Best thermal paste

[xtremedetailz]

Thermal Grizzly Kryonaught for me.

 

[Rroff]

Apply as little possibly spread as thinly as possible unless dealing with very concave/vex then consider lapping to 800/1000 grit. Temps will be great.

I’ve just built a 64 core overclocked OEM EPYC system and the core temps are amazballz.

Unless you are experienced at lapping and achieving a perfect finish, or a large surface area core like EPYC (where spreading performs better than not) you shouldn't be hand spreading AS5 - especially on Intel CPUs you should use the process as proscribed in the application notes which is usually a thin line oriented to the die.

 

[jigger]

Unless you are experienced at lapping and achieving a perfect finish, or a large surface area core like EPYC (where spreading performs better than not) you shouldn't be hand spreading AS5 - especially on Intel CPUs you should use the process as proscribed in the application notes which is usually a thin line oriented to the die.

The aim is to use as little paste as possible while maintaining complete coverage. Spreading the pastes is the only method that ensures you achieve this and gives you a simple way to gauge the amount of paste. If the ISH and/or the cooler base are not well matched then temps will suffer.

The goal of lapping isn’t to achieve a perfect finish, just to make the surface flat. Mirror tiles, taped wet and dry sand paper 400, 600 and 800 rinsing inbetween.

 

[jeffspuds]

Thin grain of rice can't say any of products matter, the cooler heatsinks even stock are adequate.

 

[Rroff]

The aim is to use as little paste as possible while maintaining complete coverage. Spreading the pastes is the only method that ensures you achieve this and gives you a simple way to gauge the amount of paste. If the ISH and/or the cooler base are not well matched then temps will suffer.

The goal of lapping isn’t to achieve a perfect finish, just to make the surface flat. Mirror tiles, taped wet and dry sand paper 400, 600 and 800 rinsing inbetween.

You are using the wrong paste - other pastes like Thermal Grizzly will get better results with that approach, AS5 is designed primarily to fill imperfect spaces via pressure mount and thermal cycles ( http://www.arcticsilver.com/pdf/appmeth/int/vl/intel_app_method_vertical_line_v1.1.pdf ) and should not be manually spread unless using large surface area CPUs (or some CPUs which have a very high quality finish as standard). While the goal is to use as little paste as possible AS5 needs a certain minimum amount to work properly which will decrease with the flatness and quality of the surfaces involved.

Do NOT spread the line of thermal compound out.

When you place the heatsink on the top of the metal cap, the line of thermal compound will spread out like the pink oval pattern on top of the metal cap shown in photo QP4. Allowing the mounted heatsink to spread the thermal compound insures proper coverage, minimizes air bubbles and allows an optimum bond line between the two surfaces.

For your average Intel quality IHS, Intel die layout and average CPU cooler quality hand spreading AS5 is sub optimal. For other situations your mileage may vary.

Instructions if using something like an EPYC CPU: http://www.arcticsilver.com/PDF/appmeth/amd/ss/AMD_app_method_surface_spread_v1.2.pdf but you are better off using a different paste really - if you have a high quality finish it'll do OK but AS5 is designed primarily for not manually spreading.

Apply some thermal compound and nothing else to the corner of the die as shown
in photo QP2. The amount of thermal compound to apply depends on the surface
area of your die. For a small single core CPU use about 1-2 cubic millimeters
(1/4 of a bb or 1/2 uncooked grain of white rice) of thermal compound. For a
large single or multiple core CPU, use 4-5 cubic millimeters (2/3 of a bb or 1
uncooked grain of white rice) of thermal compound. Spread the thermal
compound over the die as shown in photo QP3 and QP4. The small amount of
thermal compound has been carefully spread over the top of the die using a single
edge razor blade. A razor blade or the clean edge of a credit card can be used as
an application tool. You may use whatever tool is available as long as it is
CLEAN and allows you to control the application area and thickness. The flatter
the mating surface of a die and heatsink, the thinner the layer of thermal
compound required. Stock processors and/or heatsinks with normal surface
irregularities will require a layer 0.003" to 0.005 thick as shown below to fill the
resultant gaps. (Equal to the thickness of about 1 sheet of standard weight paper.)
Properly lapped heatsinks with mirror finishes will only require a translucent
haze.

 

[jigger]

You are using the wrong paste - other pastes like Thermal Grizzly will get better results with that approach, AS5 is designed primarily to fill imperfect spaces via pressure mount and thermal cycles ( http://www.arcticsilver.com/pdf/appmeth/int/vl/intel_app_method_vertical_line_v1.1.pdf ) and should not be manually spread unless using large surface area CPUs (or some CPUs which have a very high quality finish as standard). While the goal is to use as little paste as possible AS5 needs a certain minimum amount to work properly which will decrease with the flatness and quality of the surfaces involved.



For your average Intel quality IHS, Intel die layout and average CPU cooler quality hand spreading AS5 is sub optimal. For other situations your mileage may vary.

Instructions if using something like an EPYC CPU: http://www.arcticsilver.com/PDF/appmeth/amd/ss/AMD_app_method_surface_spread_v1.2.pdf but you are better off using a different paste really - if you have a high quality finish it'll do OK but AS5 is designed primarily for not manually spreading.

I’m not, I’m using the best all round paste I’ve ever found.

 

[Dg834man]

You are using the wrong paste - other pastes like Thermal Grizzly will get better results with that approach, AS5 is designed primarily to fill imperfect spaces via pressure mount and thermal cycles ( http://www.arcticsilver.com/pdf/appmeth/int/vl/intel_app_method_vertical_line_v1.1.pdf ) and should not be manually spread unless using large surface area CPUs (or some CPUs which have a very high quality finish as standard). While the goal is to use as little paste as possible AS5 needs a certain minimum amount to work properly which will decrease with the flatness and quality of the surfaces involved.

Isn't as5 conductive? they will issue that guidance to minimise people getting paste everywhere to minimise shorting risk.

 

[Rroff]

Isn't as5 conductive? they will issue that guidance to minimise people getting paste everywhere to minimise shorting risk.

Slightly capacitive not particularly conductive, it shouldn't destructively short anything but it can cause signalling errors, etc. if you bridge data lines, etc. with it.

AS5 is primarily designed to get best results when it is spread by pressure and a few thermal cycles, in some cases it isn't possible for that to happen effectively such as large surface area CPUs and/or where there are multiple distributed packages beneath the heat spreader, etc. in which case manually spreading will produce better results - but in that situation you are usually better off using a paste designed for manual spreading.

I'm a fan of it because it is a great all around paste, but I'd be surprised if jigger couldn't get better results using pastes designed for spreading in some of the application situations they've talked about and I certainly don't recommend hand spreading it with the average quality of Intel CPUs and the average mainstream heat sinks.

 

[Mr Oxide]

I bought the 10g version of Noctua NT-H2 and has been solid. I do think the 3.5g is best for anybody that building a rig or 2 and wants some spare.

If IIRC there was a 1c difference temp wise between Noctua NT-H2 and Grizzly and Grizzly just come out on top. I only seen this months later and am happy at trying Noctua.

I'm guessiing that 1c could differ from build to build and user

 

[Rroff]

I bought the 10g version of Noctua NT-H2 and has been solid. I do think the 3.5g is best for anybody that building a rig or 2 and wants some spare.

If IIRC there was a 1c difference temp wise between Noctua NT-H2 and Grizzly and Grizzly just come out on top. I only seen this months later and am happy at trying Noctua.

I'm guessiing that 1c could differ from build to build and user

I've not used TG in awhile so don't know if it is any different now, I know they've made some changes, but I found it gave slightly better performance at the expense of longevity when I used it on some builds. That was some years back though.

These days I have AS5, MX-4 and MX-6 in my PC parts - tend to use AS5 on a lot of my own stuff, MX-4 if I'm building for someone else just to avoid any arguments, etc. and MX-6 on anything like the Intel 14th gen.

 

[jigger]

A good quality paste is around 10 W/mk so a 1 metre cube of TIM can transfer 10watts~ of heat from any face to opposite. Copper has a thermal conductivity of around 400watts~ silver, 430~ and air about 0.03 IIRC. The very best paste is maybe 15.

So what you want for the “best” thermal performance and what matters more than anything by orders of magnitude, is how close you can get the copper heat spreader of the chip to base of the cooler. In a perfect world the amount of TIM is zero, the materials are inert and faces perfectly flat. In that world a metre cube of silver could passively cool even an Intel i9!

So how do we get good thermal performance? We evenly spread as little interface material as possible, avoid air and use as much mounting pressure as possible. If you lap both surfaces flat all the paste is required to do is to fill the micron size scratches with something better than air that is compressive enough to allow the surfaces to be pressed together held under tension.

All interface materials are terrible by comparison, and one being slightly less terrible is of little consequence to other factors. IMO the best TIM is the one that gets you closest to zero gap with as little waste as possible, in a reasonable amount of time for a reasonable price.

 

[Rroff]

We evenly spread as little interface material as possible

That is the thing - unless you have a high quality finish and perfectly flat surfaces, as the documentation mentions, spreading by hand with AS5 can result in too little or too much in any one place. Most people aren't going to be lapping and so on so shouldn't be manually spreading AS5 and in a situation where you should or are manually spreading AS5 you will almost certainly get better results, hence the results in a lot of flawed benchmarks where they persist with manually spreading it inappropriately then rating it as mediocre, with another paste which has been designed for it.

It is a paste which is optimal when dealing with an imperfect situation, as in your average Intel IHS, and can be beaten by other pastes under more ideal circumstances.

EDIT: End of the day if it works for you, and it sometimes seems more of an art than a science with thermal paste, then great, but I definitely don't recommend most people manually spread AS5 and have seen a lot of situations where people are getting sub-optimal performance or longevity by doing so - there was a poster on here who never seemed to make the connection between complaining they were having to re-apply it every couple of years and my criticism of their application method :s yet still persisting with using it and using it wrong.

 

[Grim5]

Anyone better than the others ?

Liquid Metal is the best hands down for performance , for example: Conductonaut

Next best is one of the new phase change thermal pads which have been shown to be better than traditional paste and last longer, for example: PTM7950

Then all the way down the ladder of performance is traditional thermal paste, has lower performance than the first two options, but tends to also be the cheapest and lasts a decent time but can dry out with high heat applications, for example: Kryonaut

 

[Rroff]

Liquid Metal is the best hands down for performance , for example: Conductonaut

Next best is one of the new thermal pads which have been shown to be better than traditional paste and last longer, for example: PTM7950

Then all the way down the ladder of performance is traditional thermal paste, has lower performance than the first two options, but tends to also be the cheapest and lasts a decent time but can dry out with high heat applications, for example: Kryonaut

Liquid metal has far too many potential downsides really unless min-maxing a setup, phase change materials can be tricky to get the best results from - although you can beat thermal paste, you can also hinder cooling performance significantly with them, or just not do any better than thermal paste.

 

[Grim5]

Liquid metal has far too many potential downsides really unless min-maxing a setup, phase change materials can be tricky to get the best results from - although you can beat thermal paste, you can also hinder cooling performance significantly with them, or just not do any better than thermal paste.

I'm gonna try PTM7950 on my PC next time I make changes, currently my cpu and GPU are using Krypnaut thermal paste. I won't use Liquid Metal even though it has the best temps, but I will try PTM7950

What would make me try Liquid Metal is if it came with some sort of chamber that could seal and hold the liquid so it can never escape, like the PlayStation 5 has, my PS5 has Liquid Metal and it's 4 years old and still runs well

 

[cosmogenesis]

Thanks for all the info

 

[jigger]

Liquid Metal is the best hands down for performance , for example: Conductonaut

Next best is one of the new phase change thermal pads which have been shown to be better than traditional paste and last longer, for example: PTM7950

Then all the way down the ladder of performance is traditional thermal paste, has lower performance than the first two options, but tends to also be the cheapest and lasts a decent time but can dry out with high heat applications, for example: Kryonaut

Nasty stuff. Impossible to remove completely and bites into metals leaving pockets for air. Also a royal PIA to apply.

I wouldn’t class gallium based TIM as a paste and I’m also dubious of some of the claimed thermal conductivity claims.

 

[goldy6660]

Another vote for thermal grizzly

 

[jigger]

That is the thing - unless you have a high quality finish and perfectly flat surfaces, as the documentation mentions, spreading by hand with AS5 can result in too little or too much in any one place. Most people aren't going to be lapping and so on so shouldn't be manually spreading AS5 and in a situation where you should or are manually spreading AS5 you will almost certainly get better results, hence the results in a lot of flawed benchmarks where they persist with manually spreading it inappropriately then rating it as mediocre, with another paste which has been designed for it.

It is a paste which is optimal when dealing with an imperfect situation, as in your average Intel IHS, and can be beaten by other pastes under more ideal circumstances.

EDIT: End of the day if it works for you, and it sometimes seems more of an art than a science with thermal paste, then great, but I definitely don't recommend most people manually spread AS5 and have seen a lot of situations where people are getting sub-optimal performance or longevity by doing so - there was a poster on here who never seemed to make the connection between complaining they were having to re-apply it every couple of years and my criticism of their application method :s yet still persisting with using it and using it wrong.

How you spread the paste isn’t really relevant, it’s all about getting as close as possible with as little as possible. Using an arbitrary method like a “pea size amount” amount for eg, then fitting a cooler isn’t as accurate as visually spreading paste across the surface and adding/removing compound as required IMO/experience.

 

[Rroff]

How you spread the paste isn’t really relevant, it’s all about getting as close as possible with as little as possible. Using an arbitrary method like a “pea size amount” amount for eg, then fitting a cooler isn’t as accurate as visually spreading paste across the surface and adding/removing compound as required IMO/experience.

AS5 is engineered so that as long as you put a reasonable amount in a blob or line it will structure itself under mounting pressure and thermal cycles to achieve optimal coverage, or as close as possible, with imperfect surfaces - hand spreading works against the way this paste works and risks having too much or too little where needed or not. On larger surfaces and/or with complex layouts below the IHS it isn't able to properly structure itself and you will need to manually spread it as a second best option - in which case another paste is a better option. On very flat surfaces with a high quality finish it matters less as as per the documentation all you need is a light haze of paste in those circumstances. It is all documented in the product datasheet/notes and promotional material.