X210 CPU cooling improvement options?

[L29Ah]

The cooling setup of X201 is inadequate to unleash the full potential of the modern i5/i7 CPUs. I wonder what are my options to improve it.
The current power dissipation can be displayed by tools like `throttled --monitor`. For instance, my X210 with i7-8550U CPU at 92°C with a stock X201 cooler and freshly-applied КПТ-8 paste at 25°C ambient temperature during `sysbench cpu run --threads=9` load dissipates around 21W.

There's a hacked version of the stock X201 cooler by a 51nb forum member "17m19" available for $65, i wonder how much power does it dissipate, and whether it would make sense to stuff thicker/even more heat pipes in there:
https://i.imgur.com/1YaJ4Wv.jpg

Also it's possible to add an external fan to assist the internal one, but i afraid if it's powerful enough to make a big difference it can actually fry the internal one (or even the motherboard) due to increased back-EMF in the fan motor (that turns into a generator that the stock circuitry may not be designed to handle correctly):
https://images-na.ssl-images-amazon.com ... L1000_.jpg

Also https://www.pcper.com/reviews/Mobile/51 ... quid-Metal suggests using liquid metal for improving the thermal interface between the CPU die and the stock cooler doesn't improve its performance by a sensible margin, and meanwhile creates a hazard since the liquid metal (gallium, to be precise) is corrosive to copper, will short any electronics it's spilled on, and may crack the CPU die if frozen due to expansion.

Any other ideas?

[L29Ah]

Oh [censored], turns out X200 and X201 coolers aren't compatible:
https://dump.bitcheese.net/files/fewibu ... 25_002.jpg
https://dump.bitcheese.net/files/xodeka ... 25_001.jpg

[xytech]

Hi guys,

If anybody needs better cooling, the fan by 17m19 of 51nb which was custom designed for the X210 is available individually through me. It's able to sustain up to 39w on the 8650u, with turbo boost maintaining 3.9GHz at maximum load.

https://forum.51nb.com/forum.php?mod=vi ... id=1871059

It's a 3 pipe copper design for the X200 and another custom designed cooling assembly for the s variant of the laptops.

Check out my other post if anybody need it for cooling. It also performs well for the 8250u too but it's not a necessity compared to the 8650u or the 8550u.

Thanks!

[L29Ah]

https://dump.bitcheese.net/files/hadyma ... n-mod.jpeg
Soldered an additional heat pipe to my X201 cooler. Got only ~13% more power at 10 minutes than the stock config. Either 17m19 did an astonishingly good job, or he has a [censored] test routine that doesn't wait for the cooling system to stabilize its power sink. Would be glad if anyone with 17m19's cooler post their results.

[axur-delmeria]

Soldered an additional heat pipe to my X201 cooler. Got only ~13% more power at 10 minutes than the stock config.

Is 13% not enough?
Are you sure that the transplanted heatpipe survived the bending and soldering processes? I've done something similar on a non-Thinkpad laptop, but the improvement was meager at best. OTOH, it was an Atom notebook that didn't have a heatpipe at all, and the fact that it's still running (though kept in an air-conditioned room) could mean that the heatpipe transplant did some good.

[L29Ah]

Soldered an additional heat pipe to my X201 cooler. Got only ~13% more power at 10 minutes than the stock config.

Is 13% not enough?
Are you sure that the transplanted heatpipe survived the bending and soldering processes? I've done something similar on a non-Thinkpad laptop, but the improvement was meager at best. OTOH, it was an Atom notebook that didn't have a heatpipe at all, and the fact that it's still running (though kept in an air-conditioned room) could mean that the heatpipe transplant did some good.

If xytech is right, it could be 100%. Also i've cleaned all the dust that i couldn't suck with my vacuum earlier so it might have given an improvement as well.

I'm not sure. I did quite bad of a bend near the fan (it's badly deformed but no leaks observed; probably i should've used a lighter to soften the metal before bending), and then wasted 4 hours trying to solder everything back since once i heat the cpu side enough, the fan side desolders too, and i don't have two clamps to hold them both together.

[L29Ah]

My current benchmark numbers btw, it would be nice if anyone with modded coolers presented theirs. The benchmark is cross-platform and crunches prime numbers, roughly representing CPU calculating power, so it should scale linearly with CPU frequency.

∞ sysbench cpu run --threads=9 --time=300 | grep 'events per'         
    events per second:  6821.08

[axur-delmeria]

I had an online discussion regarding heatsink mods, and one suggested using solder paste (for easier application) and a toaster oven (more controlled heating). I don't want to buy solder paste (requires storage at 10 degrees Celsius), nor sacrifice our household's only toaster oven.

What I do have is a spare flat iron. Maybe I can use that instead.
I can prepare solder foil by melting regular solder and letting it drop onto a metal surface, as seen in this video.

My objective is different from yours though-- I'm planning to mod an X220 heatsink such that the cooling fins are directly soldered to the heatpipe. The original design is that there's a sheet of metal between the heatpipe and cooling fins, and it's not made of copper-- this is what I suspect to be one of the causes of the X220's heat issues. The X230 rectified this by replacing said sheet with copper. I do have to put it back in one way or another, as said sheet is what holds the fan to the heatsink. I think I can still make it work somehow.

[axur-delmeria]

I'm not sure. I did quite bad of a bend near the fan (it's badly deformed but no leaks observed; probably i should've used a lighter to soften the metal before bending), and then wasted 4 hours trying to solder everything back since once i heat the cpu side enough, the fan side desolders too, and i don't have two clamps to hold them both together.

I've just succeeded in detaching and reattaching the cooling fins of a laptop heatsink using only an electric clothes iron.

I put the heatsink on a plank of plywood around 3/4 inch thick, set the iron to the highest setting, then put it on the heatpipe directly over the cooling fins. After around 5 minutes I turn off the iron and set it aside. As I lift the heatsink using a pair of pliers, the cooling fins are left behind on the plank, the heat having melted the solder.

You could fabricate a wooden jig that holds the entire heatsink in place. That way you only need to put the new heatsink over it, and the iron on top of htat.

When I tried soldering copper using a torch a few years back, I discovered that pieces need to be cleaned and prepped in order for the solder to stick. This doesn't matter much if both surfaces already have a layer of solder bonded to them, which is indeed the case in the tests I just did.

Soldering with the clothes iron is the same as desoldering, except that I keep the iron in place aften turning it off and unplugging it. This is done so as not to disturb the work pieces. In my tests, I removed the iron 5 minutes after turning it off.


EDIT
Looks like I spoke to soon. I checked the soldered parts after an hour and found that the solder joints were weak and broke easily. This means I need to replace the thermostat (or ditch it altogether) to reach higher temps.

[L29Ah]

Increased the output power by ~30% by adding a bulky 120mm fan to the RAM bay: http://vperde.l29ah.blasux.ru/dump/7367 ... oling.jpeg

Power plot, running mprime limited at 93°C (reported temperature is about 95°C), before and after: https://tinystash.undef.im/il/5QNQzQmbU ... Z7s6uEFYuG
You can see on that plot that any power benchmarks should be done on this laptop for AT LEAST 10 minutes to let the internal temperatures stabilize. My previous benchmarks and 51nb/xytech ones sadly lack this.

[atagunov]

deleted

[atagunov]

Would be glad if anyone with 17m19's cooler post their results.

Hopefully I should be able to later this summer.. It will be on a 10th gen CPU however.. and I only ever plan to run Linux - not sure what you guys are using for benchmarking

[flyingfishfinger]

I can run one. How should I measure power dissipation properly though? Currently on Ubuntu.

R

[L29Ah]

I can run one. How should I measure power dissipation properly though? Currently on Ubuntu.

R

I wrote the following haskell code to do it:

Code: Select all

import Control.Concurrent
import qualified Streaming.Prelude as S
import System.IO

toJoules :: String -> Double
toJoules = (/ 1e6) . (read :: String -> Double)

main = do
	hSetBuffering stdout LineBuffering
	S.print $ S.drop 2 $ S.scan (\(prev, _) new -> (new, new - prev)) (0, 0) snd $ S.map toJoules $ S.repeatM (threadDelay 1000000 >> readFile "/sys/class/powercap/intel-rapl:0/energy_uj")

[flyingfishfinger]

Do I need to adjust some kind of govenor? It seems I'm "only" running at 3.5Ghz instead of 3.9...

R

[L29Ah]

Do I need to adjust some kind of govenor? It seems I'm "only" running at 3.5Ghz instead of 3.9...

R

You may want to use throttled to adjust your thermal limits.

[flyingfishfinger]

This?

https://github.com/erpalma/throttled

R

[L29Ah]

This?

https://github.com/erpalma/throttled

R

Yes.
BTW you have a letter "R" stuck to your messages.

[flyingfishfinger]

BTW you have a letter "R" stuck to your messages.

It's my first initial

As far as the throttling, I can't seem to make it work right. My temps hover around 81-85C running "sysbench cpu run --threads=9 --time=600", but it still throttles to 3.5Ghz and dissipation is around 25W. I've set TDP UP in the BIOS as well as "Performance" governor, but it doesn't seem to stick anything higher than 3.5Ghz.

R

[L29Ah]

BTW you have a letter "R" stuck to your messages.

It's my first initial

As far as the throttling, I can't seem to make it work right. My temps hover around 81-85C running "sysbench cpu run --threads=9 --time=600", but it still throttles to 3.5Ghz and dissipation is around 25W. I've set TDP UP in the BIOS as well as "Performance" governor, but it doesn't seem to stick anything higher than 3.5Ghz.

R

Do you have the i7 CPU?
You can use `throttled --monitor` to see if it's a thermal throttling.

[flyingfishfinger]

Hi,
Needed some tweaking (there's no AC* in the /sys/bus/power_supply for example) but it does appear to be throttling, yes.

Temps reach 90C+ after a few minutes, but it does manage to keep the frequency at or near 3.9Ghz until that time. Now the dissipation is around 32W.

Guess I should go find my Arctic Silver and see if I can improve the situation a little...

R

[L29Ah]

Temps reach 90C+ after a few minutes, but it does manage to keep the frequency at or near 3.9Ghz until that time. Now the dissipation is around 32W.

This is awesome! What kind of cooling setup do you use? Do i get it right it's 20-ish°C outside?

[flyingfishfinger]

Hi,
I have a cool update (no pun intended...or, maybe). I replaced the stock thermal paste (some gold stuff) with my trusty (old) Arctic Silver 5 and re-ran the tests. The results are actually pretty good.

First, here's the cooler I have (the 3-pipe one) and the paste which was on there before I replaced it:





Then I ran the same test as before, 10 minutes of sysbench (see above) and monitored with "throttled --monitor". Amazingly, the machine NEVER ONCE thermal throttled and stayed at 3.9Ghz the whole time! Here is the temperature plot. It looks like it got pretty close to throttling though, so if it had run for more time it might have:



And it looks like the dissipation is now up to 36.5W!

sysbench returns 8615 events / second.

Cheers,
R

[RealBlackStuff]

That antiquated (by now) AS5 takes 200 hours to "settle in".
Instead I recommend Phobya NanoGrease Extreme Noctua NT-H1.
I used to like Kryonaut, but when that hits 80 C only once, it loses its capabilities!

[Kingair]

I used to like Kryonaut, but when that hits 80 C only once, it loses its capabilities!

Learned something today ...

[atagunov]

Instead I recommend Phobya NanoGrease Extreme Noctua NT-H1.

Not graphite pads? Just learnt they are now sold for the first time today.. Apparently people say that as long as you have high pressure..

[RealBlackStuff]

High pressure on CPUs does not apply to laptops...
More often they are held only by spring-loaded screws.

[L29Ah]

Meanwhile i've repasted my cooler with GD900. Output of https://github.com/l29ah/intel-powermon drawn with gnuplot during 10 minutes of `mprime -t`: https://tinystash.undef.im/il/3x6iqY1Kd ... 7aiuQzR7pX
That's it, 23W at the end with my crappily modified cooler at 26°C ambient temperature and 95°C CPU temperature (received via ACPI, 92°C limit set with throttled).