Just how important is thermal paste? The results may surprise some people... Tested by Texas_Ace

Ok, first off this will most likely be my last post on here for a few weeks at least as internet access will be extremely limited. So sorry if I am not able to reply to any questions. I will read them when I get back on though.

Long story short, recent events have made me curious as to the real world effects of thermal paste vs clamping force.

Well the results were more or less what I expected based on my experience and anecdotal evidence along with logically thinking about the issue.

It is nice to have to hard data to see the real world effects instead of paper numbers to guess at those effects.

There are a lot of charts below, I ran several tests back to back.

They are all using the same LED, a nichia 319A D440 6500k 70CRI LED mounted on a 16mm DTP copper mcpcb (the ones sold by Convoy), no lapping of any kind, just stock mcpcb surface.

The mcpcb is mounted to the 140W fan cooled heat sink with 2 screws that offer far more clamping force then almost any flashlight.

For the tests I ran 5 separate tests.

I did 1 test with both no AS5 thermal paste and with AS5 with as much clamping force as I could get without stripping the screws.

I then did the same tests except I only tightened the screws to light finger pressure, basically what the tips of my fingers could do. Still much higher clamping force then the vast majority of flashlights that do not screw down the mcpcb. For this test the thermal paste was applied like I always do with a thin film, just enough to cover the surface.

The last test was for thermal paste application. This time I applied the thermal paste by putting a pea sized blob of thermal paste on the mcpcb and then loosely tightening it up as you see in many cheap flashlights from the factory.

The results of these tests were pretty much as I expected:

The high clamping force tests showed virtually identical results up until the very limits of the LED, where the thermal paste pulled away very slightly. The thermal paste increased the max lumens by less then 1% in this case and only at the very edge of what the LED can handle.

The low clamping force results were also not surprising, although I will admit that the thermal paste did a bit better then I excepted, I had excepted it to end up about like the high clamping force no paste test. In this case the thermal paste gained 2.2%. This is why properly applied thermal paste works so well in flashlights with low to moderate clamping force.

The results of this last test were not remotely surprising as I have seen this in many cheap flashlights, without proper clamping force the thermal paste won’t be “squished out” from under the mcpcb, this leads to a much thicker layer of thermal paste then needed or desired. Many many lights do not have enough clamping force from the factor and have a thick layer of thermal paste under them when you pull it apart.

When you have this thick layer the performance is far worse then even the no-thermal paste with weak clamping force. In this case peak lumens dropped by 9% and they would have kept dropping. The lumens were dropping quickly as I ran the test, to keep things consistent I kept about the same timing between readings as the prior tests.

So what is the end result?

Apply thermal paste in your flashlights but worry more about the clamping force of the mcpcb to the pill.

Make sure to only apply a thin layer of thermal paste, too much is just as bad or worse then too little. plus it is messy.

I won’t be making any changes to my LED test procedures. Thermal paste is a pain to work with and has less then a 1% effect on the final numbers in my test rig with lots of clamping force.

Now for the hard real world data, enjoy, click any image to see a larger version:

Comparison of all LED’s together:

Zoomed in version of the comparison:

No thermal paste, clamped hard:

With AS5 thermal paste clamped hard:

No thermal paste, clamped loose:

With AS5 thermal paste, clamped loose:

With excessive AS5 thermal paste, clamped loose:

Hopefully people will find this useful.

You Sir, are awesome. I will be reading this more thoroughly when I have more time. Thanks a lot! :slight_smile:

Have you considered doing the test for a 50W or 100W led rather than 25W?
Thanks for the test though.

one big problem with no paste is when the surface of the star or host are not perfectly flat, then the clamp force will not lead to a good result

of course if you have a perfectly flat heat sink and star with enough clamp force the thermal path will be good

also the particle size of a thermal paste will result in a differnt result
one cheap paste you can feel the particles on your finger tips, while arctic silver 5 or other good pastes feel like a liquid

Because I like summarizing tables:

Paste Clamping Max Lumens Amps at Max Lumens
None Hard 1462 6.0, 6.25
With Hard 1476 6.25, 6.50
None Loose 1443 6.00, 6.25
With Loose 1477 6.25
Excessive Loose 1354 5.5

nice, been saying it many times, not too much paste
When I see the star of a LED with paste pushed out in a commercial light i receive just clean one of the two surfaces nd still have plenty of paste.

Thanks for that.

Also, popcorn is almost ready (homemade no less.)

If you have a “nearly perfect flat”, you even don’t need thermal paste

Thanks sac02, that makes it extremely easy to read!! And a big thank you to T_A for providing the info to read… :+1:

Thanks for taking the time to do this TA. :+1:

T_A, I do have one question.

In the last scenario of “Excessive Thermal Paste”, was any effort at all made to squeeze out the “Excessive” amount or was it just left as “Excessive”?

Nice info also, thanks for doing it as well as posting. :+1:

Never reuse thermal paste. Always apply new paste.
When you remove a led board en want to mount it again with the same thermal paste you have a big chance to have air bubbles underneath. And we all know that air is a good insulator.

Great tests TA! I’m surprised of how low the difference is between using thermal paste and using no thermal paste.

Great tests

So not a great deal of difference so long as the star is nice n tight.

Dont suppose you want to repeat the test with the star soldered do you :slight_smile:

Not being a real modder, but every (MC)PCB that comes out of the bag takes a turn on the lapping-stone. A tiny bit of thermal paste and then some wiggling until the paste emerges on the side. The screws are used for positioning the led.

Interesting test.
Did you heat cycle the AS5 or give time to allow it to cure per the instructions?

Nice job! Interesting to see how much AS5 actually helps.

The light I have in which I noticed the paste the most is my modded Sunwayman C10R . The light started as a 1xCR123 light with electronic side switch. I modded it to triple XPL HI and upped the battery to a much larger and longer 18500. I replaced the stock driver with an e-switch FET driver from Mountain Electronics.

To make room for the longer battery, I placed the driver above the LED shelf, then sat the triple copper Noctigon above the driver. I used a strip of copper bent into a c-ring to keep the star elevated at the right height.

Naturally in that configuration, heat is an issue. This is a triple with FET driver and at max power with an IMR 18500 it pulls serious amps. The only heatsinking is where the sides of the star touch the head and where the edge of the bottom rests on the copper c-ring. There is no heatsinking at all below the center of the star. Like a hollow-pill but worse, because there isn’t even a built in ledge for the star to rest on.

I found that without any thermal paste, the wires to the star in the modded light would desolder themselves in about a minute when turned on turbo even with 30 second rampdown to 50% power. To fix this I liberally spread AS5 between the C-ring and head, and where the star contacts the C-ring and edges of the head. Problem solved! The head gets hotter much faster and the wires to the star no longer desolder themselves no matter how long I run the light.

Good subject and test, thanks…

Well, to me, the reason to use thermal paste is many.
It will more or less help negate all but the worst non flat shelves and star boards.

If one sands the star and shelf flat, and uses a coarser grit sand paper/crocus cloth on glass, or better yet, a coarse flat stone, this makes for more surface area from the coarse sanded pits and crevices, for better heat transfer(only with thermal paste). The pits and crevices, thermal paste will fill in all those crevices/scratches and transfer heat well with more surface area to dissipate/absorb heat…
So don’t worry about getting a perfectly smooth glass like surface, not needed. Flat, yes, get it as flat as possible.

I’ve seen tests in other thermal applications, of perfectly machined flat smooth surfaces compared to coarse perfectly machined flat surfaces, both with thermal paste, and the coarse(scratched) surface wins out…

On the lights with 2 screws on opposite sides, imo be careful of over tightening the star board and warping it out of flat…

All thermal paste is not created equal, some works much better than others in my experience (in other heat transfer applications), but any paste is always better then none. The correct application of these thick expensive pastes is critical for best heat transfer. Not always true, but usually you get what you pay for with thermal pastes. The very cheap stuff not as good.

ARCTIC MX-4 is my favorite, not electrically conductive like arctic silver and its cheaper then AS5.


Nice test, so with a single die led you can’t go wrong under 4 amps, hence the carelesness when mounting leds of most manufacturers :slight_smile:

Correct me if I am wrong, but “IF” I am reading the T_A graphs & sac02’s summary table correctly, there is very little difference between “thermal paste clamped hard” and “thermal paste left loose”.

And both of these were at the top in results.

Am I reading and interpreting this correctly?

Great testing as always T/A! I hate thermal paste for the same reasons! I take the time to flatten and lap surfaces together whether it takes days,hours or minutes and use just a see thru ultra light swipe across the star surface of paste, then after centering and tweaking, heat cycling, lock the screws down or if I have enough clamping force between bezel, reflector and star, its good enough! Enough clamping force means I can smack the light against a thinly padded board and not loose focus! :wink: