Thermal paste on laptop CPU/GPUs (avoiding pumpouts)

Hello,
I've recently tried using Arctic MX4 and my experience has been almost identical to Noctua NT-H1 several years ago.

Within the first month or so performance is exceptional then after that ~1 month mark, cooling performance drops to worse than stock unknown thermal pastes.

Upon opening up the cooling system, I always find all of the thermal paste has been pushed out from where they need to be (pump out). In such cases, I seem to always reach for Arctic Silver 5 and I am more or less happy for the following 3~5 years.

More or less what this guy is showing : but for both Noctua NT-H1 and Arctic MX4.

Above have been my experience for Lenovo T480, T430, T420, T410, W530 but Arctic MX-4 on my Alienware 2015 seems to be holding just fine even after 2 months so far.

I know there are many resources available online regarding thermal paste for CPU/GPU, but they don't seem to apply very well for laptops. Has anyone done some extensive tests with thermal pastes exclusively for laptops?

 

Isn't it better to stick with other thermal pastes such as GC Gelid Extreme and Thermal Grizzly Kryonaut?

 

I've actually never tried either of GC Gelid nor Thermal Grizzly Kryonaut. Have you tried them on laptop CPU/GPUs?

I've tried at least 10 different brands but I always come back to Arctic Silver 5 because they are cheap, abundant and reliable. I would never go as far as liquid metal but I am willing to experiment but I wish I had some positive results.

 

Personally, I believe undervolting cpu/gpu and properly setting all other parameters in ThrottleStop can be more effective than new paste itself. I use GC Gelid, seems to be working fine, but again IMO this is just part of the set up, then comes TS settings and fans, most laptops have fans set to spin at too high temp to keep quiet, but that really makes everything run very hot. Even best paste won't help if your fans are not spinning and pushing the heat out.

 

Pete, I am using my laptops for performance programming and gaming. Sound is the lowest priority and undervolting is not an option at least for my case.

I usually attach an external fan to the laptop vent to help with the heat dissipation and the thermal paste does make a huge difference in my case.

I should also mention, both on linux and W10, I tweak fan profile to hit max speed at 60 degrees Celsius.

 

Undervolting is not an option why?
Its free performance and less heat generation that you are just throwing away.

 

Because first of all, I say again, performance programming, meaning using full CPU and GPU loads. I have a desktop counterpart with more power but I have to synchronize its clock speeds and thread uses on my "drafting" laptop.

Second, My Lenovo T480 for example has an i7 8th gen core ( https://ark.intel.com/content/www/u...-8550u-processor-8m-cache-up-to-4-00-ghz.html) with 1.8Ghz base clock without boost. Boost gets me up to ~4.0Ghz, to undervolt this CPU is like mixing water with the fuel for your car engine to keep it from overheating. 1.79Ghz for example is completely unacceptable for anyone who lives in this century and does anything more than browsing the internet.

Third... This thread is about preventing pumpout, not why we should under-utilize our machines. I undervolt my toughbook, but again, this is not the topic of this thread.

Before we go further off topic. I want to figure out if others have figured out ways to prevent pumpouts, and what kind of experiences they've had using specific thermal pastes. For me, Arctic MX4 and Noctua NT-H1 are out. Arctic Silver 5 is the only functional choice, and I am willing to try other pastes.

 

Grizzly Kryonaut is a bad choice if you experience pump out with most good thermal paste
http://forum.notebookreview.com/thr...ers-welcome-too.810490/page-803#post-11037924

Make sure that the heatsink fits perfectly on die. Warped/uneven heatsink is the main culprit for this behavior.

Try out ICD or Phobya NangoGrease Extreme. Probably the best choice outside fixing the heatsink problem. As well undervolt for lower temps.

 

@SJLPHI, you are obviously not informed about the benefits of undervolting. Especially with ThrottleStop by @unclewebb.

 

I used Artic-MX4 for years, never noticed it pumping out unless the heatsink was crooked due to thermal pads being not the correct size/stacked pads, or outright a bad heatsink.
I can confirm the Noctua NT-H1 on laptops sucks, it pumps out pretty fast.
I have had good results with kryonaut, on my lapped laptop heatsinks, never used it on stock heatsinks so cant say if it will pump out or not in stock heatsinks.

Undervolting makes the CPU run cooler, how can it be like adding gas to a fire?..

Also, getting a U CPU and expecting that it will run at max boost clock is wishful thinking.

 

If you have issues with pump out, perhaps a thermal pad might be the way to go.
Bob of All Trades tested and approved.

https://www.thermal-grizzly.com/produkte/6-minus-pad

Possible. Just got to find a way to keep it cool. Even with under volting and sticking this laptop in a freezer, it would still thermal throttle. No surprise when you are trying to run it at 150% over the 15W rated TDP.

https://i.imgur.com/eo0J744.png

Less voltage = less heat = more performance. Well worth looking into.

 

I know this thread is not about undervolting, but you are getting totally different concepts confused. Undervolting does NOT decrease clock speed or performance: it can only increase it and/or reduce temps. What you are thinking about is underCLOCKING which is a different thing entirely.

Undervolting means your CPU receives less voltage to achieve whatever clock speed it's going for. This means less heat, which means the CPU might be able to aim for a higher clock speed for longer. Of course, if you go too far then you will experience crashes because your CPU can't function at such a low voltage, so you have to see how far you can push your undervolt via trial and error.

The reason this works is that by default both Intel and AMD set the voltage at a super high value that is guaranteed to be enough for all the CPUs they manufacture, rather than individually testing the voltage that each individual unit actually requires. So this means that you, as the user, can find out the voltage that your individual, unique CPU unit actually requires, and undervolt in order to set the voltage to that, instead of the default overkill voltage.

Please research this topic better. It's free performance gains. It does NOT reduce clock speeds. You can combite undervolting with a good repaste for optimal performance.

 

Please be aware that undervolting causes instability and incosistent outcomes when doing any kind of performance programming. To safely undervolt, one must underclock.

Will you please do some research on what "performance programming" is before commenting?

 

I see, sorry

 

I'll disagree with you on that. There is room in mass produced CPUs for undervolting with zero instability/performance loss. Intel doesn't run every CPU on the bleeding edge because there would be a chance for errors. The silicon lottery is a real thing and some undervolt better than others.

 

Yes, that's the problem. It's a lottery. As a programmer, I cannot take chance of my program misbehaving due to hardware error. I do understand that there is usually a relatively large safety margin on how much one can undervolt before reaching the instability/artifact generation but I want that chance to be as low as possible even if that involves "overvolting" as the manufacturers have configured them.

I would be a very sloppy programmer if I inserted any form of undefined behaviour in my code that somehow worked on my laptop but not on other computers.

I appreciate what everyone is trying to suggest, but I am not going to explore that option at all. Can we please get back to the topic?

 

Maybe try already suggested solutions. You don't have many other choices as it is now. See post #8

 

I intend to. I am also thinking about thermal pads. I've had some really good experience with 16.6W/mK pads on my toughbook.

 

From my bad experience of alienware 15r3 uneven heatsink issue my suggestion is you use ICD thermal paste and Gelid thermal pads.. The most important part of the thermal pads is that they are squishy/soft so when you press down the heatsink it will have better contact.. hope this helps..

 

I am a programmer too, working on parallel computing with CUDA and HPC too, and the only question I want to raise is: why on earth are you doing 'performance programming' on your laptop, and your desktop? Haven't you got a terminal to SSH into and work there, or some form of virtualised instance with a supercomputer?

At any rate, I also work with low-level systems, and no, undervolting does not impact stability, provided that said undervolt is not too extreme. Again, experimentation is needed, but without loss of generality, it is fair to say that undervolting provides a universal benefit.

 

I draft on my laptop then send the main program to the desktop to handle the brunt end of my program. I secure shell into desktop from laptop and I have limited licenses and legal agreements limiting the number of cores I can use.

Undervolt DOES impact stability and artifact generation if your code depends on racing conditions. I question whether if you actually have worked on parallel computing if you say there is no loss of generality from undervolting your devices. Are you programming at the systems level, or some sort of python script?

In any case, I am not sure if you understood what "pump-out" is and this thread is to address the pump-out effect, which can and will happen even if the CPU is undervolted, underclocked or even toothpaste is used as a thermal interface.

 

Yes: C and assembly, and as already mentioned, CUDA.

No, not Python.

I don't suppose the passive aggression was necessary. At any rate, undervolt or not, your choice, but I was under the impression that if you wanted exact performance metrics, you'd program, debug and measure performance on the production machine itself. Your moving from laptop ↔ desktop already negates any 'synchronisation' you are attempting, because the hardware is so different. I certainly do parallel computing, and the general consensus appears to be that it is extremely difficult to guarantee equal performance, even on two identical machines.

Moving back to the pump-out effect, the largest determining factor appears to be heatsink pressure, which is already quite small to begin with on notebooks. One way to mitigate this is to use the thermal pastes that are known to be really thick, which are Phobya Nanogrease, and IC Diamond (thickest of the lot). IC Diamond does have the potential to scratch the silicon dies, though.

 

Okay good, C and assmebly then we're on the same page. Off topic, both my laptop and desktop are pushed to 4GHz and they are 8 threaded quad cores using the same speed and capacity RAM. Drafting of course does not really apply for using Cuda/OpenCL but main chunk more or less is synchronized and is relevant. I hope this is the last time I read and write about anything other than the topic at hand.

I am sorry but I am rather being purposely offensive because even the title says what the discussion is about and more than 50% of the comments in the thread are more than completely irrelevent to the topic at hand. I much rather be offensive than to be off topic again.

Regarding pump-outs, I appreciate your input. Have you tried Phobya Nanogrease or IC Diamond on laptops? There is a general lack of information regarding thermal paste specifically for laptops and I am most interested in personal experiences because many reviews such as ( https://bestcpucoolers.com/laptop-thermal-paste/) is obviously written by people who never tried the pastes themselves or wrote a single word about pump-outs.

 

IC Diamond, yes. It is highly abrasive, so if you want to use it, be careful when applying/removing it. It contains diamond particles, which will scratch your die and heatsink if you are not careful. Another option are the Panasonic Soft-PGS/IC Graphite pads, or the Thermal Grizzly Carbonaut carbon pads. Both of these are solid, and hence would not be subject to said effect at all. Their efficacy is a controversial topic.

Still unnecessary: discussions with several human participants tend to be that way, and meander through several interrelated topics as the discussion progresses: it is natural, and becoming offended or acting passive-aggressive does not help. A polite 'may we please stick to the topic' would have sufficed. Furthermore, undervolting would certainly affect the pump-out effect, as your CPU would thermal-throttle less, and achieve peak performance at lower temperatures, thus reducing the thermal expansion/contraction that your heatsink and CPU would undergo, hence pumping out the thermal paste less.

 

I see, my local suppliers actually do offer IC Diamond and I may try them... After the warranty expires on my laptop. Do you know how often they need to be reapplied? I've actually read about the Graphite pads, I've heard more or less bad about them but no personal experience.

True, but I was trying to be polite and even went as far as to explain why I am not interested in undervolting in the beginning and the discussion kept on going far away from the topic.

Also, true reducing thermal generation = less thermal fluctuation = less possibility of pumpout. It is also true that thermal expansion/contraction caused by a few degrees Celsius is negligible in comparison to tens. Meaning, the best possible undervolting may give you ~5~10 degrees Celsius, going from 90 to 80 degrees Celsius, which is a negligible thermal change compared to room temperature to 80 degrees Celsius. Which in turn means that undervolting has negligible effects when there is already a pump-out effect caused by >40 degrees Celsius difference when the CPU is under load then set to idle.

You can disagree and or feel offended by anything says about anything, but if being polite has not worked, being offensive often does. I wouldn't go as far as Linus Torvalds but he gets his point across... fast and effectively.

I will say it now politely again. Let's stay with the topic of thermal paste we have personally tested on our laptops and if and/or how we resolved the pumpout issue.

 

It is a known fact that Linus Torvalds, Richard Stallman, Joel Spolsky, Eric Steve Raymond (well not this guy so much) and similar persons are certainly very intelligent, and have produced high-quality software, but are scumbags, and have very little EQ whatsoever. This isn't the 1980s or 90s where sheer intellect allowed one to brute-force his/her way through discussions.

Furthermore, the reason why people here were determined to press undervolting, was because you were (and still are, to a certain extent) being patently wrong about it. It is also why I asked my initial question: if your workload is highly time-dependent, hihgly hardware-dependent, highly real-time, and highly synchronised, there ought to be zero reason why you are even drafting code on your notebook, and not programming directly on the production machine. Merely having the 'same clock and memory frequency' is insufficient, and as a 'systems guy', you ought to know this.

Furthermore, undervolting is a tried-and-tested method to improve performance while improving thermal performance, simultaneously. It's a free performance upgrade.

As for the graphite pads, yes, the consensus is that they aren't as good as the best thermal pastes, but they are average, and will continue to stay average, unlike pastes, which, well, pump out.

 

In my personal experience, Kryonaut is best for laptop CPUs.

Phobya nanogrease is certainly best for thermals, but suffers from extreme pump out on laptops. I have to redo the paste job every week with nanograse on my Ranger's CPU. That's how bad it is. It's not good for applications on small dies, but does well on larger surface areas such as on GPU dies and a desktop CPU's IHS.

Kryonaut on the other hand does not seem to pump out, or at least does so very slowly. The only problem with Kryonaut is that it degrades if subjected to temperatures over 80°C for a long period of time.

In short, my goto is Kryonaut for laptop CPUs, and Nanogrease for GPUs and desktop CPUs.

 

Ionising_Radiation

All I can say to you is that you proved my point. As I was politely asking you to be on topic yet again, you go off and brought your "expertise" on something you have absolutely no clue about. You don't know what projects I am working on, and I don't care to explain. By the time such message is conveyed, the earth would have stopped rotating.

I am not sure if you understand what thermal expansion and contraction is but pumpout is not an artifact for thermal throttling. A small +/- 5 degrees Celsius fluctuation throttling is negligible compared to +/-40 degrees Celsius idle-load cycle in terms of thermal expansion/contraction meaning pumpout is not only not limited to thermal throttle, it is simply property of improper thermal interface.

The discussion at hand is thermal interface for laptops, and undervolting will not solve the problem at all because pump out will happen regardless whether if I under/over/stock volt my cores. I recommend you to take some physics classes for future reference. For the time being, please learn the language.

Pumpout = thermal paste pumpout = !(performance gain from undervolting).

Also:

Reputation/ What you have heard = !(your personal experience)

 

See below:

I responded in kind as you continued the tangential discussion, too. You have also continued with the ad-hominems and passive aggression (r.e. taking physics classes: I have majored in that, too, so I know what I'm talking about).

Even so, I concede that my response above was sharper than necessary.

Moving back to pump out: you can also consider bending your heatsink plate, or sanding it down to a fine finish, as other members here have tried. Swapping out the heatsink screws might help. Pump out is a function of heatsink pressure, which is already low to begin with, in notebooks.

 

Let me ask you a question. Have you actually tried any of the suggestions in this thread? Lots of good thermal pastes were suggested, checking for a warped/bent heatsink was suggested. Other than that there are no other solutions to your pumpout issue (it's a function of viscosity and heatsink fitment). If you are unwilling to try them then I guess there is no point in any of the suggestions. Typical thermal pads won't work on a CPU/GPU, but the carbon options will. Even though they aren't as good as a good thermal paste, it won't pump out.

 

custom90gt.

No I have not, but I have a list of pastes and pads that I intend to try thanks to some of the releveant posts.

Also, none of my thermal system are bent as far as I can tell.

 

Typically they are only subtly bent and you need something that is very true to see the bend. My only word of caution is if it is bent be very careful should you choose to rebend it because the heatpipes are EASY to bend/pinch.

Also Toms hardware has a list of thermal paste viscosity:
https://www.tomshardware.com/reviews/thermal-paste-comparison,5108-12.html

 

The carbon nanotube pads will perform even worse with a bent/crooked heatsink..
Can be as simple as thermal pads that force the heatsink to not seat parallel to the die.

By the way heatsinks are created(soldering heatpipes into cooper plates) they are always uneven due to the stress caused by the heating and cooling, I gained around 8ºC just by lapping my GT75 heatsinks..

 

I personally have tried Phobya Nanogrease, its (really) good the first few days, then it degrades fast. I cant seem to find IC Diamond at reasonable price (almost 30€ for the paste with shipping)

Other pastes I have tried were mx4 and GC Gelid Extreme.

 

Thermalright TFX doesn't degrade like Kryonaut does, is excellent for sub-zero, and doesn't pump out very easily either.

 

I'm gonna have to try out that paste then. I have a whole collection of thermal pastes going now .

 

Is this the one?
https://www.amazon.es/dp/B08CGPD9RN/ref=cm_sw_r_cp_apa_i_4LMpFb8V4E4Q2

it feels like every month this forum seems to recommend a different paste lol

 

Yes.

 

We have a lot of knowledgeable people on these forums that constanty find new awesome things for us to try.

 

How is your experience with this paste?
Phobya NanoGrease Extreme decreases performance pretty shortly on my current laptop.

 

Either your heatsink is bent, or its not making proper contact..
Kryonaut lasts 1 year in my laptops and here everyone says that it lasts 2 days..

 

Here buy this

@Papusan @Spartan@HIDevolution @Mr. Fox @jc_denton

https://www.amazon.com/dp/B08F78P2J3/

 

We've already ordered it for testing just a couple of days back: https://www.thermal-grizzly.com/en/products/359-kryonaut-extreme-en

 

Is there any software that will show me a graph of my cpu temp over time, min and max is no good to me as i want to see the temp my cpu drops to when the fan kicks in.

Thanks.

edit: I downloaded the 30 day trial of ArgusMonitor and works great.

 

Grizzly dropped the notice "halts the drying out process at temperatures of up to 80° Celsius" for the new paste. They probably found out that such statements won't boost the sales

The new paste is most likely thicker and mixed up for extreme LN2 cooling. In same way as Kingpin KPx. Fancy colors sell.

GenericLogViewer from Hwinfo. Free and not trial software.

GenericLogViewer - Version 5.3

 

Either that or Thermal Grizzly improved their Kryonaut paste using a new formula, and it no longer suffers from thermal degradation issues.

This new Kryonaut Extreme paste looks very similar to Thermalright TFX specwise. Visually, it kinda looks like the glue inside pink elmer's glue sticks.

 

Well, Arctic Silver 5 may be old and reliable but pumped out after ~1 month mark on my T480. Next up : https://www.digitec.ch/en/s1/product/thermal-grizzly-cryonaut-1g-1250wm-k-thermal-grease-5614395

I checked with some laser tools and microscopes, no bends in the cooling system.

 

What about Thermalright TFX ?

 

I will get to them eventually... Basically the current idea is to try different ones as one paste fails and I am going for what's most easily available locally first which happens to be cryonaut.

Besides, it takes at least a month to make a fair judgement of each paste.

 

I'm curious what you mean by you checked with "some laser tools and microscopes" and didn't find any bends? If your paste is pumping out your heatsink is either not level or does not have enough pressure (most likely a combination of both). What does your paste look like if you do a test mount?