Thermal Compound Application Methods Benchmark

[klemm]

I think it would be nice to do a follow up video to this where you got the temps up higher (80C Range) just to see if that will effect the results, but then we go down the road of different pastes as well etc etc. So this video is a good benchmark for the average PC user.

 

Good Job.

[cabritar]

Luke, what were the fan profiles set to ?

• Was it set to temperature curves?
• Was it set to CPU utilization?

I think in some cases you could have 2 pastes, 1 working better than the other but the water cooling has to work harder in one instance VS the other.

 

I assume this variable has been taken into consideration but it wasn't mentioned in the video.

 

I like this series. Testing "common knowledge" with the scientific method gets me likes, comments, and favorites.

 

Keep it UP!

[fueg]

There was no "does not really matter" choice in the strawpolls

[Progressor]

8 hours ago, aflyinghotpocket said:

I would like to see a followup to this video Where Luke tests the following:

 

• Different types of thermal paste (are some better then others?)
• Different types of cooling methods (liquid vs. fans)
• Completely naked CPU (no thermal paste) with and without cooling 
• Different CPU?, AMD chips tend to run a bit hotter then Intel (in some situations). 

I know there are WAY to many combinations to cover in a ~10-12 minute video, but I would like to see more variety with this Workshop. 

Great video as always guys!

 

4 hours ago, jasonvp said:

I'd pay money to see this one.  TIM manufacturers don't have as much at stake as far as how you apply the product.  But they certainly do when it comes to which product you choose.  I'll bet the results from the testing would ruffle some feathers.  My gut tells me that it doesn't matter much. ;-)

 

Regarding the different brands and types of thermal pastes, you guys should look it up, there are plenty of benchmarks on the interwebs(I like the ones on Tom's Hardware). And yes, there is huge difference between between types/brands. Some of the benchmarks even include crazy shit like toothpaste and other gooey stuff. Which can make a pretty funny video provided that LMG decide to make a video on that. Also if they haven't already they can make a video on all the types of heat transfer solutions, like different types of thermal pads and thermal pastes(like liquid metal, silicone, metal oxide, ceramic based)their intended uses(for example some need time to "cure" and their properties improve over time, some are good under extreme low temperatures like under liquid cooling use, etc.) and advantages/disadvantages.

[PianoPlayer88Key]

I've only watched a couple minutes  of the video and read the first several comments in this thread thus far, but my thoughts so far are...

 

First, I saw the 16.8 C ambient temp in the warehouse, converted it to F (62.24) ... brrr!!  I would love to see the test done in a warmer environment. For example, in summer our house often reaches 30-32 C (86-89 F) indoors, and I think one hot summer several years ago it could have been 38-40 C (100-104 F) in the house.  Maybe test in a 50 C / 122 F environment?  (Just keep the humidity low, I'd tentatively suggest below 20% otherwise you could be quite uncomfortable especially if it's high like 70+%.)

 

Re: comments about 47 C CPU readings on most tests ... I also, with someone else, suggest setting the fan to a constant speed.  I too agree with using a cheaper cooler, like the 212 Evo which I have, or stock.

And for heating up the CPU - on my i7-4790K, if I run small FFT (or max heat) in Prime95 with the newest versions, I've had my CPU temps hit 100 C within a few seconds, and on one occasion (I think in summer or early fall - it stays warm here fairly late in the year) it was thermal throttling at 3.7 GHz.  Btw I used the pea method, and it was my first time ever applying paste.

 

I'd also like to see tested how it affects temps if you have the case in an enclosed cubby hole like some desks have.  What results might you get if you used a cheap case with little airflow, only a single 80mm fan, a 6700K OC'd so its TDP matches the FX-9590, with a stock cooler borrowed from an i3 cause the 6700K I think doesn't come with one, Prime95 small FFT, and two R9 390X2s in XFire, overclocked a similar percentage, with a similar heat/stress test?

 

I also like @aflyinghotpocket's ideas.  His third one about the naked CPU reminded me of another idea... How good of a gaming experience could the LTT group get, even if you had to use older less-demanding games, if you ran the CPU and GPU completely naked? This would mean no paste, no heatsink or watercooler, delidding the CPU, etc. Also everything else in the computer would be passively cooled if at all - no fans, just silence.  Bonus points for testing in something like JonnyGuru's hot box. Of course keep software safeguards in place like thermal throttling, etc.  I have yet to learn of a CPU that could safely run at several thousand C. (I think it wouldn't be possible to exceed the melting point of tungsten, unless it's under high atmospheric pressure?)

[xnamkcor]

5 hours ago, DXMember said:

In the intro you promised X shape, there's none in the video

Where's the circle?

Where's the smiley face?

 

Learn from the best?

https://www.pugetsystems.com/labs/articles/Thermal-Paste-Application-Techniques-170/

 

also where is CISCO application method?

The "x" was the one that looked like a +, but rotated about 45 degrees.

[Zodiark1593]

I'm a fan of the dot or line. Easy to gauge, and a lot less work than spreading it around.

[Tricklet]

On 2016-02-25 at 9:42 AM, Progressor said:

Come on Luke, how can you test this at 47 degrees C and expect any results.. Sometimes it looks as you guys have no idea what you're doing. The purpose of a thermal compound is to transfer heat. HEAT, yes guys HEAT! You need to push the thermal paste to it's maximum heat transfer limit, that's when other factors such as spreading start to make the difference, at 47 C there is simply not enough heat transfer for the spreading method to "bottleneck" the thermal paste. I'll repeat that again, you need high enough temperatures(80 C should do) so the spreading method becomes a bottleneck(at least the incorrect applying methods) to the heat transfer!

I think this is incorrect IMO. Looking at the properties of IC Diamond stating temperature range @ -40*C -- +85*C. But only from 130W so not good for 5000-series CPU in that case.  

This should validate 47*C to be correct anyway. Once you have a good layer to even out imperfections in the two mating surfaces, nothing else matters. The TC will transfer heat unaffected between above mentioned temp ranges and should not matter if its 47 or 80*C as long as the radiator can handle the wattage; and every other option except "too little thermal compound" was used.  

 

What I think is actually going on is the TC being so good these days it can eliminate sloppy application practices. This was not the case back in the day.

[PianoPlayer88Key]

I have the Cooler Master Hyper 212 EVO on my i7-4790K, and I used the "pea" method last January (2015) when I applied the CM paste.

 

I recorded this video a short while ago that shows my CPU temps running Prime95.  For the first part of it, I was trying to run FireStrike, Valley & Heaven, and Cinebench simultaneously.  (I had FS, V & H cranked all the way up on quality settings, but running at 720p resolution.)  I noticed that while Prime95 was running, the temps would actually be considerably lower with Cinebench running than with it stopped.  (Also I'd like to see more reviewers run multiple applications simultaneously when benchmarking, so it reflects the use of people like me who run a bunch of things at once. )

In the later portion of the video, I stopped the other benchmarks, and ran Prime95 by itself.  Notice the CPU temperatures... (brrr...NOT!)

 

 

 

In the video, I was using P95 version 28.7.  I have learned some time ago that new versions of Prime95 stress the Haswell and newer CPUs in a way that iirc wasn't intended.  I forgot to include the next test in the video, but off-camera, I ran the same isolated stress test, but this time with Prime95 26.6.  My temps didn't get much over 65°C, if at all.

 

But, considering the temps I got with the newer version of P95, would a different application method of paste make any difference?  Or would a different cooler work?

 

Hey I was thinking, too.  This would be quite out of reach for me, but what CPU cooler would the LTT team use, so they could overclock a 4790K or 6700K by 50% (referencing the 3DMark Steam achievement "OC Madness"; or so its thermal wattage / TDP / whatever matches the stock FX-9590 (or even a 50%-OC'd FX-9590), whichever is higher/est), run the newest Prime95 test in an environment where the ambient temp is ~50°C (like what is possible outdoors in some places in southern California & western Arizona in summer) ... and the max temps on their CPU are only getting up to what a typical cold winter night outdoors in Canada would be?

I was going to suggest targetting CPU temps like what Antarctica can get, but decided to make it a bit more achievable.  (I live in southern California, although not where temps reach 50°C - typically we routinely see 38°C in summer, and have seen 47°C at least once in my lifetime.  I know LTT group is in Canada, although I haven't researched the nighttime winter temps there, I'm guessing it could get below 0 to -20°C or so.)

Speaking of Antarctica though ... if someone had a computer in a place like Vostok Station, I wonder how fast of a CPU and GPU you could run *WITHOUT* any cooler at all, even delidding the CPU?

 

[iamdarkyoshi]

[NumLock21]

On 2/25/2016 at 1:40 PM, DXMember said:

In the intro you promised X shape, there's none in the video

Where's the circle?

Where's the smiley face?

 

Learn from the best?

https://www.pugetsystems.com/labs/articles/Thermal-Paste-Application-Techniques-170/

 

also where is CISCO application method?

You beat me to it. Was going to post that, showing that X is the best, not in terms of temperature, but how well the thermal paste spreads. There is the X shape in the video.

[Progressor]

On 2/26/2016 at 11:03 AM, Tricklet said:

I think this is incorrect IMO. Looking at the properties of IC Diamond stating temperature range @ -40*C -- +85*C. But only from 130W so not good for 5000-series CPU in that case.  

This should validate 47*C to be correct anyway. Once you have a good layer to even out imperfections in the two mating surfaces, nothing else matters. The TC will transfer heat unaffected between above mentioned temp ranges and should not matter if its 47 or 80*C as long as the radiator can handle the wattage; and every other option except "too little thermal compound" was used.  

 

What I think is actually going on is the TC being so good these days it can eliminate sloppy application practices. This was not the case back in the day.

I'd disagree. In my experience the thickness of the layer of paste between the CPU/GPU and the heatsink/cooler does affect the temperature especially in my case on my laptop, having just a "good layer to even imperfections" is not enough if you don't consider the thickness of that layer, which is affected by the spreading method . Since the cooler in my laptop by default uses a thermal pad(terrible solution) which simply didn't work I was left with few solutions:

- replace the thermal pad with a pure copper pad with the same thickness - this gave me the "worst" results, In my opinion due to requiring two layer of thermal paste(one between the cpu/gpu and the copper shim, and another one between the copper shim and the cooler).

- remove the thermal pad and instead of putting a copper shim in it's place just try to tighten the screws of the heat sink so the gap between the cpu/gpu and heatsink is smaller, use more thermal paste to fill the gap. This gave similar but slightly better results, still not enough.

- remove the thermal pad, bend the heat pipes so there is no gap between the cpu/gpu and the cooler, tight the screws as much as possible and put normal amount of thermal paste. This gave the best and actually satisfying results that allowed me stress the laptop as much as possible while reaching max 63cpu/74gpu C. For comparison with the thermal pad I couldn't even boot the laptop into windows(doesn't really applies since we're talking about thermal paste), and with the copper shim(two thin layers thermal paste) and too much paste methods I was still reaching 80/90 C under stress.

 

All testing was done on an already undervolted cpu(no underclocking, no stability issues) and using pea method of spreading with ArcticCooling MX4 paste, high mounting pressure on "copper shim/two paste layers" and "bent heatpipes/one thin layer" methods.. low mounting pressure on "one think layer of paste" method. 

So, in my experience the thickness of the paste and amount of layers did affect the temperatures, which applies to spreading methods in a desktop cpu. Now in most cases where we're talking desktop CPUs with high pressure mounted cooler, the pressure should be high enough to displace the excessive paste in "X" and "too much paste" methods, so I'd agree that unless your cooler is with low mounting pressure there should not be difference between pea, line, X and too much paste methods(as I said the high pressure will displace the excessive paste out). We all agree on the "too little paste method", so we're left with the method where you smear the paste over the entire surface of the heat spreader/naked cpu, this would create bubble pockets in the layer so it should have different results(for the worse since there is less paste in in contact to cpu/cooler) provided you use high temperatures to bottleneck the thermals paste as I said in my first comment.

 

In addition, it never occurred to me(none of my shit rigs have this functionality) that there are other variables which can compromise the results in the video. As other users suggested, differences of fan speed and cpu/gpu clock that are automatically applied depending on the temperature in order to reduce it.. can make so the results seem the same too. Those should have been set to a constant, specific fan speed and clock(provided they weren't).

 

I'd really like to see them make another video, this time done properly to prove me and the others complaining right or wrong.

 

 

[DXMember]

On ‎2016‎.‎02‎.‎26‎. at 10:27 PM, NumLock21 said:

You beat me to it. Was going to post that, showing that X is the best, not in terms of temperature, but how well the thermal paste spreads. There is the X shape in the video.

a + is not an X

[xnamkcor]

57 minutes ago, DXMember said:

a + is not an X

Now I have to research the neurological or mental causes of being unable realise the rotation of an object and recognise it.

[Zerim]

This is something that, like the fan obstruction tests, people want to matter, when it really just doesn't. This is probably a side effect of the fact it's one of the two or three things that even approaches requiring thought in building a computer (along with not touching the pins of the CPU and attaching the I/O panel cables). 

 

Personal computers are not precision equipment. If precision actually mattered, manufacturers wouldn't let the user do it. To the extent it does matter, very few users would notice or benefit from the difference. 

 

I've found the definition of a hobby, I guess

[xnamkcor]

23 minutes ago, Zerim said:

This is something that, like the fan obstruction tests, people want to matter, when it really just doesn't. This is probably a side effect of the fact it's one of the two or three things that even approaches requiring thought in building a computer (along with not touching the pins of the CPU and attaching the I/O panel cables). 

 

Personal computers are not precision equipment. If precision actually mattered, manufacturers wouldn't let the user do it. To the extent it does matter, very few users would notice or benefit from the difference. 

 

I've found the definition of a hobby, I guess

Good point, he forgot the "Thermal Paste On Pins" "pattern".

[DXMember]

9 hours ago, xnamkcor said:

Now I have to research the neurological or mental causes of being unable realise the rotation of an object and recognise it.

do you rotate the die under the IHS as well?

[xnamkcor]

24 minutes ago, DXMember said:

do you rotate the die under the IHS as well?

Considering the contact point is either in the center or a line, The rotation hardly matters. And if you wanted to be super-specific about the "concentration"(that will get spread across the entire surface anyway because of the final distance between the heatsink and the CPU and the amount of thermal paste used), you would just use the line pattern.

 

With exception of: No Paste, A line following the extreme border of the CPU, or using chunky peanut butter, prefering + over x is like buying a 1000 USD HDMI cable.

[DXMember]

31 minutes ago, xnamkcor said:

Considering the contact point is either in the center or a line, The rotation hardly matters. And if you wanted to be super-specific about the "concentration"(that will get spread across the entire surface anyway because of the final distance between the heatsink and the CPU and the amount of thermal paste used), you would just use the line pattern.

 

With exception of: No Paste, A line following the extreme border of the CPU, or using chunky peanut butter, prefering + over x is like buying a 1000 USD HDMI cable.

yeah but due to the quality of the production you can't draw these conclusions from this specific video

it's like saying + and x are the same is like building a car that doesn't comply to co emission requirements and saying it does

[DeScruff]

I'm late, but a topic I'd like to see if there is ever another 'thermal paste' workshop episode:
Does reusing thermal paste really affect temperatures?
Ive seen people say you should always clean and reapply, others say its fine if your not overclocking, or some people say it doesn't affect anything at all.
Ive also heard arguments about the age of the thermal paste, like if it was applied a month ago it will be too dry to reapply, with others saying the heat will "reflow" the paste and get rid of air bubbles.

A likely scenario where this comes into question is if the heatsink you have is large and covers a ram slot or two, and you need to remove it to add/replace ram.