Linked by Thom Holwerda on Mon 7th Aug 2017 20:14 UTC
Hardware, Embedded Systems

Several years ago, we published a round-up of thermal pastes that started with Thermal Paste Comparison, Part One: Applying Grease And More and concluded with Thermal Paste Comparison, Part Two: 39 Products Get Tested. Since it's so hot outside (at least in our U.S. labs), we're trying to cool so many new CPUs and GPUs, and readers keep asking for it, we decided to combine and update those stories, adding a range of new thermal pastes and pads.

Thermal paste and how to apply it are probably more divisive than anything else in technology. So many different methods, old wives' tales, folklore, and god knows what else.

Order by: Score:
Comment by grandmasterphp
by grandmasterphp on Mon 7th Aug 2017 22:27 UTC
grandmasterphp
Member since:
2017-05-15

It all comes from the same factory in china. It makes very little difference what brand you buy.

I've been building my own kit of the last 20 years and everything works fine if you get "the generic paste" from PC World or something with a a fancy name.

EDIT: I had no idea that their conclusion supported mine before reading it. It just a know thing if you build your own kit.

@Thom do you build your own?

Edited 2017-08-07 22:30 UTC

Reply Score: 2

RE: Comment by grandmasterphp
by Alfman on Tue 8th Aug 2017 00:17 UTC in reply to "Comment by grandmasterphp"
Alfman Member since:
2011-01-28

grandmasterphp,

@Thom do you build your own?


I'm also curious, what kind of rig you got Thom?

My own personal computer is a mid tower with a gigabyte mainboard with dual core e3110 xeons. I've got an acer laptop as well. Both are due for an update, but alas the lack of disposable income...

Reply Score: 2

Thom_Holwerda Member since:
2005-06-29

I've always huilt my own PCs, and earlier this year I took it to the next level by designing and building my own custom watercooled system.

https://twitter.com/thomholwerda/status/881878748303380480

Reply Score: 1

RE[3]: Comment by grandmasterphp
by Alfman on Tue 8th Aug 2017 14:05 UTC in reply to "RE[2]: Comment by grandmasterphp"
Alfman Member since:
2011-01-28

Thom Holwerda,

I've always huilt my own PCs, and earlier this year I took it to the next level by designing and building my own custom watercooled system.

https://twitter.com/thomholwerda/status/881878748303380480


BTW the video never loaded for some reason, only the screen shot. Anyways yes watercooling and overclocking is big with hardcore gamers it seems. You have very little in the way of expansion cards and drives, looks like you could go with a smaller form factor if you wanted to.

Is it normal to run the CPU and GPU on the same loop? I'm not sure which direction the water's going, but it seems not ideal that the heat output from one enters the other. Not criticizing, but just that I have no experience with water cooling and don't know much about it.

Reply Score: 2

Thom_Holwerda Member since:
2005-06-29

Loop order doesn't matter; the coolant temperature is uniform all over the loop (physics!). The form factor could technically be smaller, but that would force a smaller radiator, higher fan RPM, etc.

Reply Score: 1

RE[5]: Comment by grandmasterphp
by Alfman on Tue 8th Aug 2017 15:32 UTC in reply to "RE[4]: Comment by grandmasterphp"
Alfman Member since:
2011-01-28

Thom Holwerda,

Loop order doesn't matter; the coolant temperature is uniform all over the loop (physics!). The form factor could technically be smaller, but that would force a smaller radiator, higher fan RPM, etc.


There would have to be some temperature difference between the temperature entering the CPU cooler and that exiting the CPU cooler, but I guess the heated water is taken away fast enough that the difference is negligible.

I didn't see clearly from the screen shot, but it looks like the entire top of the case is the radiator? If so, that is quite a large surface, I didn't realize it would be that large.

I've thought about building a water rig, but frankly none of my personal computers ever really warranted it. On servers perhaps it could be useful, but I'm not even sure it would be allowed by colocation facilities, haha. Server fans are especially loud, but the servers are usually physically isolated. I imagine some specialized datacenters might be be equipped for watercooling an entire rack with external pumps & radiators.

Sometimes I wonder if we should be finding a better way to use the heat waste to do other useful work.

Reply Score: 2

Thom_Holwerda Member since:
2005-06-29

There would have to be some temperature difference between the temperature entering the CPU cooler and that exiting the CPU cooler, but I guess the heated water is taken away fast enough that the difference is negligible.


"A" difference is obviously there, but the way fluids work and the speed of the coolant flow means it's so small it's often even difficult to register. You have to set up an SLI system, max the CPU and GPU load, and only then can you measure maybe a 1-2 degree difference in CPU/GPU temps between your loop being GPU>CPU>rad or GPU>rad>CPU.

I've thought about building a water rig, but frankly none of my personal computers ever really warranted it.


Neither did my PC, actually - it's a 7700K and GTX 1070, both running at stock, non-overlocked speeds. I did it for the fun, the challenge, and the looks. Also it's really expensive. The cooling parts alone ran me ~€1000.

Edited 2017-08-08 15:52 UTC

Reply Score: 1

RE[7]: Comment by grandmasterphp
by Alfman on Tue 8th Aug 2017 18:17 UTC in reply to "RE[6]: Comment by grandmasterphp"
Alfman Member since:
2011-01-28

Thom Holwerda,

Neither did my PC, actually - it's a 7700K and GTX 1070, both running at stock, non-overlocked speeds. I did it for the fun, the challenge, and the looks. Also it's really expensive. The cooling parts alone ran me ~€1000.


Rock on ;)


If you want to see an example of tons of work on water cooling overkill, linus tech tips did a video series on watercooling his office and how it all went awry.
https://www.youtube.com/watch?v=EV6oYfcAwLM

I'll give him credit for trying to "go big", but he wouldn't do it again, haha.

Reply Score: 2

RE: Comment by grandmasterphp
by Kochise on Tue 8th Aug 2017 07:35 UTC in reply to "Comment by grandmasterphp"
Kochise Member since:
2006-03-03

Some even tested toothpaste with some success.

Reply Score: 2

RE[2]: Comment by grandmasterphp
by Alfman on Tue 8th Aug 2017 13:39 UTC in reply to "RE: Comment by grandmasterphp"
Alfman Member since:
2011-01-28

Kochise,

Some even tested toothpaste with some success.


How do you know?

For the sake of curiosity I wish they would have tested the same heatsinks with no heatsink compound at all. Without that datapoint it's not obvious to me that the toothpaste helped at all, it may have acted as an insulator between the surfaces for all I know.

Reply Score: 2

RE[4]: Comment by grandmasterphp
by Alfman on Tue 8th Aug 2017 18:49 UTC in reply to "RE[3]: Comment by grandmasterphp"
Alfman Member since:
2011-01-28

Kochise,

Those are amusing, especially the last link ;)
Edit: I don't know if the water based compounds would remain as effective over time after the water evaporates.
But I still don't see where any of those substitutes are compared to using no compound at all?

Sure you can use mayo, but did it actually help? That's my question.

Edited 2017-08-08 18:54 UTC

Reply Score: 2

quatermass Member since:
2005-08-03

The shiny metal surfaces aren't that smooth on the micro millimetres scale. So we put a paste on to fill in the valleys of each surface.

Trouble is most people put far too much on.

In the Military work I used to do, we had a special applicator tool that basically made a 0.1mm thin layer.
You can't do this by hand without such a tool.

Reply Score: 1

RE[6]: Comment by grandmasterphp
by Alfman on Wed 9th Aug 2017 01:12 UTC in reply to "RE[5]: Comment by grandmasterphp"
Alfman Member since:
2011-01-28

quatermass,

The shiny metal surfaces aren't that smooth on the micro millimetres scale. So we put a paste on to fill in the valleys of each surface.

Trouble is most people put far too much on.


I'm not questioning that thermal paste improves thermal conductivity. However the authors tested toothpaste, and that got me curious how it compares to using no compound.

For example:
artic silver 5 = 37.2
toothpaste = 48.7

Of course I'd use thermal paste to get the max cooling, but it was just a hypothetical question: does toothpaste help at all or would it be just as good to not use anything? Obviously the data provided is inadequate to answer that question. I tried looking this up, but most people making claims have no experience/empirical data to back their claims:

http://www.tomshardware.com/answers/id-2631540/test-cpu-thermal-pas...

http://www.tomshardware.com/forum/292749-28-thermal-paste

Those who have tested cooling without thermal paste generally report not having any issues whatsoever, and even the temps are well within safe operating parameters.

https://www.youtube.com/watch?v=yX6AN4vWZP0

https://www.youtube.com/watch?v=RAr2wKZ_nes

https://linustechtips.com/main/topic/205743-cpu-without-thermal-past...

Edit:
This guy actually demonstrates what happens when one removes the heatsink entirely.
https://www.youtube.com/watch?v=Xf0VuRG7MN4
It's pretty much what you'd expect: Some CPUs throttle down, maintain a low temp, and recover when the heatsink is returned. Others have no throttling and just go up and up in smoke as the temps climb to 370C!

Edited 2017-08-09 01:28 UTC

Reply Score: 2

RE[7]: Comment by grandmasterphp
by Kochise on Wed 9th Aug 2017 05:09 UTC in reply to "RE[6]: Comment by grandmasterphp"
Kochise Member since:
2006-03-03

Old AMD Athlon Palomino that not had any throttling capabilities. Since Barton this is solved.

Reply Score: 2

RE[7]: Comment by grandmasterphp
by Megol on Wed 9th Aug 2017 22:08 UTC in reply to "RE[6]: Comment by grandmasterphp"
Megol Member since:
2011-04-11

quatermass,

"The shiny metal surfaces aren't that smooth on the micro millimetres scale. So we put a paste on to fill in the valleys of each surface.

Trouble is most people put far too much on.

"

This exactly. In many cases the goop with worse on the paper performance will actually perform better IF it is easier to spread out (=viscosity) and the user have read about the advantages of a thin layer.


I'm not questioning that thermal paste improves thermal conductivity. However the authors tested toothpaste, and that got me curious how it compares to using no compound.


I actually knew someone that tested toothpaste for a short while. It worked reasonably well and better than using no paste at all. However even toothpastes have different formulations so one can't make a general claim. The layer used in the above case was very thin and it was tried for some hours IIRC which means outgassing wasn't a significant factor. Using it for longer periods will probably reduce the paste to a dry powder with little heat conduction.

Reply Score: 2

RE: Comment by grandmasterphp
by Megol on Wed 9th Aug 2017 21:58 UTC in reply to "Comment by grandmasterphp"
Megol Member since:
2011-04-11

It all comes from the same factory in china. It makes very little difference what brand you buy.


No it makes a great difference if the product is as stated (but that can of course not be guaranteed in general). There are different compunds used for different things that have different characteristics. Large gap (relatively)? Use A. Need the best thermal transfer possible for thermal pastes? Use B. How coarse are the surfaces? If they are very coarse use C, if they are very fine then use D.

But then I generally use industrial products not made in China by the lowest bidder. IME even expensive top of the line unobtanium containing greases produced for overclockers are generally worse than the higher performance (for the application - see above!) inexpensive stuff for industrial use. If not at the time of application then after some months to some year - thermal pumping tend to destroy the performance of overclocker stuff.

Reply Score: 2

RE: Comment by grandmasterphp
by bassbeast on Sat 12th Aug 2017 01:56 UTC in reply to "Comment by grandmasterphp"
bassbeast Member since:
2007-11-11

I have to disagree as I've worked on several systems that were being refurbed after years in some of the worst conditions you can put a PC in, working in warehouses and construction zones where there is tons of dirt and crap clogging up fans and I've been able to tell those that used Arctic Silver pastes versus the generic stuff because over time the generic quickly loses its ability to cool while the Arctic Silver brand pastes keep on chugging.

I've used both the regular and the ceramic and while I personally prefer the ceramic (its easier to remove the cooler down the line if the unit gets an upgrade and I've found it keeps it a couple of degrees cooler than the AS regular) I've found both kinds work VERY well and more importantly KEEP working very well years down the line.

Reply Score: 2

Comment by ahferroin7
by ahferroin7 on Tue 8th Aug 2017 12:14 UTC
ahferroin7
Member since:
2015-10-30

The brand and variety are divisive (but shouldn't be, as the article implies), but applying it shouldn't be if you're talking to anyone with a basic knowledge of thermodynamics. You need exactly enough to get a uniform layer as thin as possible between the heatsink and processor lid, and shouldn't spread it otherwise you incur a higher risk of air bubbles which in turn reduce it's effectiveness. There's solid, well understood, and well tested science behind correct application.

Reply Score: 2

Why oh why?
by quatermass on Tue 8th Aug 2017 22:40 UTC
quatermass
Member since:
2005-08-03

Ok, how come no reviewers ever look into the thermal rating of the paste as this one is easiest to fake?

This is the thermal conductivity of it. Usually rated in Wm/K. The higher this is, the more heat it conducts through it.

You want this to be well over 10.

Reply Score: 1

Arctic Silver for me...
by Drunkula on Wed 9th Aug 2017 12:36 UTC
Drunkula
Member since:
2009-09-03

It's the only one I've used in years. Works great for me! Funny they rated it so low.

Edited 2017-08-09 12:46 UTC

Reply Score: 2