Which Thermal Paste to buy and apply (Traditional and Liquid Metal)

Gpus have always had significantly less thermal density than cpus. Sure maybe one day it'll reach a point where it reaches the same thermal density as cpus but at the moment its not as bad as cpus.

The stock 10th gen 8 core i7s seem to be able to hit 4ghz all core while consuming only 61w based on @Jarrod'sTech 's review. For reference my undervolted 6 core i7 draws 67w at all core. Not sure whats going on here.

Then you'll love ampere for sure. They are going with big dies on Samsung's super mature 10nm node (they will be advertising it as 8nm) instead of going with 7nm.

 

More information has surfaced. @Jarrod'sTech 's review has incorrect information.

The 8 core i7 pulls about 100-105w when all cores are running at approximately 4ghz



Kudos to @B0B

 

This sounds more realistic.

 

You are incorrect. sTIM+IHS is a hindrance, not a benefit, as it increases thermal resistance. Direct die cooling (die->TIM->coldplate) is superior to die->sTIM->IHS->TIM->coldplate. And at least on the Intel desktop side, the die thickness was increased to withstand the rigors of the soldering process and for longevity over years of thermal cycling, which increases thermal resistance even further.

 

Perhaps I should rephrase since everyone is missing the point.

Direct die cooling is superior, no doubt about that. The issue here is the use of paste.

What laptops are doing is direct die cooling using paste which may become a significant thermal bottleneck as cpu die sizes get smaller and their thermal density increases.

 

Liquid metal isn’t a cure-all for laptop thermal issues. Even assuming a coldplate with perfect contact with the die and LM, it may simply move the thermal bottleneck to other areas such as heatpipes, fin area/density, fans, or airflow.

 

Oh there will always be a bottleneck, but the most significant one now is the thermal interface material.

15-20°C improvement in thermals is nothing to sneeze at and represents a massive bottleneck.

 

15-20°C improvement over a quality thermal paste sounds difficult to believe.

 

Have you used liquid metal before?

 

Since 2014.

 

What results did you see then? On what laptops/desktops?

I've been using it for 7 years now and most laptops I've used it on gave me a 15-20°C drop. The only exceptions were macbooks which have atrocious cooling but at the very least they stopped throttling.

 

On my current laptop, and before that a Clevo P650SG and Lenovo Y500. It was at most a 10C drop, usually closer to 5C, compared to a good conventional paste. I've never heard of 15-20C drop unless it was compared to dried stock paste or something. Or on a de-lidded desktop CPU (where the improvement comes more from the glue removal than the LM).

 

Yes I was comparing LM to stock paste, not dried ones mind you, just stock paste on new laptops.

 

Reading comprehension...

 

Eh, small typo.

Anything else you'd like to say?

 

In any case

Your recent post already agrees with my statement about thermal density which is why I said liquid metal may be the only way forward to deal with increasing thermal density of cpus.

 

And I've seen plenty of examples of recent thin-and-light gaming notebooks where LM is not enough to keep the unit from thermal throttling, because either the manufacturing tolerances on the heatsink are too shoddy to work with LM, or the thermal bottleneck lay elsewhere in the cooling design. Like I said, LM does not in my experience perform worlds better than a good conventional aftermarket paste, so expecting something like a .7" 4lb laptop to be able to dissipate 100W of CPU heat is just ludicrous, LM or not.

 

Nobody is expecting worlds better performance but running at 80°C instead of 90°C is still a significant improvement over quality paste is it not?

And the thing with thermal bottlenecks is that the less bottlenecks there are, the better. And right now the biggest bottleneck is the TIM which can be easily remedied.

Of course, a beefier heatsink is always better but on thin and lights there's only so much you can do.

As for the laptops with bad cooling design, it's simple, just don't buy them. There are plenty of other options in the market.

Alienware laptops they have good cooling provided you get one with a good heatsink but as you said, they have QC issues with their heatsinks.

Here's an alienware m15r2 running an overclocked i9 9980hk at 4.45ghz and 120w
http://forum.notebookreview.com/threads/full-speed-9980hk-with-no-throttling-new-m15-r2.830612/

 

Knowing the limitations of your heatsink can help to make the right choice, in terms of what TIM to use.

 

Testing Thermaright TFX. Seems to be 1C better than Kryonaut after a few hours application. Various Japanese posters said it needs about 5 days of curing time. Some of the best results said it was up to 5C better than Kryonaut after. Consistency seems to be like Phobya Nanogrease Extreme. Definitely thicker than Kryo. Might be a good alternative for those who need a thicker paste.

I also ordered some of that ZF-12 paste from aliexpress for $9 and won't arrive for awhile. May have been a questionable purchase as there seems to be a ZF EX out now too. (Thermagic ZF extreme) but seems to only be available in South Korea and Vietnam

 

Liquid metal on a mag 15 only dropped temperatures on the CPU by 5 degrees compared to stock, and 2 degrees compared to mastergel maker. Additionally, it actually raised temps on the GPU by a couple degrees.

 

What temperature is this at? LM conducts heat better when temps are 70-80c, due to it's make up.
Otherwise it could either be contact issue or LM diffusing into copper heatsink. If you can provide a picture of the application it will be easier to comment on.

 

I haven't tried it myself, but it's from Bob of all trades review video at 10:40. It's going from 79C - 75C. Although I did miss the fact that the LM is about 500mhz higher, so the temperature difference is actually bigger than that.

Sent from my Pixel 3a using Tapatalk

 

I have to say I'm very impressed with liquid metal, I ordered some galinstan from eBay and it arrived today (thanks the_periodic_element_guy).

I was expecting to see a marginal improvement over IC diamond in my P375SM, blew my mind when the cpu temp dropped from ~90C max to ~80C max.

I believe this may be due to the IC diamond being too thick for this application, however in other cases it may do better.

This meant I could crank the 4710mq up to 1.185V and 4ghz (was at 3.7ghz before ).

Keen to try on the p870dm when I get my hands on it.

 

@Papusan @iunlock @Mr. Fox

Hi @ all,
I hope you can help me out. I used Coollaboratory Liquid Ultra on my pimped up Alienware M17x R4 which went very well for at least 9 months. Since a few weeks I noticed raising temps and decided to take a look and to my surprise I found a completely dried up/hardend/bakened layer of LM on the copper heatsink. Although I covered the CPU-Die as well with LM it was almost only found on the copper heatsinks-side. Only a few dots on the die which are dried up as well.
Now I tried to clean it up with 99% alcohol but to no avail. Should I use 4% hydrochloric acid as Der8Auer suggests or is there another way to get rid of the remnants?

Thank you for a short reply!

 

Hi brother. Use +2000 sandpaper and sand it carefully. Use quality isopropyl instead for water. A finish with (FLITZ) metal polish will make the die to shine (clean of with isopropyl after the metal polish job). Remember the copper will still be grayish aftewards but this doesn't matter for the cooling capacity.


Rockitcool offer this kit ROCKIT COOL CLEANING KIT (read the tip in the link).

Remember the greyish color on the copper is an huge advantage for your next Liquid metal job.
https://imgur.com/a/2xHnNVn

 

Thanks brother! So there is no other way to get rid of it than to sand it "away"? I'm asking because I have some hardened LM-spots on the die as well and I will never ever lay hand on it with sandpaper!

It's the first time that I had LM which "dried out" (that fast)....

 

Then clean/rub of with quality isopropyl then polish it with metal polish. Remember finish with iso afterwards. +2000 sanpaper carefully ain't dangerous. You can even get it with up to 3000 grit.

This is the results after I removed Intel's sTIM (SOLDER) on 9900K




Dried up that fast means not the best fits from HS on die.

 

Ay ay ay, caramba! Will do that, thanks!

What I don't get is: I applied LM on both sides (Die and HS). When I took the HS off the LM was almost completely on the HS side. So I would assume the contact must have been pretty solid otherwise I should have found it dried out on both sides, don't you think?

 

The copper will try suck up liquid metal as long it have contact with the other part. Tight and good pressure combined with already saturated copper will give best results for longeivity. The thermal cycling will open miniscule gap between HS and die with not perfect fits/good enough pressure. The winner will be the copper who will be filled up.

It could also be that you didn't got/get the LM to stick properly on die when you applied it. Often a problem with first time application (due the added protection barriere on the silicon).

Edit. Another option is to rub in liquid metal (Instead use of Quicksilver - Solder Remover) properly before you remove it/clean of again as mentioned above. In same way as we remove the Intel sTIM on 9900K die.

 

All right! Last question: as an absolute emergency solution for the next few days till Metal Polish and 2000er sandpaper arrives, is it "save" to just apply normal Kryonaut over it? Because I need the Läppi right now. Or will it explode right into my face?

 

Use what you have of paste. Clean of the best you can before you put on the paste What other thermal paste brand do you have? Why not put back LM?

 

You're AWESOME! Thank you for your support and advice!

 

I have ICD, Kryonaut and Liquid Ultra. I am afraid, putting additional LM on the dried one on top will put more mess on it as it already is.

 

Remember we use Liquid metal to remove Intel sTIM on Intel's newer chips. Then maybe go for ICD in the meantime.

Good luck brother. And if you have the time, post temp results after you have applied the thermal paste. Even better if you have old results to compare with

 

I forgot to mention. I have changed from Coollaboratory Liquid Ultra for many years ago. The same have many others + Silicon lottery. Coollaboratory haven't been equal good with quality the last years. Either pushed out many bad batches with Liquid Ultra or they have changed the recipe. I changed to Conductonaut, I mean in early 2015 and before SL's transition to same LM brand. As a heads up.

 

I have still two Liquid Ultra syringes left from 2015. That's why I thought it a good idea to use them still.

Gesendet von meinem CLT-L29 mit Tapatalk

 

Then use what you already have and re-apply new when the need is there. Just remember the great advantage with the greyish color on heatsink (less liquid metal will suck into the copper)

 

I'm considering buying the Coollaboratory Liquid Metal Pad - does anybody have any ideas?
My 10210U can sustain about a 33W constant loading under Prime95, whilst being pegged at 97C. NGL, was kinda hoping it would do better. I'm not really willing to go down the liquid metal route, as I want to be able to toss my laptop around without worrying about spillage.
I tried searching this thread but there isn't much info. Does anybody have any tips for using the metal pad? This is your chance to discourage me
Been using NT-H1 for a while and it hasn't seemed to degrade. The below screenshots are with MX-4 after a few months.

 

For such a low power part a high quality paste like mastergel maker should do fine. The bottleneck here is the cooling solution of the laptop.

 

Bitspower have introduced new thermal paste. Baircool TP-1 Thermal Compound (6.5 W/m-K thermal conductivity).

The new paste is rated among the less viscous compounds, so this looks like being a awful thermal paste for notebooks with its sloppy mounting pressure over the die/IHS. I expected something better than this from Bitspower. Official link to the product... Baircool TP-1 (2.5G/5G)


Bitspower Also Introduces the Baircool TP-1 Thermal Compound techpowerup.com | Today,

In addition to the CPU+VRM monoblock for the Z490 AORUS Master, Bitspower also introduced the Baircool TP-1 thermal compound. This is a graphene-based, non-silicone thermal interface material that boasts of being "long term stable," non-corrosive, electrically non-conductive, and does not require burn-in after application, with a short dry-out period. The compound offers 6.5 W/m-K thermal conductivity, and 0.009 K-in²/W thermal impedance. With under 600,000 CPS, it's rated among the less viscous compounds, and has a specific gravity of 2.5 g/cm³. Its dielectric strength is rated at 0.9 KV/mm. The Baircool TP-1 comes in a 2.5 g tube priced at USD $9.70, and a 5 g tube at $13.80.

 

Howdy everyone, I've got an idea I want to pitch and get some feedback on it.

I have an Alienware laptop with the very toasty 9750H in it. I've repasted it a few times, first with Arctic MX-4 and then again with IC Diamond. Thermals are okay, nothing great, and in order to keep them in line I've resorted to using TS and limited the all-core turbo from 4 GHz to 3.8 GHz. With a -.140 core and -.160 cache undervolt, I have temps in the low 80s under load.

My idea: Repaste with LM and create a high temp silicone RTV dam around the CPU and GPU packages to prevent spillage. I've noticed with this particular unit that the heatsink doesn't sit flush no matter what I do. I've got core to core differences of nearly 10 degrees and I'm guessing there's some sort of imperfection in the heatsink that's preventing a flush contact. If I seal the CPU and GPU with RTV, I can use the LM as a liquid buffer to fill any gaps inbetween the two assemblies and, in theory, not worry about it running out and frying anything because the silicone should create an airtight compartment around the dies.

I'm using Loctite gasket maker, which is rated up to 650F (peak) and something like 400F continuous.

Thoughts?

 

How easily removable would that be though? You might need to top up the LM if the initial amount is not enough. But otherwise your idea is viable.

 

To be honest, it seems like a lot of work.
Could it pay off by further dropping temperatures into the low-mid 70-ies?
Possibly, but considering your temps are already in the low 80-ies, I wouldn't necessarily bring liquid metal into the mix.
One of the ways to prevent spillage of liquid metal would be to use simple foam barriers around the cpu.
Short of that, you could try using Kryonaut thermal paste instead as it might be better compared to Arctic MX-4 or IC Diamond for that matter.

 

The RTV would be on top of the Kapton tape. It should peel right off.

In reply to @Deks, this particular 9750H hasn't had the best thermal performance. I'm currently running on a test bed of LM while I wait for the Kapton tape to come in. It's helped core-to-core temperature differentials to sub 10 degrees, but thermal performance still isn't stellar as in I can't run the all core boost beyond 3.8 GHz without limited throttling (advertised is 4.0). At 3.8, running Cinebench 20, it'll reach the low 90s, anything more than that and it throttles. I've tried Kryonaut and haven't noticed any appreciable difference between that product and ICD (other than ICD being rock frigging hard).

I'll swing back by in a week or two and update with how well, or bad it turned out.

 

Between Kryonaut and ICD, I'd suggest using Kryonaut in that case as I heard ICD hardening up before and it being an issue.
As for everything else... it seems like a problem with cooling assembly not being particularly adequate... then again, Intel CPU's aren't particularly good at maintaining their TDP seeing how those values are good for baseline, but not for boost clocks.

 

For an 8 core Intel, maybe. But an i7 9750h should hit 4ghz all core in stress tests without issue. Mine hits 4ghz all core in prime 95 small fft (@72°C) without issue and in a much lighter laptop to boot. How? Liquid metal

Alienware is notorious for their poor heatsink QC. The poor heatsink uniformity causes makes using LM the normal way unfeasible and using paste is also bad as the unevenness will accelerate paste pump out, in which case kryonaut is a much worse choice than IC diamond due to its low viscosity.

Alienware laptops also have a high pl1 and pl2 so the cpu can sustain 4ghz without needing to throttle down due to insufficient power so thermals are the only limiter here.

 

That's exactly what I've noticed when repasting my laptop. The heatsink and thermal pads are just crappy. The differences between cores before I used LM was horrible, like 10-15 degree differences and one core that consistently ran hot. Things are tidier now but I know I can do better. I figure if I can create a dam around the CPU where the LM can't escape, I can fill the gaps and create a space where it's a pure LM buffer between the heatsink and CPU die. The only thing I'll need to watch for is the copper reacting to the gallium; I figure this is something I'll need to repeat a few times depending on how long I have the laptop.

Either that or hope someone makes an aftermarket heatsink like I've seen for the bigger, older Alienwares.

 

Might be better to get a new heatsink, or at least try to lap or remove the bends of your current one, because a pool of LM wont stay between the CPU and the heatsink for much time at all..

 

@papusan,@Falkentyne, @Mr. Fox, @Fire Tiger, @Spartan@HIDevolution

I have made a comparaison between Kryonaut and Thermalright TFX on my GPU only. I had no overclocked.I ran CPU at lowest 36x on all test.
This is my first time using this Thermalright and i have to say it is extremly difficult the spread this paste on a GPU (thicker then phobya nanogrease in my opinion) I had to use a heat gun to spread it evenly on the GPU. Kryonaut was one week old and Thermalright one day old.

Here are the results:


I think (not sure) that thermarright runs about 1C cooler compare to Kryonaut. It will probably last longer as it compare to phobya nanogrease extreme.
If i had to choose between this paste and phobya nanogrease extreme, i would go with the cheapest one.