like we should have a graph of thermal conductivity showing the results at say 5 or 10 degree intervals between idle temps and thermal throttling (maybe 115C for the top end) to give a reliable, relatable and comparable results table
Okay, we had some really extreme examples of manipulation like testing from LN2 to 250C but what about within like 100C? That would be the worst realistic scenario + little extra on top. I mean if we take 10 or so more well known thermal paste (pastes not LM) makers then during the real life scenario (gaming pc) testing the range stays pretty much within 1 degree which is nothing. So in that sense all the decent pastes are pretty much identical as far as the immidiate temps go. So now comes in the longevity - if the take those 10 or so pastes, apply them betweeen two sheets of metal plates, put in the same chamber with controlled ambient temp and heat one side up to like 90C and then measure the other side after week, month, 2 months etc. so see if some of the pastes tend to fall off.
Steve, you are an amazing person, with a goodwill, the amount of work you make for us is outstanding. Best wishes for you, thanks for everything you're doing!
I work in the R&D field. We just recently started evaluating high conductivity PCBs. It's exactly the same in that market. The numbers don't matter. We have our own standardized test board to test XY as well as Zup and Zdown thermal conductivity, dielectric breakdown, inductive coupling and foil adhesion strength. We've tested the gambit from the aluminum backed PCBs from JLCPcb to a company that charges many hundreds of dollars for the same 10 dollar board. The cheap ones are not as bad as the datasheets make them look. Layout is more important than substrate to an extent. Once you get into high frequency applications, then the more expensive boards start to shine, but that's 10's or megahertz in a power board. Not very common.
At least from my perspective, I find "all of these things perform basically identically so you should just buy whichever one is cheapest" to be a useful conclusion from a review. I'd totally be up for GN testing things where that would be the likely conclusion.
Roman definitely identifies the problem with thermal conductivity numbers. They are meaningless without temperature. I've worked in several chemical labs, some in drinking water treatment. I've seen pH's reported with no temperature association, tried to explain, to everyone. "Nope we don't report temperature!" One operator actually let a sample sit around and warm up to lab temp, in order to raise the pH number to get in spec so they wouldn't have to adjust anything. (Why write the temp? Hahaa! Not so funny on the drinking water end!) So this point hits home for me. On a side note, I came into position where I could correct this, at least in one town!
Thanks for that, that must be the reason why I've never really noticed any difference between different pastes. In the future I will put more weight into reviews when buying paste.
Roman is such a cool and honest guy. When he says it doesn't really make sense to buy the extreme paste...instant respect and trust! Great interview Steve.
Well that was very interesting. As someone with training in engineering, I'd latched on hard to the thermal conductivity number - that's a thing I'm aware of and, when it's measured properly and reported accurately, it's ...an actual physical property of material that exists as much as its tensile strength and resistivity and such. Y'know, things that tell you literally exactly how it performs in that way. Theoretically it could be the single most solid number we talk about in computing short of things like the physical transistor count and such. Good to know that, as reported, it's just marketing BS. And yes, if you're reading this, Der8auer, I would very much be interested in a paste that performs well and can maintain its properties long term. Even if it's not up to your top-of-the-line stuff, endurance is a huge factor for me (and I suspect many other users) as I want something that will last for several years with as little maintenance as I can get away with. I'd love to swap out parts every generation or two, but the cost of major upgrades like that is just beyond what my income (and saving habits, granted) make possible, so my computers tend to go several years without the cooler being pulled off the CPU.
You can calculate that your cooling with say a CM Hyper 212 Evo the total thermal impedance die to air is about 0.35K/W. You just divide temperature delta air to die (full heat path) by wattage consumed, for whatever cooler you have, i expect that you'll find a figure somewhere above 0.2K/W. This figure is of course not the thermal impedance of the cooler, but the whole thermal system including die, heatspreader, heatsink, all intermediate interface materials. You can then multiply the thermal conductivity of the paste by area, divide it by thickness (sub 0.03mm typical), and what comes out is something on the order of 200-500W/K, or alternatively sub 0.005K/W. Imagine burning 200-500W on your CPU, and the paste layer thickness creating a 1°K gradient. Because the thermal conductivity describes the wrong thing, it describes a THICK slab of paste, where surfaces don't matter, while what you have is a THIN layer of paste where most of the thermal impedance occurs at the contact surfaces, and it's not characterised at all. As someone with training in engineering you could have always made these sorts of calculations and not get yourself mislead.
@@SianaGearz Ah, yeah, the issue is I have the wrong kind of engineering training. I studied mining and mineral processing engineering, which is about rock masses, explosives, ventilation, and the various physical and chemical processes used to separate out desired minerals from waste minerals. Very different from what the material engineering students were studying 2 floors down from my classes. So I only got a very basic introduction to thermal conductivity in first year thermodynamics.
Thing is thermal conductivity doesn't matter. Noctua paste has shit conductivity (real) at about 2.5W/mK and it performs well. Kerafol industrial paste has good conductivity (real) of almost 10W/mK but happens to result in much higher CPU temperature. Because what matters is not thermal conductivity but thermal impedance in deltaK/W. m in metres in thermal conductivity is the layer thickness, but larger particles result in larger layer thickness combined with higher nominal conductivity, but worse real performance. Let's say you measured the resulting layer thickness and multiplied conductivity by that and inverted it, you have thermal impedance, right? Wrong. Because thermal impedances from die to the air form a chain, and there's paste somewhere in the middle, but the paste doesn't contribute one impedance, but 3. There's two boundary impedances from CPU heatspreader to paste, and from paste to cooler coldplate. And these aren't characterised at all. Let's say you have a paste with a carrier liquid which doesn't wet the metal very well, you'll have a surprisingly bad boundary impedance; or its mix of particle sizes doesn't nest well in the surface texture, that's going to lower those figures as well. You understand what the problem is? Thermal conductivity adequately describes an infinitely thick slab of paste where boundary impedance can be ignored, but the real PC pastes usually differ in their performance pretty much only in how good their boundary impedance is. The total thermal impedance figure is actually super predictive, used in engineering all the time. You can add up individual impedances in the chain and predict the die temperature from wattage and environment temperature that way, and you can also measure the thermal impedance of the whole system and by varying individual parts, estimate the individual impedances. Though test setups for that to get good individual numbers can get really difficult. And as you note, these can be temperature dependent as well.
@@SianaGearz So why not provide the thermal conductivity / temperature curve for a specified paste thickness and the most common heat spreader/cold plate materials?
Two things: 1) I get what he's saying in terms of how you can "game" the number for marketing purposes which may or may not have any "real" value for an end user/customer when the product is used/deployed in their specific application. There's the mantra that "numbers don't lie", but as an engineer, there's also the mantra of "we can make the test tell you what you want it to tell you as well". But that's also where discussions about the methodology becomes critical, and very few people ever spend time, talking about the test/simulation methodology and/or data collection methodology (to see whether there are glaring issues with it or not). Thermal testing of this kind is actually impossible to do, physically, because the physical thermal couples have their own physical properties (size/dimension, material properties, etc.) to contend with. You can literally get a Ph.D. just by studying the testing methodology and how to try and minimise the influence or the impact that the physical test itself imparts on the test that you're trying to run. 2) Thermal conductivity IS required though, for my CFD simulations. But what he said about thermal conductivity being a function of temperature -- there aren't very many CFD codes that I am aware of, that is actually able to accept k(T).
Just built my new machine, 7950X with an ASUS X670E-A (I know, I bought it before the current issues) and a RTX4090. I used hydronaut as my TIM but to be honest I don't think it would matter what I chose. That 7950X runs so bloody hot (at load) due to it's IHS, if I want it any cooler my only option would be a delid and direct die which I'm not 100% comfortable in doing. Was an interesting build, 128GB of 6400 DDR5 that can NEVER run at 6400, the best I got was 4800 (5000 with occational boot failures.) I do like the 011D Dynamic XL case though. Just don't buy a shift version of a PSU though, you'll have to remove a drivebay, LOL. I was waiting for the new 011D Evo XL which now won't be out until October so got fed up waiting for it.
I LIKE long form, in-depth, content. 50 minutes I've stood on street corners intensely conversing twice as long! A LOT of us like longform; my favourite English tech tuber used to regularly go 60 minutes + and I loved that; now thirty minutes is not a regular thing. Shorts? Go away! We WANT in depth and we want our content creators bouncing of one another! My favourite content has been when y'all invade one anothers content and we get something impromptu and when y'all decide that since we are here, lets do something longer and more specific together. I know we wrag you pair, but it is rather marvelous and very entertaining! Successful enthusiasts with serious cashflow and turnover! Hot damn; we are so lucky to have you! All the best and onwards and upwards, guys.
I remember when I changed the alphacool Sub Zero thermal paste, that was very good and I had perfect temperatures, for the Alphacool Apex thermal paste, I had a great improvement because of the better thermal conductivity. Made the same (really I did here first) in the Gpu, where everything was a good, a perfect starting point, and I had 5, 6 degrees more of improvement. Thermal conductivity of thermal paste and thermal pads was more important of what I thought in a first time. Liquid metal I will try in the future. Great meeting!
TMIs really only do one thing: smooth out bumps and voids in the IHS and cold plate surfaces by replacing air (basically zero thermal conductivity) with paste (alright-ish thermal conductivity). If you do the math for a cold plate/paste/IHS system with realistic paste layer height and thermal conductivities, the cold plate is like 0.02 K cooler than the IHS. The thermal resistance (superior metric imo) is tiny.
@4:35 To go a bit 'Jay', I make a car analogy here: The most fun cars don't have the best 0-60 mph or 0-100 km/h numbers, which most manufacturers marked with. They go for the best acceleration and feel, that isn't in line with the highest acceleration, as most of the time you spend in certain speeds, that the 'faster' car really is not geared for.
Mit Acht KEKW Jokes aside, I've actually steered away from Kryonaut since I cba to reapply thermal paste multiple times a year, as the community does point out longevity issues with the product, even if it is the "best" thing out there. Same with Liquid cooling, it's most likely better than my big tower cooler, but there I can just visually check without any hassles if a fan has failed and don't have to hunt for eventual rogue fluid drops in the case or try and listen if the pump has failed.
I still want to see (just as a "what if") a heatsink or AIO made almost entirely out of machined silver. Not "silver coated", I mean uncoated, corroding SILVER. Just to see what the difference in conductivity and heat dissipation looks like. The price would be impractical, but it would still be fun to see. I mean, I know you can buy .999 electrolytically refined silver ingots, and there a lot of hobbyists with CNCs at home. Curious if it would actually make a noticeable difference.
For me, MX4 is king. Marketing says it last 8 years. I had MX4 for more than 2 years on an air cooled FX 8320 OC to 4.8GHz, thermals didnt change from first to last day. After that I got a FX 9590 and installed with the arctic gooey with the same air heatsink, excellent temps until I upgraded to ryzen. I have machines using MX4 for 5 or 6 years doing the same thermals than the first day. -It is cheap, got a 20g tube for 10 bucks. - It performs, No problem with an air cooled FX 9590. -Easy to apply, Like hot butter. -Last long time. Like a lot
I can't remember when I pasted up this PC. Maybe 2017? Yeah sounds about right. Temps are still good. I'll see how it is in a couple more years. If I see temps go up I'll repaste it.
(insert "thanks Steve" meme here) I love these collaborations, specially this way, chill and comfy, even more, as friends, making the videos very enjoyable, beyond the informative content. Thank you guys! Successes and blessings!
Gotta mix those particle sizes. Some large particles to increase conductivity, some much smaller particles to fill the gaps so it still acts like a paste. I think that is what Tech Ingredients did.
@@thefirstloser In my opinion Kryonaut is great for CPUs with a heatspreader but has too much pump-out effect for anything direct-die. Pump-out is why I don't use MX-6 even though newer should be better.
Love it when Roman do piece discussion with you, guys! Wish it was a more common thing! But, well, distance is a biatch! Now, on about the testing: On die application Pump-out effect is the biggest issue and the one that should be focused the most when testing pastes. A lot of "high" end pastes perform between a few degrees of each other when applied on a IHS, but, when it comes to die application, few passes the test after some thermal cicles because of pump out effect. Id love to see you guys do a huge round up on this subject with the most popular pastes.
A very thorough test of thermal pasts would probably take a couple years and a super consistent way of applying it. Testing by applying to a system and doing a heat tests immediately and again some time down the road. The trick is making a lengthy lab test into content.
Thermal conductivity outside of marketing hyperbole is a study I've taken to the extreme for composite jet engine parts; I've often thought about adapting my process for the ultimate heat sink for a CPU, but continue to run into practicality hindrances such that performance would be so far beyond what is currently accepted as 'high' that it would almost be a waste of materials. That, and as of the present, it couldn't be made for any price the typical PC enthusiast would call acceptable. Being designed to withstand extreme temperatures in constant use for a given lifecycle at 3000*C, for a PC part it would essentially be an infinite shelf life with no discernable wear. The flip side is that I couldn't manufacture it for less than $200 without going to industrial scale, which is not the direction I have in mind for the company or this particular process.
Look i think $200 for an ultra performance item - even if i don't need it and probably nobody "needs" it, seems outright cheap. There's a few of those people out there, someone is buying up all the 4090s don't they - some of these people actually put the performance to use in one manner or another; but others don't really do that and just buy because they want the absolute best thing no matter the cost. I think you can mark it up to a grand and give it a shot, i wish you good luck. There are all manner of mildly outrageous things of extremely niche use.
@@C-M-E I wouldn't really trust myself to do that anywhere near competently. I had an introduction to marketing in school (2 weeks) so uhhh better than nothing i guess?
It's been so long I don't even remember what paste is on my CPU. It's a 4790k purchased when Skylake came out and I refused to pay the memory premium. It's been spending its life at 4.4 with a vcore offset in the .9s. With a CM 212 on it. Even the thought of having to worry about reapplying thermal paste on CPU during it's normal lifecycle is enough for me to never purchase a paste that has that as a concern.
Thermal conductivity is just the material property, the inverse of thermal resistivity. The thickness (and area) of the material also has a huge impact on the amount of heat transfer you get, and the layer of thermal compound is typically very thin and spread wide over the IHS. Because of this, as long as the material is generally a thermally conductive material it will transfer the heat effectively; with how thin the layer is, there are diminishing returns by only changing the material. A material with a high thermal resistivity (low thermal conductivity) will of course inhibit the heat transfer, such as air gaps, contaminants, or just a bad material, like the plastic film layer you get on a new cooler that people sometimes forget to remove. I think someone on youtube (possibly Linus?) performed some testing many years ago with other materials such as toothpaste, and they worked pretty much as well as some of the leading thermal compounds on the market, at least over a very short period of time (as you say, there are other factors relating to longevity). So I completely agree, thermal conductivity basically doesn't matter. Just use the thermal compound that comes with the cooler unless you're going for overclock records.
The boundary layer between two different materials has a thermal impedance of its own. And this is why thermal conductivity doesn't matter, since it doesn't describe that boundary, just the conductivity through the bulk of the material, and the thermal impedance of the paste layer is an order of magnitude smaller typically than the boundaries to the both adjoining metals.
How good is Artic MX 6, cus the reviews are amazing on that paste? Long lasting, cheap, amazing cooling... it has it all. I hear the MX 5 was bad tho, so yeah. I should have gone with Noctua 2, but oh well. So far so good with the Artic one.
Remember graphite thermal pads , pepperidge farm remembers . They were the talk back in the day about how long lasting they were and they somehow stopped being mentioned all together .
(And because we didn't expand on it here: Under identical test conditions or for base material properties, thermal conductivity is of course a real thing and is useful -- say comparing aluminum to copper -- but it's when you start making pastes and tuning the properties just for a marketing number that Der8auer needs an anger translator!)
Check out our previous video with Der8auer speaking on this topic here: ua-cam.com/video/CCqxE-5Ct3w/v-deo.html
And check out Der8auer on UA-cam (auf Deutsch): ua-cam.com/users/der8auer
Und auf Englisch: www.youtube.com/@der8auer-en
Well its not something JDEC can step in and solve
We just need to hope that manufacturers start to do the test at 95c and base the specs sheet off of that
Because we have moved from CPU’s ramping up over time to them ramping down instead
Hence pegging at 95c and so the paste needs to be spec’ed for 95c
No mentions of thermal resistance. Still remember - Kryo33 made a point (they probably have a lab you talked about) that paste does not contain silicone, or any additives, and viscosity comes from finesse of the material, and, I speculate, that smaller the grains > closer to each other > better thermal transfer, because role of paste is not to conduct, but to transfer the heat. Often parameters come in two values conductivity and resistance - such as conductivity 13W/mK, thermal resistance
Had my sister use household lubrication mineral oil as "paste" on a CPU for 3+ years, because she is living in annother country and was unable to get her hands on real paste. So, I talked her trough all steps of building a PC and she did not have thermal paste. It was a custom "as cheap as we can make it, but new parts" office PC and I could not visit her. It is still running, albeit she did buy some paste after a few years and applied it.
I am a big fan of liquid metal. Used it on my AMD K6-2 the first time. Loved it. But yeah, there is a lot of fancy talk about heat paste, which a lot of people do not need. Even for overclocking gamers it is better to get a good paste and refresh it after a while, instead of getting the "best of the best" and never touch it again.
I miss the nerdy OC forums, which tested all the waterblocks, heat pastes and cascading compressor cooling solutions, with zero bias.
There's also thermal pump out that can occur that most reviewers won't see in the short term.
@@Niko-qk3su smaller particle size can only do so much as its fully bottlenecked by the carrier material and how many of those particles even actually touch each other
Its why copper shim modding Nvidia 3000 series is as good as it is
when a German and American tech nerd have a playdate, this is what they talk about
hahaha, and it's apparently something we talk about every single time we meet up!
Allo allo
And we do listen carefully because it's is friggin interesting. Sucks to be a nerd sometimes 🤣
well put.
Hasta la vista baby, I drive..
Props to GN for featuring a lot of other channels/creators during computer.
Computer is my favourite trade show
Ah yes. Can't wait to go computer next year.
@@XxXnonameAsDXxX Noice👌🏻,🤣
@@XxXnonameAsDXxX I celebrate computer all year round
I'm glad computer is exist.
Roman: "If I turn this into dust, and add a little bit of silicon oil..."
Steve: "What're you, a f**in' alchemist?"
You guys have the best conversations, always such a great blend of tech information and relaxed humor. Thanks GN & der8auer!
Well, Intel certainly is!
"Community voted bromance of the show"
5 seconds in and I'm already dead!!! 😂🤣😂
closet got ya eh?
Coming from the audio industry, don't get me started on the lack of standards. Great talk. 👍
How do you mean golden cables do not make sense?!
@@Lodinn I use platinum cables only.
@@NeroKoso I use Flacanium with 0% seasalt impurities. Makes the artists actually appear in the room.
On the hardware or software side?
@@sinsiliuxs120 probably both
Steve - "we talked long enough to need to split them"
The audience - "No, you didn't."
Snowflake's kibbles budget thinking about adsense: "Yes, you did."
@@SirGingerOfKnight *Spot-on!*
Jesus loves you so much and He is waiting for you with open arms, please accept Him in your heart and life while He is near Jesus loves you so much and He is waiting for you with open arms, please accept Him in your heart and life while He is near
Kinda like how NVIDIA misleads people with stuff like the 4070 ti being "3x faster than a 3090 ti"
ooooooooof
@@GamersNexus it's like they think people in the market for these high end GPUs are stupid.
That wasn’t misleading tho, it was a blatant lie
not a single soul
@@vvb890 it's both
Der8auer's channel has some fantastic content, especially for higher-level enthusiasts. He's also an actual engineer, so some of his commentary can get pretty involved and nuanced (as you see here) so prepare to have a second window open for WIkipedia~
I love how he talks a lot about engineering stuff in some of his videos, it really gives an unique touch to them. One of the tools he used to test a coldplate costs 7000 euros, lol, true engineer stuff.
Recently did a PowerColor GPU factory tour.
Just remember he split English and German videos to different channels.
“Anyway... temperature’s fake and uh... energy doesn’t exist.”
I knew it!
Earth flat, IHS' are flat therefore the Earth is an IHS?
I lmao so hard to this. Great guys and great video
@@Pegaroo_ big brain moment. does this make the atmosphere TIM?
@@warmachine5835 no TIM is all CGI propaganda from NASA
Yo! Thank you for being so kind after I ran into you at the airport and said hello. I appreciate the welcoming hello.
Thanks for saying hi and for watching the content!
"der8auer spelled mit Acht" This is very important and we should only call him like this from now on
I thought his name is called as Der AchtAuah, learning that it was Bauer messed it up for me
What does it mean? Is it “at eight”?
@@Quizack It means he replaced the "B" in "Bauer" with a number 8.
@@andreasu.3546 Oh yeah now I get it. Appreciate the answer. Thanks
THANKS STEVE!
Thanks Roman!
Hydronaut hits all that and then some. and easy to apply
@@warpedphreak how is hydronaut different than kryonaut? I've heard of both but idk what the difference is
arctic
@@TheAechBomb kryonaut was designed for sub zero cooling, and it excels at that, but people usually don't daily drive sub zero, and instead frequently change the heat sink and repaste, meaning that they didn't need to design kryonaut for longevity. I don't know exactly how long it remains in an uncracked state, but it's very short compared to pastes that are designed for long term use, such as hydronaut.
@@insu_na Interesting and very useful reply to a question I didn't ask but did need the answer to, thankyou! Glad someone asked it too!
Wow that ended so fast, I feel like I need the uncut 50 minutes.
Good to know that thermal conductivity numbers are about as useful as TDP numbers. Just make everything up to get numbers you like!
Vote for uncut 😮
can't say it enough, these 2 need more content together.
You guys have such good chemistry. Always a joy to watch you guys.
Thermal conductivity has a very common standard using astm D5470 which is similar to the use case. The supplier should note the standard used. Laser flash and other metroligies are basically random number generators.
Also, the number that matters is thermal impedance. Thermal conductivity is better number for heat spreaders while impedence tells you how good of a tim the material is.
These things are true but lab results communicated ineffectively to end users who are not sophisticated and only want to know "does it work?" "How long?" "Is it easy to use?" and of course "price?" just do not help overmuch.
Standards can be misused as well as correctly used. Sometimes the guy who wants to build a computer just wants IT TO WORK OKAY.
That I think is probably the largest segment of the market.
@@phlogistanjones2722 the largest segment of the market is OEMs and ODMs. Selling to individual builders is barely any market share.
@@CutaneousSensationsI would argue that the same points that matter to consumers matter even more to OEMs and ODMs, longevity, ease of application, dangers like with Liquid Metal and how it’s conductive electrically, and price. For average Joe those things matter but less, I’d spend 30$ more to get 2deg C and have to spend an extra 10min applying it, for an OEM that just won’t do when they are looking at fractions of Pennies in their build. If I was an OEM I would look #1 at longevity, needs to be at least 5 years, next pick the top 5 lowest priced pastes that fit that criteria and trial a couple units of each in my existing assembly process and evaluate the results. Then I would go with the cheapest finest paste that performs well enough to stay within the cpu manufacturer specs for temperature.
@@radugrigoras all suppliers worth their salt do provide basics understanding on how their materials survive for 1008 h bake at 125c, temperature cycle B 1000 cycles, and uhast 85/85 1008h. Sadly this will not translate directly and the OEM/ODMs will have to verify it in their test vehicles and products since the use conditions will be unque. So basically they already do that. Only small material suppliers don't provide this data.
@@radugrigoras also, you are crazy to think they don't try to find the cheapest paste that will meet their minimum requirements. That's their goal. They work with the material suppliers to develop this.
"Der Bauer mit Acht" i lol'd heartily
Met Der8auer few years ago at Gamescon and he gave me a signed delidded 7700k chain, free ifixit kit and a hug, super nice guy.
After I got a new CPU I tried to put that IHS on my old 7600k because reasons, didn't go to well.
What a great dude! Roman always seemed like such a cool guy to me, so I’m happy to hear he’s like that IRL. What a shame the IHS transplant didn’t work though.
Holy Moly! Did Der8auer just take a dump on Thermal Grizzly? Dude's got balls.
His own company
I bought Thermal Grizzly paste with the thought of “If it’s good enough for overclocking it’s good enough for me”It was also part of the preparations for a planned Pc upgrade that never ended up happening.
If you're not overclocking or running a beast, any of the aftermarket pastes will be far beyond "good enough".
At that point, it's worth getting stuff for alternate specs, like longevity or ease of use or cheaper price.
@@ashkebora7262 I was planning on upgrading to a 12900k (the 13th gen wasn’t released yet) at the time I bought it, but that never ended up happening. And so I now have an I7 8700 with an overkill cooling solution. Better than letting the money go to waste. Maybe when my financial situation gets better again I’ll get a 9900k and overclock that. At least my CPU stays cool even with my room being nearly 40*C during summer 😂
@@Anonymus-ih7yb ahhhhh, I see. Yeah, running a 12900k definitely counts as planning on running a beast!
Ahh well, it'll hopefully at least last longer. I also try to overkill cooling, as unnecessary as it is for a lot of CPUs.
Though with how power-dense the tiny cores are getting these days, it's arguably _always_ a good idea to use as good of a paste as you care to afford, even if it won't need 200W+ of total cooling.
Misleading and marketing go hand in hand.
Not always. Some of the most effective marketing in this industry is based on just actual fact. Noctua, as an easy example, is famous for its focus on performance and letting performance and engineering sell the brand. It works. It's just not common because it requires more difficult-to-acquire skills.
@Gamers Nexus I've taken a more jaded view of this because for every good example (Nocuta) you have 10 more bad ones. Not discounting the good but acknowledging the majority. Cheers!
@@GamersNexus They'd sell more if their fans weren't "baby poo brown"... they are really good performance wise though!
I agree with Gamersnexus here. In big industries, people would only do very conservative figures as it can go south very quickly.
The company I worked for would always understate the machine capacity (they would be like around 10% below with additional cuts based on the expected down time per day)
@@RogueWraith909 I feel like they sell at all BECAUSE they're "baby poo brown", tbh. It's not like Noctua hasn't done black fans before. Turns out few people cared about those because everyone's already associated Noctua with the "baby poo brown", and it's unique compared to the millions of "more marketable" black and RGB fans out three.
you 2 are my favourite tech youtubers!
Two of my favourite tech channels right there! The amount of amazing knowledge that you two (and your teams) have shared with the community is just incredible. Thank you for helping to make me an informed consumer. Just 3 years ago I had never built my own PC, but then I started watching tech channels like GN, Der8auer, LTT, and Level One Techs, and now I’ve built two with my own hands! Learned so much about form factors, platforms and their chipsets, CPU and GPU architecture, PSUs and their safety features/certifications, basic wiring of connectors, PCIE generations and specs, and even how TIMs work. I can’t overstate just how much I’ve learned from you two. Thanks so much Steve and Roman! Back to you Steve!
love derbauer! great content
-Thermal conductivity values will tend to decrease with increasing operational temperature. Though in ranges of 20-100 C they may increase before decreasing at higher temperatures (this can be somewhat material dependent however)
-High thermal conductivity values are reported for solid sheet based TIMs usually because they are composed of graphene or graphene-based materials. The issue with this value is that graphene has exceptional thermal conductivity IN THE PLANE of the 2D graphene lattice. But thermal conductivity NORMAL to the graphene plane is significantly lower.
-The reason that the solid piece of copper interfaced with the heat source has approximately the same thermal conductivity as the heat source is based on the heat transport. You can think of the system as a multi-compartment system with a rate of heat flux for the thermal paste and another rate of heat flux for the copper sink, the reason the SYSTEM has the same heat flux rate as the paste is that it is the rate determining portion of the heat transport process. Like comparing flow rates of water through chicken wire fence versus a fine mesh.
-The thermal conductivity of the copper in silicone oil paste being low is based on, once again the rate-limiting process, which is heat transport through the silicone oil (silicone oil has a much lower thermal conductivity than copper). Increasing particle size of the dispersed metal increases thermal conductivity by increasing interfacial contact area between the heat source and metal, and sink and metal, and will be dominated by the high thermal conductivity of copper. Past a certain particle size however, you will see a decrease in interfacial contact area between the TIM and the surfaces of the parts (since the larger particulate will not easily fill the micro voids and imperfections in each surface).
-I'm not sure what they meant by the delta between the cooler and die, if its the separation distance though, thermal conductivity is a measure of Watts per Meter * Kelvin. The meter refers to the thickness of the material through which heat transports. So it follows that if you are transporting heat through a greater thickness of the same material, that the thermal conductivity will decrease.
This is a great video and shows the real issue with current marketing and listed performance specs used by manufacturers and resellers.
Even within a manufacturer's product line, they are all over the place. Other times there are apparent manufacturers which are actually just resellers and rebrand products as their own, sometimes claiming a different w/mk rating than that actual manufacturer without using any more additives. I no longer trust w/mk ratings at all and rely on testing to tell the truth.
A good example of ratings not being accurate within a manufacturers product line would be T-Globals TG-PP10 thermal Putty (rated at 10w/mk) vs *TG-NSP80 (Rated at 8.3w/mk, *NOT SAFE for electronics as it conducts electricity). In this comparison the TG-PP10 conducts heat almost as well as Gelid GP Extreme pads, where as *TG-NSP80 would theoretically beat all currently available thermal pads (with the exception of 3D Graphite pads). Luckily we now have some extremely high performing Thermal Putties that don't conduct electricity and can effectively be used to replace Thermal pads on VRAM, VRM, etc.
For anyone interested in Thermal Putties I currently recommend Upsiren UX Pro as the best performing, although that may change as we discover/test more putties in the future.
I still have a Prolimatech PK-1 Thermal Compound. I once build a Pc for my aunt it was an AMD Athlon 4850e (Dual core). When I bought it, it was relatively new. So could have been mid 2009. Summer 2022 I took the Pc apart that was used till then. I demounted the cooler and the paste was still like day one. It was not dryed out, it was still thick and sticky. I could have remount the cooler directly back on and it would have worked flawlessly I'm sure. Really impressed with the longevity.
Just a little shutout to Prolimatech.
Computex 2023 was what was needed!
So glad to see many of my tech reviewers able to share space and ideas.
Especially after the last few years on this planet.
I like Der8auer, a smart young fellow, I am subscribed to his channel.
He is spot on about manufacturers picking a sweet spot for their products out of the minimum to maximum 'Curve' , to make it look like it's the best.
Cool discussion / video .
Cheers
It stands to reason that they want to cast their product in the best light possible. No one's out there saying buy our stuff because it sucks unless they're selling vacuum cleaners.
I feel like we should have a graph of thermal conductivity showing the results at say 5 or 10 degree intervals between idle temps and thermal throttling (maybe 115C for the top end) to give a reliable, relatable and comparable results table
Thanks Steve.
If u guys made a podcast, i will listen to it all day
Respect is earned. der8auer is famous for good reason. This was an excellent video and needs further explanation with GN styled charts. 😀
data has been out there a long time. wtf you schmucks been waiting for? invitation?
I would love to see more TIM benchmarks. *nudge nudge*
This talk just shows how much they are needed.
I hope you will upload full conversation. It’s always good to hear two enthusiast talking.
Okay, we had some really extreme examples of manipulation like testing from LN2 to 250C but what about within like 100C? That would be the worst realistic scenario + little extra on top. I mean if we take 10 or so more well known thermal paste (pastes not LM) makers then during the real life scenario (gaming pc) testing the range stays pretty much within 1 degree which is nothing. So in that sense all the decent pastes are pretty much identical as far as the immidiate temps go.
So now comes in the longevity - if the take those 10 or so pastes, apply them betweeen two sheets of metal plates, put in the same chamber with controlled ambient temp and heat one side up to like 90C and then measure the other side after week, month, 2 months etc. so see if some of the pastes tend to fall off.
Gotta say, I love Der8auer's sleeve, he's clearly a man of culture!
I would like to know what gate address that is??
@@Borg1701 ua-cam.com/video/qCFQGfLCsKQ/v-deo.html He says its from the episode "Window of Opportunity"
@@Yoshi92 Thanks, I was curious as well. Great episode too!
Very interesting talk, i loved the video. A little different but also very useful!
Steve, you are an amazing person, with a goodwill, the amount of work you make for us is outstanding. Best wishes for you, thanks for everything you're doing!
I work in the R&D field. We just recently started evaluating high conductivity PCBs. It's exactly the same in that market. The numbers don't matter. We have our own standardized test board to test XY as well as Zup and Zdown thermal conductivity, dielectric breakdown, inductive coupling and foil adhesion strength. We've tested the gambit from the aluminum backed PCBs from JLCPcb to a company that charges many hundreds of dollars for the same 10 dollar board. The cheap ones are not as bad as the datasheets make them look. Layout is more important than substrate to an extent. Once you get into high frequency applications, then the more expensive boards start to shine, but that's 10's or megahertz in a power board. Not very common.
At least from my perspective, I find "all of these things perform basically identically so you should just buy whichever one is cheapest" to be a useful conclusion from a review. I'd totally be up for GN testing things where that would be the likely conclusion.
Roman definitely identifies the problem with thermal conductivity numbers. They are meaningless without temperature. I've worked in several chemical labs, some in drinking water treatment. I've seen pH's reported with no temperature association, tried to explain, to everyone. "Nope we don't report temperature!" One operator actually let a sample sit around and warm up to lab temp, in order to raise the pH number to get in spec so they wouldn't have to adjust anything. (Why write the temp? Hahaa! Not so funny on the drinking water end!) So this point hits home for me. On a side note, I came into position where I could correct this, at least in one town!
Great to see Germans and Americas Hardware and OCs #1 👍
It's always good to have Broman onboard. And twice the ranting, oh my.
Dude at the end i was no way this video is ending it has only just begun. Can't wait for pt.2
The amount of physical mirroring with how y’all touch your hair and hold your arms is adorable
derbauer the guy who recently tried to flame AMD over energy efficiency...... so professional
I never noticed that Der8auer has a Stargate tattoo, that's pretty cool
i love it too see you two guys work together..... RESPEKT to BOOTH !!!
The thumbnail for this video is hilarious! It looks like Steve is disgusted at the amount of thermal paste used and Der8auer is amused by it.
4:15 I was actually wondering exactly this when I was shopping these, why some products didn't have thermal conductivity numbers. Glad you covered it!
Thanks for that, that must be the reason why I've never really noticed any difference between different pastes. In the future I will put more weight into reviews when buying paste.
yeah...we gonna need to see that whole convo bro. in advance Thanks Steve!
Roman is such a cool and honest guy. When he says it doesn't really make sense to buy the extreme paste...instant respect and trust! Great interview Steve.
anyone reading basic data knew this a long time ago. the fact someone waited til now to believe it because "he said so" is just friggin sad...
I love watching stuff like this, getting the little bit of understanding from the people engineering something. Can't wait for the other parts.
Well that was very interesting. As someone with training in engineering, I'd latched on hard to the thermal conductivity number - that's a thing I'm aware of and, when it's measured properly and reported accurately, it's ...an actual physical property of material that exists as much as its tensile strength and resistivity and such. Y'know, things that tell you literally exactly how it performs in that way. Theoretically it could be the single most solid number we talk about in computing short of things like the physical transistor count and such.
Good to know that, as reported, it's just marketing BS.
And yes, if you're reading this, Der8auer, I would very much be interested in a paste that performs well and can maintain its properties long term. Even if it's not up to your top-of-the-line stuff, endurance is a huge factor for me (and I suspect many other users) as I want something that will last for several years with as little maintenance as I can get away with. I'd love to swap out parts every generation or two, but the cost of major upgrades like that is just beyond what my income (and saving habits, granted) make possible, so my computers tend to go several years without the cooler being pulled off the CPU.
You can calculate that your cooling with say a CM Hyper 212 Evo the total thermal impedance die to air is about 0.35K/W. You just divide temperature delta air to die (full heat path) by wattage consumed, for whatever cooler you have, i expect that you'll find a figure somewhere above 0.2K/W. This figure is of course not the thermal impedance of the cooler, but the whole thermal system including die, heatspreader, heatsink, all intermediate interface materials.
You can then multiply the thermal conductivity of the paste by area, divide it by thickness (sub 0.03mm typical), and what comes out is something on the order of 200-500W/K, or alternatively sub 0.005K/W. Imagine burning 200-500W on your CPU, and the paste layer thickness creating a 1°K gradient. Because the thermal conductivity describes the wrong thing, it describes a THICK slab of paste, where surfaces don't matter, while what you have is a THIN layer of paste where most of the thermal impedance occurs at the contact surfaces, and it's not characterised at all.
As someone with training in engineering you could have always made these sorts of calculations and not get yourself mislead.
@@SianaGearz Ah, yeah, the issue is I have the wrong kind of engineering training. I studied mining and mineral processing engineering, which is about rock masses, explosives, ventilation, and the various physical and chemical processes used to separate out desired minerals from waste minerals. Very different from what the material engineering students were studying 2 floors down from my classes. So I only got a very basic introduction to thermal conductivity in first year thermodynamics.
Der Bauer spelled with eight is just so funny, thanks Steve, back to you, Steve!
One Love!
Always forward, never ever backward!!
☀️☀️☀️
💚💛❤️
🙏🏿🙏🙏🏼
The thermal conductivity curve over temperature should be the standard, just as with PSU efficiency
Thing is thermal conductivity doesn't matter. Noctua paste has shit conductivity (real) at about 2.5W/mK and it performs well. Kerafol industrial paste has good conductivity (real) of almost 10W/mK but happens to result in much higher CPU temperature. Because what matters is not thermal conductivity but thermal impedance in deltaK/W. m in metres in thermal conductivity is the layer thickness, but larger particles result in larger layer thickness combined with higher nominal conductivity, but worse real performance. Let's say you measured the resulting layer thickness and multiplied conductivity by that and inverted it, you have thermal impedance, right? Wrong. Because thermal impedances from die to the air form a chain, and there's paste somewhere in the middle, but the paste doesn't contribute one impedance, but 3. There's two boundary impedances from CPU heatspreader to paste, and from paste to cooler coldplate. And these aren't characterised at all. Let's say you have a paste with a carrier liquid which doesn't wet the metal very well, you'll have a surprisingly bad boundary impedance; or its mix of particle sizes doesn't nest well in the surface texture, that's going to lower those figures as well. You understand what the problem is? Thermal conductivity adequately describes an infinitely thick slab of paste where boundary impedance can be ignored, but the real PC pastes usually differ in their performance pretty much only in how good their boundary impedance is.
The total thermal impedance figure is actually super predictive, used in engineering all the time. You can add up individual impedances in the chain and predict the die temperature from wattage and environment temperature that way, and you can also measure the thermal impedance of the whole system and by varying individual parts, estimate the individual impedances. Though test setups for that to get good individual numbers can get really difficult. And as you note, these can be temperature dependent as well.
@@SianaGearz Interesting read. I will still call you a nerd
@@SianaGearz Alphacool uses Thermal Impedance along with few other metrics and the exact ASTM standard used (at least for the Apex TIM)
@@SianaGearz So why not provide the thermal conductivity / temperature curve for a specified paste thickness and the most common heat spreader/cold plate materials?
@@GewelReal Thank you
I need a der8auer and Steve podcast.
Somebody at the show saw them and said get a room.....so they did and we got broke back motherboard.
broke back mounting?
Two things:
1) I get what he's saying in terms of how you can "game" the number for marketing purposes which may or may not have any "real" value for an end user/customer when the product is used/deployed in their specific application.
There's the mantra that "numbers don't lie", but as an engineer, there's also the mantra of "we can make the test tell you what you want it to tell you as well".
But that's also where discussions about the methodology becomes critical, and very few people ever spend time, talking about the test/simulation methodology and/or data collection methodology (to see whether there are glaring issues with it or not).
Thermal testing of this kind is actually impossible to do, physically, because the physical thermal couples have their own physical properties (size/dimension, material properties, etc.) to contend with.
You can literally get a Ph.D. just by studying the testing methodology and how to try and minimise the influence or the impact that the physical test itself imparts on the test that you're trying to run.
2) Thermal conductivity IS required though, for my CFD simulations.
But what he said about thermal conductivity being a function of temperature -- there aren't very many CFD codes that I am aware of, that is actually able to accept k(T).
Here they are!
I was waiting for this video :D
Now I hope there's a video with Gordon too!
Just built my new machine, 7950X with an ASUS X670E-A (I know, I bought it before the current issues) and a RTX4090. I used hydronaut as my TIM but to be honest I don't think it would matter what I chose. That 7950X runs so bloody hot (at load) due to it's IHS, if I want it any cooler my only option would be a delid and direct die which I'm not 100% comfortable in doing.
Was an interesting build, 128GB of 6400 DDR5 that can NEVER run at 6400, the best I got was 4800 (5000 with occational boot failures.) I do like the 011D Dynamic XL case though. Just don't buy a shift version of a PSU though, you'll have to remove a drivebay, LOL. I was waiting for the new 011D Evo XL which now won't be out until October so got fed up waiting for it.
Been using Arctic Silver FOREVER... No plains to spend any more.
I LIKE long form, in-depth, content. 50 minutes I've stood on street corners intensely conversing twice as long! A LOT of us like longform; my favourite English tech tuber used to regularly go 60 minutes + and I loved that; now thirty minutes is not a regular thing. Shorts? Go away! We WANT in depth and we want our content creators bouncing of one another! My favourite content has been when y'all invade one anothers content and we get something impromptu and when y'all decide that since we are here, lets do something longer and more specific together. I know we wrag you pair, but it is rather marvelous and very entertaining! Successful enthusiasts with serious cashflow and turnover! Hot damn; we are so lucky to have you! All the best and onwards and upwards, guys.
I remember when I changed the alphacool Sub Zero thermal paste, that was very good and I had perfect temperatures, for the Alphacool Apex thermal paste, I had a great improvement because of the better thermal conductivity. Made the same (really I did here first) in the Gpu, where everything was a good, a perfect starting point, and I had 5, 6 degrees more of improvement. Thermal conductivity of thermal paste and thermal pads was more important of what I thought in a first time. Liquid metal I will try in the future. Great meeting!
Fantastic! I keep ordering based on this rating they slap on pastes and this has taught me a lot
TMIs really only do one thing: smooth out bumps and voids in the IHS and cold plate surfaces by replacing air (basically zero thermal conductivity) with paste (alright-ish thermal conductivity). If you do the math for a cold plate/paste/IHS system with realistic paste layer height and thermal conductivities, the cold plate is like 0.02 K cooler than the IHS. The thermal resistance (superior metric imo) is tiny.
"Don't go chasing thermal values, Listen to the Reviewers that you're use to" - Der8auer. in tune of Waterfalls - TLC
Always great content! You guys cant tell me if Der8auer had red hair he wouldn't look like the Shermantator! (American Pie)
"Number one community voted bromance of the show" is something I didn't expect to hear lol
I love Roman's choice of tattoos! Stargate AND Doctor Who!
Indeed!
@4:35 To go a bit 'Jay', I make a car analogy here:
The most fun cars don't have the best 0-60 mph or 0-100 km/h numbers, which most manufacturers marked with.
They go for the best acceleration and feel, that isn't in line with the highest acceleration, as most of the time you spend in certain speeds, that the 'faster' car really is not geared for.
6:35 my immediate thought upon hearing the particle sizes was
"Do you prefer creamy thermal paste? Or chunky thermal paste?"
Lmao. “Would you like your Kryonaut with or without pulp?”
This fits right in line with IgorsLab recommendation for thick and very soft thermal pads but with low/medium thermal conductivity for GPUs!
love love love Der 8Auer, should only been shown in soft focus and slow motion
Mit Acht KEKW
Jokes aside, I've actually steered away from Kryonaut since I cba to reapply thermal paste multiple times a year, as the community does point out longevity issues with the product, even if it is the "best" thing out there. Same with Liquid cooling, it's most likely better than my big tower cooler, but there I can just visually check without any hassles if a fan has failed and don't have to hunt for eventual rogue fluid drops in the case or try and listen if the pump has failed.
I still want to see (just as a "what if") a heatsink or AIO made almost entirely out of machined silver. Not "silver coated", I mean uncoated, corroding SILVER. Just to see what the difference in conductivity and heat dissipation looks like. The price would be impractical, but it would still be fun to see.
I mean, I know you can buy .999 electrolytically refined silver ingots, and there a lot of hobbyists with CNCs at home. Curious if it would actually make a noticeable difference.
Der Achtauer ist ein richtiger Held! 😊 Love these vids by Gamer Ne10us, keep it up!
Had a chuckle to myself. If you didn't know the protagonists of this video, one would assume the ranting Derbauer was on the right in the thumbnail.
For me, MX4 is king.
Marketing says it last 8 years. I had MX4 for more than 2 years on an air cooled FX 8320 OC to 4.8GHz, thermals didnt change from first to last day.
After that I got a FX 9590 and installed with the arctic gooey with the same air heatsink, excellent temps until I upgraded to ryzen.
I have machines using MX4 for 5 or 6 years doing the same thermals than the first day.
-It is cheap, got a 20g tube for 10 bucks.
- It performs, No problem with an air cooled FX 9590.
-Easy to apply, Like hot butter.
-Last long time. Like a lot
I can't remember when I pasted up this PC. Maybe 2017? Yeah sounds about right. Temps are still good. I'll see how it is in a couple more years. If I see temps go up I'll repaste it.
(insert "thanks Steve" meme here) I love these collaborations, specially this way, chill and comfy, even more, as friends, making the videos very enjoyable, beyond the informative content. Thank you guys! Successes and blessings!
Gotta mix those particle sizes. Some large particles to increase conductivity, some much smaller particles to fill the gaps so it still acts like a paste. I think that is what Tech Ingredients did.
"And number one bromance of the show" had be start laughing in like the first 4 seconds. That's a record. Thanks Steve.
I like these interviews with technical experts, that give their views on topics like this
My go-to paste is from a different German company. Arctic MX-4. Order it in a 20g tube and it's a very good value for good paste.
I will try the expensive Grizzly Kryonaut Extreme soon. My go to would be Artic MX-4 too or Noctua.
@@thefirstloser In my opinion Kryonaut is great for CPUs with a heatspreader but has too much pump-out effect for anything direct-die. Pump-out is why I don't use MX-6 even though newer should be better.
@@chronicalcultivation Good to know, thanks.
two legends of the scene. hell yes.
Love it when Roman do piece discussion with you, guys! Wish it was a more common thing! But, well, distance is a biatch!
Now, on about the testing:
On die application Pump-out effect is the biggest issue and the one that should be focused the most when testing pastes.
A lot of "high" end pastes perform between a few degrees of each other when applied on a IHS, but, when it comes to die application, few passes the test after some thermal cicles because of pump out effect.
Id love to see you guys do a huge round up on this subject with the most popular pastes.
A very thorough test of thermal pasts would probably take a couple years and a super consistent way of applying it. Testing by applying to a system and doing a heat tests immediately and again some time down the road. The trick is making a lengthy lab test into content.
maybe in 10 years LTT labs will have finally started to do testing!
We need more 8auer and GN Tech Talks and Rants!
I have been waiting the entire Computex for this!!! BEST BROMANCE ON UA-cam!
Thermal conductivity outside of marketing hyperbole is a study I've taken to the extreme for composite jet engine parts; I've often thought about adapting my process for the ultimate heat sink for a CPU, but continue to run into practicality hindrances such that performance would be so far beyond what is currently accepted as 'high' that it would almost be a waste of materials. That, and as of the present, it couldn't be made for any price the typical PC enthusiast would call acceptable. Being designed to withstand extreme temperatures in constant use for a given lifecycle at 3000*C, for a PC part it would essentially be an infinite shelf life with no discernable wear. The flip side is that I couldn't manufacture it for less than $200 without going to industrial scale, which is not the direction I have in mind for the company or this particular process.
Look i think $200 for an ultra performance item - even if i don't need it and probably nobody "needs" it, seems outright cheap. There's a few of those people out there, someone is buying up all the 4090s don't they - some of these people actually put the performance to use in one manner or another; but others don't really do that and just buy because they want the absolute best thing no matter the cost. I think you can mark it up to a grand and give it a shot, i wish you good luck. There are all manner of mildly outrageous things of extremely niche use.
@@SianaGearz I appreciate the spurning and have thought about doing a limited run kind of thing. Now if you can master the marketing for me... 😁
@@C-M-E I wouldn't really trust myself to do that anywhere near competently. I had an introduction to marketing in school (2 weeks) so uhhh better than nothing i guess?
I love the collaboration You have with all the other UA-camrs! It's always a treat to see you get together for a video!
It's been so long I don't even remember what paste is on my CPU.
It's a 4790k purchased when Skylake came out and I refused to pay the memory premium.
It's been spending its life at 4.4 with a vcore offset in the .9s.
With a CM 212 on it.
Even the thought of having to worry about reapplying thermal paste on CPU during it's normal lifecycle is enough for me to never purchase a paste that has that as a concern.
Thermal conductivity is just the material property, the inverse of thermal resistivity. The thickness (and area) of the material also has a huge impact on the amount of heat transfer you get, and the layer of thermal compound is typically very thin and spread wide over the IHS. Because of this, as long as the material is generally a thermally conductive material it will transfer the heat effectively; with how thin the layer is, there are diminishing returns by only changing the material. A material with a high thermal resistivity (low thermal conductivity) will of course inhibit the heat transfer, such as air gaps, contaminants, or just a bad material, like the plastic film layer you get on a new cooler that people sometimes forget to remove.
I think someone on youtube (possibly Linus?) performed some testing many years ago with other materials such as toothpaste, and they worked pretty much as well as some of the leading thermal compounds on the market, at least over a very short period of time (as you say, there are other factors relating to longevity).
So I completely agree, thermal conductivity basically doesn't matter. Just use the thermal compound that comes with the cooler unless you're going for overclock records.
The boundary layer between two different materials has a thermal impedance of its own. And this is why thermal conductivity doesn't matter, since it doesn't describe that boundary, just the conductivity through the bulk of the material, and the thermal impedance of the paste layer is an order of magnitude smaller typically than the boundaries to the both adjoining metals.
I can't believe it's taken me this long to notice that Der8auer has the gate address for Abydos tattooed on his arm.
Love you guys! :D
How good is Artic MX 6, cus the reviews are amazing on that paste? Long lasting, cheap, amazing cooling... it has it all. I hear the MX 5 was bad tho, so yeah. I should have gone with Noctua 2, but oh well. So far so good with the Artic one.
Remember graphite thermal pads , pepperidge farm remembers . They were the talk back in the day about how long lasting they were and they somehow stopped being mentioned all together .