Testing Another YouTuber's Thermal Compound!

[nicklmg]

Buy ASUS Crosshair IV

On Amazon (PAID LINK): https://geni.us/yWUw

On Newegg (PAID LINK): https://geni.us/5NQp

 

Buy Thermal Pads

On Amazon (PAID LINK): https://geni.us/OgsFmP

On Newegg (PAID LINK): https://geni.us/gG89WiG

 

Buy AMD CPUs

On Amazon (PAID LINK): https://geni.us/LQl0K

On Newegg (PAID LINK): https://geni.us/JWgwL

On B&H (PAID LINK): https://geni.us/4BFYb3N

 

Buy Noctua NT-H1 Thermal Paste

On Amazon (PAID LINK): https://geni.us/ouQyxKV

On Newegg (PAID LINK): https://geni.us/pYRc

On B&H (PAID LINK): https://geni.us/n0ULm

 

Buy Thermaltake Black Widow

On Newegg (PAID LINK): https://geni.us/8O1IdXf

[kelvinhall05]

Testing another YouTuber's "thermal compound"

 

 

[seon123]

2 minutes ago, kelvinhall05 said:

Testing another YouTuber's "thermal compound"

N O

[Lord Nicoll]

3 minutes ago, kelvinhall05 said:

Testing another YouTuber's "thermal compound"

 

 

Gee thanks, I always wondered how many people remembered that

 

[kelvinhall05]

4 minutes ago, Lord Nicoll said:

Gee thanks, I always wondered how many people remembered that

 

I think about it every day /s

 

 

I do think about it whenever people bring up thermal paste alternatives, lol.

[Lord Nicoll]

2 minutes ago, kelvinhall05 said:

I think about it every day /s

 

 

I do think about it whenever people bring up thermal paste alternatives, lol.

Some of them do actually work, but it's better to buy a fuck load of HY510 or GD900 and just have so much you'll never run out for a while, I bougth 4 31g tubes and haven't even used a whole one up. 

[kelvinhall05]

3 minutes ago, Lord Nicoll said:

Some of them do actually work, but it's better to buy a fuck load of HY510 or GD900 and just have so much you'll never run out for a while, I bougth 4 31g tubes and haven't even used a whole one up. 

I've been using the same 4g tube of Arctic MX-4 since July 9th, 2017.

 

EDIT: Still has some left.

[GaWran]

any chance for the cpu to be 'lapped' ?

[RejZoR]

Did Linus miss the memo, but FX-8150 really wasn't ever considered as "inferno". That was the FX-9590 title. Those 9590 chips could pump out up to 500W of heat if my memory serves me well and they came with AiO as standard.

[Computernaut]

This stuff could be good for attaching those little heatsinks to the non-GPU parts of a graphics card for when you have an LN2 pot on the actual die or if you were to attach a CPU cooler to it.

[Nystemy]

I myself would have used a heat sink. (A fairly cheap aluminium extrusion, quality doesn't technically matter in this case.)
A power resistor and 2 thermal couples to bond onto the side of said power resistor and onto the heat sink.
Apply the compound to the resistor's mounting side and screw it onto the heat sink.
Before placing some insulation around the resistor itself, to ensure that as much heat as possible is actually carried away by the heat sink and not radiated off the resistor itself.

With 2 thermal couples we can measure the temperature difference between the resistor and the heat sink. (A smaller amount of temperature difference means a better thermal interface, at the same amount of power that is.)

Here just to make life "simple", we can laser cut or engrave an 0.5-1mm thick acrylic sheet to make as a paste guide that we place between the resistor and heat sink.
This gives us a known paste thickness, arguably gigantically thick compared to what most people usually go for. But we want to performance test the compound, so more thickness, means a greater temperature difference. Though, here I would need to give Linus a big complaint. Since the thermal compound thickness weren't controlled during testing. And a difference between 0.08mm and 0.09 might not seem huge, but that is 12.5% a fairly measurable difference. So tightening the screws more or less can alter the results. (And for everyone thinking 0.08mm is thick, a typical sheet of paper is 0.1mm)

If we also know the surface area of the interface, then we can based on known input power, and delta temperature, we can calculate the watts per meter Kelvin value of said paste. The unit stands for "an interface that is 1 square meter in area, 1 meter thick, at 1 watt will have X delta C.

So the formula is simply: Constant × Delta C × Area / Thickness = Watts

Using a CPU might seem more logical though on the surface, but it has a slew of issues.
Controlling power consumption of a CPU, it varies not just depending on clock speed, but also application behavior, what instructions are being called, cache misses, thread switching order, etc. And non of this you have control over unless you roll your own OS and study x86 software developer manuals for the specific CPU you are targeting. (resistors and a power supply is cheaper and far quicker and gives you tons of control)
Though, in most cases, one isn't even in control of how many threads a CPU is currently running, what stuff the OS has decided to pre load into RAM, etc. Or what other background tasks are fiddling away in the background.

Though, its easier please the non engineering viewer with a look at a "real world" example. Even if thermal testing of heat sink solutions practically don't use actual CPUs since it makes little sense in practice.

Steve over at gamer's nexus for an example has a more proper thermal testing setup for performance testing both heat sinks and thermal compound.

Though, since this video is about evaluating just the thermal compound, then emulating the localization of heat under the heat spreader on multi die CPUs isn't really needed to be fair. (Since a better compound will be better regardless. But for testing a cooler, it can have a bigger impact due to heat pipe placement and such.)

In the end. The main things done wrong here are:

1. No real control of power going through the thermal interface material. Nor even actually measuring it during the tests.
2. No control of thermal compound thickness. (This will impact the performance of the interface.)

So the results are largely inconclusive due to poor testing practices.
Its still a bit indicative, if done repeatedly for a few test runs. (Also a note back to GN, they do multiple test runs, despite having likely the best thermal test setup among reviewers, because there is still room for errors with simple things as slight mounting differences.)

[Nystemy]

On 6/8/2020 at 10:17 PM, Computernaut said:

This stuff could be good for attaching those little heatsinks to the non-GPU parts of a graphics card for when you have an LN2 pot on the actual die or if you were to attach a CPU cooler to it.

Small secret with LN2 overclocking of graphics cards.
They put heaters on the back of the cards to actively heat the memory, since it freezes over otherwise.
More cooling isn't really needed for them.

Same thing goes for system memory in a lot of CPU overclocking scenarios.

[RejZoR]

Btw, the CPU could still be lapped to remove the epoxy... Maybe video material for the next time lol.

[SolarNova]

2 hours ago, nicklmg said:

~SNIP~

Would have to put this on the 'fail' pile of vids im afraid.

 

Should have tested vs other thermal epoxy.

Should have used a CPU which can get up to the 80's C mark.

Should have used a Delidded CPU so u dont have to worry about loosing the CPU.

Should have mixed a larger amount of compound to ensure a better mix and apply it all at the same time.

 

This seemed , rushed. its obvious u meant well trying to find a 'hot' CPU, but when u realized u could only go up to 70c, u should have altered the plan accordingly.

 

It would seem the bigger your channel / company is getting, the less ur ability to adapt to situations like this.

 

[DavidKorb]

@nicklmg I can’t seem to find the video on floatplane, does it have a different title over there? 

[Nystemy]

On 6/8/2020 at 10:33 PM, SolarNova said:

Would have to put this on the 'fail' pile of vids im afraid.

 

Should have tested vs other thermal epoxy.

Should have used a CPU which can get up to the 80's C mark.

Should have used a Delidded CPU so u dont have to worry about loosing the CPU.

Should have mixed a larger amount of compound to ensure a better mix and apply it all at the same time.

 

This seemed , rushed. its obvious u meant well trying to find a 'hot' CPU, but when u realized u could only go up to 70c, u should have altered the plan accordingly.

 

It would seem the bigger your channel / company is getting, the less ur ability to adapt to situations like this.

 

That this largely is a failure is indeed true.

Though, as I already outlined in the book of a post I made earlier, one should preferably not use a CPU, but rather a resistor and a power supply, and some thermal couples.
And one should also have control over thermal paste thickness. And for performance testing thermal compounds, a thicker layer is technically beneficial, since then the thermal gradient caused by the compound is larger and thereby easier to measure and quantify. Not to mention that its easier to reproduce.

So there is a slew of ways of improve the testing done to actually make it useful.
Currently its more anecdotal than anything to be fair.

[SolarNova]

@nicklmg @LinusTech

 

Since i dont use youtubes comments section I want to address this here.

 

In reply to a comment you stated

"We don't hate the product. People asked us to test this on a CPU and we did. For obvious reasons there would be no controlled way for us to test two different thermal epoxies on the same cpu. - LS"

 

As i mentioned in my previous post. you can in fact use the same CPU. Just use a delidded CPU, im sure u can find a spare IHS from somewhere. Your then swapping out the IHS , not the Die itself which is whats producing the heat. You should have very little to no variance in CPU temp from the LM-TIM between die and IHS, so ur sorted.

[Nystemy]

On 6/8/2020 at 11:28 PM, SolarNova said:

@nicklmg @LinusTech

 

Since i dont use youtubes comments section I want to address this here.

 

In reply to a comment you stated

"We don't hate the product. People asked us to test this on a CPU and we did. For obvious reasons there would be no controlled way for us to test two different thermal epoxies on the same cpu. - LS"

 

As i mentioned in my previous post. you can in fact use the same CPU. Just use a delidded CPU, im sure u can find a spare IHS from somewhere. Your then swapping out the IHS , not the Die itself which is whats producing the heat. You should have very little to no variance in CPU temp from the LM-TIM between die and IHS, so ur sorted.

No reason to use an integrated heat spreader.

To performance test thermal paste all one needs is:

1. A heat source with known power that primarilly only releases its thermal energy through our interface of interest.
2. A known thermal interface thickness.
3. A heat sink to get rid of the heat on the other side of the interface.
4. Some thermocouples will work fine.

That the heat comes from a CPU, GPU, or a power resistor doesn't matter.
As long as we have control over the amount of power flowing through our thermal compound.

Even the area we spread the thermal paste over doesn't matter, as long as the area remains the same for all tested thermal compounds.
Though, larger is typically easier to reproduce, same goes for thickness, thicker is also typically easier to reproduce.

I would advice against using actual PC hardware for performance testing thermal paste, since there is frankly too many variables that influence power consumption for it to be a usable test.

In short, @AlexTheGreatish could CNC a 1mm acrylic sheet for thickness and area control of the thermal paste. So that those two variables remains constant for all tested compounds.
A heatsink can be an aluminium extrusion and a fan. (With a pocket drilled/milled for a thermocouple.)
And the heat source can be a box with a 25 watt power resistor surrounded by some insulation to ensure it isn't radiating away heat to its surroundings but rather sends it through our thermal paste. The resistor will also need a thermocoupled attached to it, so that we can measure the temperature on both sides of the thermal paste. This can though be thermally glued onto the side of the resistor package rather easily. _{(25 watts might sound pathetic, but with 1 mm of thermal paste, that suddenly looks like 250 watts or more if we were to use a more realistic thermal paste application. But here thickness makes the measurement easier, and it doesn't effect the numbers in the end, so technically, could go to 2-5 mm if one feels like it, but 1 mm should be sufficient.)}

This would only take around an hour to build. (Though, filming it will likely double that.. Not to mention that one will likely need more than 1 jig due to the thermal compound of interest being a thermal adhesive....)

If one feels like converting the measured numbers into W/mK then the formula is:
Watts × Thickness / Area / Delta C = W/mK
 

[nerdoholic_n8c]

I happen to have used the Arctiv Silver product myself AND watched TEs video months ago.
 

I will just throw some stuff in here since I doubt it will reach any useful "like ratio" on YT, and maybe clear things up for some folks.
 

(1) thermal paste vs thermal epoxy
~5,3 - 12,5 W/mK = thermal paste
> www.thermal-grizzly.com/en/products/16-kryonaut-en
> www.gamersnexus.net/guides/2137-thermalpaste-types-conductivity-and-more
 

7,5 W/mK = Arctic Silver thermal adhesive
> www.conrad.com/p/arctic-silver-asta-thermal-adhesive-75-wmk-7-g-max-temperature-150-c-150367
 

(2) RTFM
You didn't watch the TE video (9:21) and you didn't even read the fucking manual (8:52) that comes with it.
Shame. Shame. Shame. ^^
If you don't have the time to watch it, sure, I get that, but that's what you have staff for.
 

(3) Mc Donalds knife for mixing
And then on that little mass.
Like, really.
You aren't moving 75% of the stuff since it's just clinging to the plastic...
 

(4) "We don't hate the product. People asked us to test this on a CPU and we did."
Just read this here.
I don't need you to be on Steve's level, that's totally fine with me.
But this was 70% Verge.

Feels like you half-assed this video from a comment recommendation without doing as much as 10 minutes of research.
 

(5) Heating the mixture during the curing process
Wanted to check if room-temperature curing epoxy becomes faulty when exposed to heat during curing - just seems to speed up the process though!
 

"However, adding heat to the epoxy systems that normally cure at room temperature, will accelerate the rate of cure and boost the performance profile."

"Q: Are there specific applications where adding heat during or after cure is advantageous?

A: In nearly every application the answer would be yes, particularly in aerospace, where the requirements are so rigorous and the qualification process is so demanding. One can easily choose a two part room temperature curing product, but it's crucial for the engineer to know that adding heat will result in a better system."

> https://www.masterbond.com/q-and-a/how-properly-cure-epoxy-aerospace-industry
 

Doesn't seem to be a blank "just heat it, bruh" though, there are limits depending on the material.
But seems like that is at way higher temperatures than a CPU will (hopefully) ever reach
> http://rcsailing.net/forum/index.php?threads/post-curing-of-epoxy.5519/
 

Don't take this as zealotic "must prove wrong", I am simply not a chemist and wanted to know. There you go, we both learned something!
 

(6) "What is it used for?"
https://www.youtube.com/watch?v=WJnrMNCahxc#t=13m5s
Felt like a real world example might be useful here.
 

E.g. I used the Arctiv Silver for modding my GPU with a Morpheus II, where the AS solved my issue of "how the fuck do I get the VRAM heatsinks to stick??" (copper mind you, and the aluminium ones already were close to falling off with pads...).
 

That's the goddamn great thing about this.
It is an ADHESIVE with GREAT thermal attributes!
 

I remember semi-raging while googling for this shit back in September when I modded my 5700XT.
Nothing useful came up, only some 1-2 W/mK shit like this https://www.silverbead.de/Silverbead-Waermeleitkleber-Thermal-Glue-Adhesive-fuer-Heatsinks-LED-VRAM-VRM-CPU-GPU
Guess how happy I was when I stumbled upon the AS? (hey! Yeah, you! With the dirty mind! I see you!)
 

- - - - - - -
 

To sum up:
If you make a 10m video out of a recommendation, please at least have someone watch/read and _fully understand_ what is presented in that recommendation, so Linus can get a proper summary to jump around with in front of the camera.
 

And yes, I clearly care for TE, they are a fucking bombastic amazing channel.
SUBSCRIBE NOW OR I WILL CALL THE POLICE; AND SMASH THE LIKE BUTTON (on ALL their videos) FOR THE YOUTUBE ALGORITHM!
> https://www.youtube.com/user/TechIngredients

Edited June 8, 2020 by nerdoholic_n8c
linked LTT comment on YT

[LinusTech]

1 hour ago, nerdoholic_n8c said:

I happen to have used the Arctiv Silver product myself AND watched TEs video months ago.
 

I will just throw some stuff in here since I doubt it will reach any useful "like ratio" on YT, and maybe clear things up for some folks.
 

(1) thermal paste vs thermal epoxy
~5,3 - 12,5 W/mK = thermal paste
> www.thermal-grizzly.com/en/products/16-kryonaut-en
> www.gamersnexus.net/guides/2137-thermalpaste-types-conductivity-and-more
 

7,5 W/mK = Arctic Silver thermal adhesive
> www.conrad.com/p/arctic-silver-asta-thermal-adhesive-75-wmk-7-g-max-temperature-150-c-150367
 

(2) RTFM
You didn't watch the TE video (9:21) and you didn't even read the fucking manual (8:52) that comes with it.
Shame. Shame. Shame. ^^
If you don't have the time to watch it, sure, I get that, but that's what you have staff for.
 

(3) Mc Donalds knife for mixing
And then on that little mass.
Like, really.
You aren't moving 75% of the stuff since it's just clinging to the plastic...
 

(4) "We don't hate the product. People asked us to test this on a CPU and we did."
Just read this here.
I don't need you to be on Steve's level, that's totally fine with me.
But this was 70% Verge.

Feels like you half-assed this video from a comment recommendation without doing as much as 10 minutes of research.
 

(5) Heating the mixture during the curing process
Wanted to check if room-temperature curing epoxy becomes faulty when exposed to heat during curing - just seems to speed up the process though!
 

"However, adding heat to the epoxy systems that normally cure at room temperature, will accelerate the rate of cure and boost the performance profile."

"Q: Are there specific applications where adding heat during or after cure is advantageous?

A: In nearly every application the answer would be yes, particularly in aerospace, where the requirements are so rigorous and the qualification process is so demanding. One can easily choose a two part room temperature curing product, but it's crucial for the engineer to know that adding heat will result in a better system."

> https://www.masterbond.com/q-and-a/how-properly-cure-epoxy-aerospace-industry
 

Doesn't seem to be a blank "just heat it, bruh" though, there are limits depending on the material.
But seems like that is at way higher temperatures than a CPU will (hopefully) ever reach
> http://rcsailing.net/forum/index.php?threads/post-curing-of-epoxy.5519/
 

Don't take this as zealotic "must prove wrong", I am simply not a chemist and wanted to know. There you go, we both learned something!
 

(6) "What is it used for?"
https://www.youtube.com/watch?v=WJnrMNCahxc#t=13m5s
Felt like a real world example might be useful here.
 

E.g. I used the Arctiv Silver for modding my GPU with a Morpheus II, where the AS solved my issue of "how the fuck do I get the VRAM heatsinks to stick??" (copper mind you, and the aluminium ones already were close to falling off with pads...).
 

That's the goddamn great thing about this.
It is an ADHESIVE with GREAT thermal attributes!
 

I remember semi-raging while googling for this shit back in September when I modded my 5700XT.
Nothing useful came up, only some 1-2 W/mK shit like this https://www.silverbead.de/Silverbead-Waermeleitkleber-Thermal-Glue-Adhesive-fuer-Heatsinks-LED-VRAM-VRM-CPU-GPU
Guess how happy I was when I stumbled upon the AS? (hey! Yeah, you! With the dirty mind! I see you!)
 

- - - - - - -
 

To sum up:
If you make a 10m video out of a recommendation, please at least have someone watch/read and _fully understand_ what is presented in that recommendation, so Linus can get a proper summary to jump around with in front of the camera.
 

And yes, I clearly care for TE, they are a fucking bombastic amazing channel.
SUBSCRIBE NOW OR I WILL CALL THE POLICE; AND SMASH THE LIKE BUTTON (on ALL their videos) FOR THE YOUTUBE ALGORITHM!
> https://www.youtube.com/user/TechIngredients

You know Tech Ingredients thinks we tested it okay, right?

 

 

 

 

[Warin]

2 hours ago, LinusTech said:

You know Tech Ingredients thinks we tested it okay, right?

 

 

 

 

Cranky fan boys pay attention to the comments? Not on my internet!

I don't think you will need to apologize for anything.  This time.

[xnamkcor]

So, Linus was asked to test this on a CPU, and people are complaining he didn't do an artificial benchmark using a heat source?

[poochyena]

Am I understanding this right? They are selling a product they didn't even test themselves?

[SolarNova]

9 minutes ago, poochyena said:

Am I understanding this right? They are selling a product they didn't even test themselves?

Your not getting this right.

They did test. Its a good Thermal Adhesive compound.

 

 

[lewdicrous]

1 hour ago, poochyena said:

Am I understanding this right? They are selling a product they didn't even test themselves?

Here's their video if you want to watch it.

 

Tl;dr, They tested thermals, but they didn't use CPUs.