The Official Thermal Interface Material thread

[80-watt Hamster]

MX-6 on my rig, I doubt I change it out for 5+ years honestly (or whenever i upgrade fully). I will never understand peoples fascination with thermal paste. lol

It's mostly a thing with Overclockers and people who frequently work on PCs.

It's our version of motor oil.

 

[Shrek]

Arctic MX-4 vs MX-5 vs MX-6 Thermal Paste Review The technical data (ocinside.de)

The higher the viscosity the better the performance?

Glad they don't play the thermal conductivity game.

 

[lilhasselhoffer]

Facinating is Der8auer's observation that the finer one grinds the base metal/oxide, the worse the thermal conductivity; makes me realize that surfaces play a very important role.

Help me understand...because I don't read this the same way.

What I heard was that if you start with a solid piece of metal you've got one thermal conductivity. If you then break the thing up into small bits, and turn it into a paste, the silicon oil used to put it into a semi-solid suspension dramatically decreases the thermal conductivity. This makes sense, as you are suspending a good thermal conductor inside a matrix of both terrible and mediocre thermal conductors (air bubbles and oil).

Theoretically, assuming you have a functionally infinitely fine conductor and dramatically decrease the amount of oil (think sintered component) you'd come closer to the thermal conductivity of a solid piece. The practical issue is that when you create a finer metal powder you'll likely do a number of bad things. Metals oxidize in an oxygen rich atmosphere...so you no longer have copper but cupric or cuprous oxide. Your air will be trapped between the oil and metal...functionally making it an insanely good insulator. Etc...


For my money, and if you have the equipment, you should instead try something fun. Diamond is quite good at conducting heat. A fine powder of it is generally thought of as industrial waste, so it's surprisingly cheap. The hillbilly/ghetto mod high end thermal paste several year ago was to take dielectric grease, mix with a diamond powder, heat into a semi-liquid state, and place under a vacuum until the air bubbles suspended in it come out of solution. The resulting paste was about as good as you could get, the dielectric grease is basically incapable of hardening, and the result was leagues better than whatever solid gunk Intel was pasting onto their coolers....though the shelf life sucked and the cost was very high if you couldn't repurpose some equipment.
Of note: diamond powder is a lung irritant and will cause damage if ingested. The paste in question had a shelf life of about a week before the diamond would come out of suspension if left in any quantity. The paste will damage the IHS and heat sink interface...that diamond is basically very fine sandpaper...so prepare to lap if you want to reinstall the thing. Finally when the dielectric does harden it's gummy...



I'd suggest that the take-away from most of this is that if you double the thermal conductivity you can double the thickness of the material without influencing heat transfer by conductivity. Engineering toolbox equations If you are looking at a processor it's pretty easy to double or triple your thickness with a bad design....think the new contactor brackets that are currently in-use for Intel stuff. It's silly to say this...but I think I'd suggest that lapping the IHS, lapping the contact plate, and using mediocre paste is likely infinitely more reasonable for most. It's definitely easier and more fool resistant than plowing huge amounts of effort into doubling or tripling thermal conductivity through chemistry and trying to make-up for mechanical interface issues. My two cents is that I've never seen enough difference to justify premium options...instead of using moderate options and good preparation. Lapping sucks...but I've never had such a bad result that I felt I needed premium options.
All of this said, I also don't overclock to the moon. I'd prefer to not hit thermal throttling limits, I'm fine with underclocking in most instances, and usually by the time my tech sucks enough to not be viable it's time that the middle level replacement is so good in comparison that I'm satisfied with a slight underclock out of the box to make room for stock settings being an upgrade in the future.

This is all coming from an idiot who bought into X79. The upgrade from there to a middle tier 3700 was huge...and the RAM fixed 5700x that replaced it was transcendent with M2 and stable clocks that didn't crap the bed with multi-thread... and the platform cost was less than the CPU cost without adjusting the x79 upward for inflation. All of this leads me to not want to plow huge money into anything...because barring price gouging it's better to buy middle tier and replace every few generations rather than buy high tier, invest in everything from cooling to maintenance, and do all of this to overclock a fraction of a GHz whereas generational improvements mean the same performance is always had for less (barring the current Nvidia shenanigans, or AMD's marketing bizarro moves).

 

[Shrek]

Theoretically, assuming you have a functionally infinitely fine conductor and dramatically decrease the amount of oil (think sintered component) you'd come closer to the thermal conductivity of a solid piece.

This reasoning is based on the assumption that the soild/oil interface has no thermal resistance, I don't think this is the case; as the surface area increases so also does the thermal resistance (I don't think air bubbles play a role). Typically themal greases use metal oxides, not metals Thermal paste - Wikipedia

Der8auer Rants About Misleading W/mK Marketing (Thermal Conductivity) - YouTube
6:20
"If you increase the particle size then you will also increase thermal conductivity."
5:20
Concering IC diamond
"and then they perform the same as..."


Here is some rough and ready reasoning: in the densest sphere packing, 26% is empty space (in this case oil) and this is regardless of the sphere size; so one might expect the thermal conductivity not to depend on grain size... unless surface effects were significant; bigger grains lead to less overall surface area.

 

[lexluthermiester]

Glad they don't play the thermal conductivity game.

Agreed! Real world performance is what counts.

 

[nomdeplume]

Here is some rough and ready reasoning: in the densest sphere packing, 26% is empty space (in this case oil) and this is regardless of the sphere size; so one might expect the thermal conductivity not to depend on grain size... unless surface effects were significant; bigger grains lead to less overall surface area.

It would appear he is addressing the physical structure. As in roughly broken shards being inferior to more accurately cut one that has improved upon or impacted less the native properties of the unfinished material.

 

[Waldorf]

@lilhasselhoffer
not everyone is looking for 10GHz OC.
if i spend money on hw above mid grade, i dont like the chances of bottle necking anything, incl cooling/heat transfer.
and once you run stuff like ryzen 5xxx/7xxx, with stuff running at higher temps (continuously), i wont save 5$ and use toothpaste grade crap.

and not sure if the diamond stuff was more of a DIY thing, as i remember using diamond (powder) based paste over 10y,
purchased from MC (not some rare "one location" shop on another continent), with decent results.
just not good enough to live with the chance of "sanding" the HS (print gone) etc, especially when working on other ppls stuff.

 

[maxfly]

I use mx-4 on client builds exclusively, for no other reason than its friggin bulletproof ($10, 20g tube ftw!). Apply it and forget it. Prior to that it was AS Ceramique for the same reason. Stuff never degraded, altho if the tube sat too long it would separate.

I use either mx-6 or Prolimatech pk-3 for my own rigs (usually the latter) but pk3 is getting expensive and hard af to find consistently. I've got at least 6 tubes of other stuff (mx-5, TG, TF7 etc)that I tried but they just weren't as all-around consistent as pk3. The only paste I will never buy again is Thermal Grizzly. It generally works fine (sometimes it hardens up in the tube)but the mark up is just stupid. I honestly don't know how they stay in business.

 

[Waldorf]

@maxfly
ahh, those were the days..
i remember back in early 2000s i first started replacing parts on my desktop, best was AS 5.

but around 2010 i went back to having a (gaming) desktop instead of laptop,
Arctic and shin-etzu stuff just "looked" better and seemed to be more "up to date".
and having used lots of arctic products for previous builds, i decided to switch.

would still prefer LM on my rig, but tired of the staining, and im not swapping parts in regular intervals anymore,
and im done pulling cpus out of its (locked) socket..

 

[MachineLearning]

Blob, sausage or solid? Apply thermal paste to the GPU correctly - we measured! | Reminder | Page 3 | igor´sLAB

Since people keep asking: Today’s test will once again show in measurement series how to properly apply a purpose-built thermal paste on large GPUs. The selection of a suitable paste and the…

www.igorslab.de

 

[80-watt Hamster]

@maxfly
ahh, those were the days..
i remember back in early 2000s i first started replacing parts on my desktop, best was AS 5.

but around 2010 i went back to having a (gaming) desktop instead of laptop,
Arctic and shin-etzu stuff just "looked" better and seemed to be more "up to date".
and having used lots of arctic products for previous builds, i decided to switch.

would still prefer LM on my rig, but tired of the staining, and im not swapping parts in regular intervals anymore,
and im done pulling cpus out of its (locked) socket..

I still like AS5. Its longevity remains top-tier to my knowledge.

 

[Bruizer]

Longtime lurker. Gotta show some love to GELID GC-Extreme. Been using it for years (switched from MX-4) and I find its still relevant (IMO). Great spread, temps, and longevity. Though a bit pricier. It's doing great on my 5950x (w/ mild PBO settings) in conjunction with the Thermalright FC140.

 

[Waldorf]

@80-watt Hamster
sure thing, except in last few years i was upgrading stuff so often, that for the last couple cpu swaps, the LM stuff hadnt even "cured"
but with the 5800x i decided for normal paste, LM was starting to get on my nerve, being "repelled" from the HS when applying, making proper coverage harder.

 

[80-watt Hamster]

@80-watt Hamster
sure thing, except in last few years i was upgrading stuff so often, that for the last couple cpu swaps, the LM stuff hadnt even "cured"
but with the 5800x i decided for normal paste, LM was starting to get on my nerve, being "repelled" from the HS when applying, making proper coverage harder.

I hear ya, and have had those phases myself. But also phases and/or machines that didn't have their HS removed for 5+ years. There actually isn't any AS5 in my kit right now. Between trying a couple of new things and the tubes that came with various HSFs, there hasn't really been occasion to buy any. I think I will next time that oppo arises, though.

 

[Shrek]

the LM stuff hadnt even "cured"

I had no idea that liquid metal hardened.

 

[chrcoluk]

I am probably happy enough using either latest MX or Noctua paste. Noctua paste comes with some nice paste removal wipes as well due to Noctua's policy of providing extras.

 

[Space Lynx]

I had no idea that liquid metal hardened.

it doesn't.

 

[Bones]

it doesn't.

No it doesn't harden, it just eats most everything around it up.

 

[Waldorf]

@Shrek
i might have used the wrong wording, but unless you apply and remove LM multiple times within a few month
(so the gallium can saturate HS and cooler), it will "harden/dry" (whatever you wanna call it),
making cooler removal much harder than with any paste i've ever used, and chances of pulling the cpu out of its (locked) socket much more likely.

and i didnt have any cooler that wasnt coming with a copper base (or nickel coated),
so im not talking about the "welding" happening with alu.

@Space Lynx
why? cause it never happened to you?!
(had a couple of times where others were present when i was removing cooler/block, observing "stuck" cooler...)

 

[lexluthermiester]

I had no idea that liquid metal hardened.

As a rule, it's not supposed to. However, if impurities are leeched out of the heatsink or IHS, it can happen. Chemistry/Metallurgy is funny like that. Remember, some metals are porous and can bleed/leech certain materials under certain circumstances. Liquid metals tend to be a sponge as well. The process usually takes a year or more to take effect. However, as long as good contact is made, heat transfer is not affected.

it doesn't.

Um, yes it can, it's just not common for most use-case-scenario's. I've only ever seen it once and it was a combination of the IHS and the heatsink having aluminium content which was leeched into the LM TIM and caused it to solidify. It was still easily removed, but the CPU IHS had to be removed and a new heatsink used. After going bare die, the system in question was back up and running without skipping a beat and could OC a bit better.

 

[Space Lynx]

As a rule, it's not supposed to. However, if impurities are leeched out of the heatsink or IHS, it can happen. Chemistry/Metallurgy is funny like that. Remember, some metals are porous and can bleed/leech certain materials under certain circumstances. Liquid metals tend to be a sponge as well. The process usually takes a year or more to take effect. However, as long as good contact is made, heat transfer is not affected.


Um, yes it can, it's just not common for most use-case-scenario's. I've only ever seen it once and it was a combination of the IHS and the heatsink having aluminium content which was leeched into the LM TIM and caused it to solidify. It was still easily removed, but the CPU IHS had to be removed and a new heatsink used. After going bare die, the system in question was back up and running without skipping a beat and could OC a bit better.

I was referring to people who use it normally, like a PS5 system for example. Open a PS5 in 40 years, and it will still be liquid in form. I'd bet money on it anyway, but yes if I knew we were talking about idiots who combine it with metals they shouldn't, my answer would have been different, lol

 

[Waldorf]

@Space Lynx
thats the problem. you never know who reads stuff

 

[Space Lynx]

@Space Lynx
thats the problem. you never know who reads stuff

this is very true, I have a bad habit of assuming sometimes. :/ I have been working on getting better at that though in recent months.

 

[lexluthermiester]

but yes if I knew we were talking about idiots who combine it with metals they shouldn't, my answer would have been different, lol

All things being fair, it's not always idiots doing this. The system in question was owned by someone who is very experienced, knows what they're doing and followed the directions given. They came to me because they wanted a second set of eyes on the problem. The problem in that situation was an unknown(and undocumented) level of metal content in the parts being used. It was just one of those things that happens even when the all precautions are taken.

 

[Waldorf]

@Space Lynx
lol, didnt mean you.

but really, unless i removed and reapplied it 2-3 times on new cooler/hs combo (at least with the coollaboratory LM i used),
it would "dry" and fuse the parts together.

and i can exclude that i got a bad batch (or similar), as it didnt happen after i repeated application multiple times within say 6-8 month,
with the hs absorbing some of it, and staying "liquid" for +1y before i sold the cpu.



but i think for the future it wont matter much anymore which of the "top" 10 TP to use,
when the trend continuous, with high heat in small areas and higher (overall) operating temps, like we see on Ryzen.