Thermal paste performance degradation

Hello everyone,

I've had some troubling thermal issues with my Clevo P650 for a while (laptop cooler, CPU undervolt, maximum fan profile - 90°C GPU when gaming, and GPU clock throttling), so I decided to repaste with a Noctua NT-H1 paste. It went splendidly, after the repaste, the temps with the same circumstances were around 79-80 °C, with much higher GPU clocks.

The repaste was done at the 17th of January, approximately one and a half months ago, and yesterday, I did some benchmarks, and overall, just monitored the system under the same circumstances as before, and my temps quickly rose to 88-90°C, with lower clocks than expected (though not as bad as before the repaste).

I cannot really think of any major differences that has happened over the course of the 1.5 months that could cause this temperature increase. Dust buildup shouldn't be a problem, since I cleaned it before testing.

What could be the cause of this temp increase? Could it be that the paste dried out? Or it's too liquid?

 

http://forum.notebookreview.com/thr...nd-apply-traditional-and-liquid-metal.806840/

Go to the above link. We (Yes, including me) found that, whilst some thermal pastes performed well on desktops with high mounting pressure, some pastes like MX-4, NT-H1, AS5, were garbage on laptop CPUs due to the lack of mounting pressure. Initial results might have been good but once the CPU has been jostled around the thermal paste might goop out and goop back in, causing air gaps.

 

So what's the consensus with laptop thermal paste? Is there anything like the manufacturers stock paste available at reasonable prices?

I tried MX-4 some years ago during a CPU upgrade and it just pumped out after a few months, I tried Céramique 2 and that seems to have held with good temps but its not as thick as the OEM stuff (my laptops don't really leave the house much so they don't see lots of transport stress, YMMV).

At this point I think its better to sacrifice a few degrees in temperature for paste that'll give a longer lifespan and resist the stresses of the laptop being transported, but I've yet to find any on places like amazon.

Most of what I find online is aimed toward desktop processors with integrated heat spreaders rather than the bare dies in our laptops, there is a deliding community who deal with bare dies in desktops but still desktops don't really move or flex as much. Then you see all these people re-pasting their GPU's which also have bare dies, but there never seems to be follow ups 6 months down the line with pump out reports.

tl;dr What readily available thermal paste is most like the thick OEM stuff for bare die laptop use?

 

There isn't a paste that can last very very long since you need to repaste every 2-3 years for best thermal performance.

 

Is there a tl;dr version of that? I don't really have time to go through an entire multi page thread and I bet most people coming across my post won't either.

 

Click the spoiler on first post to get extended list of thermal paste and its updated regularly based on user feedbacks.

 

I don't really care about the best thermal transfer, I think the OEM stuff is good enough and was chosen to give ok thermals and resist cracking/pump out when subject to the stresses of laptop life.

I have one laptop that's 10 years old and has never been re-pasted or messed with and its still fine thermally.

 

And not forget Uneven/warped heatsink (a well known problem on notebooks). Combine this with low mounting pressure and you will see a very fast degradation of low quality thermal paste (pump out due thermal cycles).

 

Hmm these look like run of the mill desktop pastes at first glance with emphasis focused on thermal conductivity numbers for marketing, have these people actually reported back years down the line without repastes?

Not trying to discredit you here but maybe we're just re-pasting for different reasons, personally I only re-paste when absolutely necessary such as removing the heat-sink to unblock vents or processor upgrade. IMO the OEM phase change materiel is much better suited for the laptop environment compared most stuff I've found on amazon.

This^^^ My parents 12 year old Dell Inspiron 1501 has a very noticeable warp in the middle from over a decade of daily use. When I removed the heatsink to upgrade the processor the stock phase change materiel actually looked ok and hadn't pumped out at all, which is pretty good when you think of all the stresses and flexing its been through to end up bowing down in the middle.

Like I said MX-4 didn't last long in there and whilst with ceramique 2 the temperatures haven't shot up like they did when MX-4 pumped out, I haven't opened it up to look so I assume its ok 3 years on. I'm sure MX-4 is perfectly ok for desktop processors, but I haven't tried it.

 

id say the most high performant fire and forget paste would be IC Diamond. its a bit less performant as Kryonaut or GC Extreme, but it lasts basically forever and its viscosity is great to balance out warped / uneven heatsinks

 

Most OEMs use 3M's thermal paste aka IC diamond like Paste. I think the composition of CM maker gel nano is similar unlike ICD it doesn't etch the die surface.
Actually thermal paste gets dried or worn out when you overclock or use turbo boost quite often so Hot->Cold-> Hot cycles degrade the paste.
Which laptop are you repasting? Those paste mentioned in the list can be applied on desktop,laptop,ps3,xbox or even raspberry boards.
I can only recommend Gelid extreme, CM Maker gel nano, Phobya Nanogrease and ICD if you're unsure about heatsink being warped or uneven.

 

Best pastes to use for very bad heatsinks is IC7, but this has been proven to be abrasive.
Other good paste is Coolermaster Nano Maker, but even this will dry out slowly with a bad heatsink, although slower.
Warped and weak mounting heatsinks kill all pastes besides IC7. If you can mod your heatsink (sanding, adjust thermal pad thickness) for a fully secure balanced mount without gaps, even if low pressure, then liquid metal compound would work. But then so would most other pastes (Kryonaut, Phobya Nanogrease, Maker Nano, Gelid Extreme), although LM would last the longest if the heatsink is fully balanced, flat and no warps or uneven pressure. Just do what Vasudev said, or try fixing your heatsink flatness and balance first.

 

please consider that the reported "abrasiveness" of ICD causing scratches on dies does in no way affect either performance or functionality of the chips its applied on. i havent heard / read or seen a single case where hardware was actually damaged by ICD application so this is a purely cosmetical flaw

 

It becomes a problem when someone decides to put on LM and eventually it will slip off and cause mobo to be dead.
@Falkentyne @Papusan @Mr. Fox are clearly the ones who have tested most pastes.
I saw @judal57 etched CPU after applying ICD 7 the mark is still and makes LM to run off because of it. He had to use nail polish and 2-3 coats of resin to protect against LM run-off.

 

Thanks for this, I think it might be useful to add this list, or some version of it to the pinned thermal paste post, in case someone is looking for a paste for laptops with bad heatsinks.

For my case, my laptop came with Kryonaut, then twice repasted with NT-H1, and all these cases ended up with getting quite bad performance in a few weeks, so I think my heatsink might be warped or really low pressure, even if I didn't notice it explicitly when repasting. I'm going to try either gelid, cm makergel, or phobya, does anybody has any concrete recommendations from these 3? My goal is good performance, for at least 6 or preferably 12 months.

 

I think you got Clevo from Obsidian PC right?

 

lol omg thats insane. what kinda scratch was that?! when it comes to CLU, conductonaut or pure gallinstan, the fluidity / viscosity is more than enough for the paste to not just "run off" because of some superficial scratches unless ure talking Phobya LM, which I dont recommend anyways because its fluid as f*** and thus runs off no matter if you have scratches or not ^^

 

Just applying ICD, 7 or 24, does not cause etching. It is usually the removal because of too much force. If done properly I know it is fine. I have used it multiple times on soft copper shims with no markings, you just have to take care, and agreed I am just a bit more of a fuss pot here. While nothing lasts forever on a proper installation after 1 year I pulled it to look a the condition and it was like brand new, not dried out pumped or any thing else.

 

Juan got bad batch of icd 24 that etched the surface.Mind you, he is perfectionist.
ICD is best with one minus point for me and very expensive at my place.

 

Sadly no, now wish I had, at the time I just tried to find a reseller closest to me, and found one in germany with a really good deal, I got mine from them. I suspected they might have applied the paste badly or something, since the temps were a bit unusually high even after getting the laptop. Why do you ask?

 

Was curious, that's all!
If you read previous posts, I posted a link for best thermal paste in the market.

 

Use nail polish (3 coats) on the SMD resistors around CPU/GPU to protect from LM getting on them and shorting them out and destroying or making the solder melt and come off.
This will NOT stop LM runoff from getting on the PCB around the CPU/GPU or motherboard. To stop that, you use foam dams.

Foam DAM for BGA cancer CPU:


Foam dam on LGA CPU and MXM GPU:
This is @Mr. Fox 's picture, not mine!

 

Where can we buy FOAM dams?

 

It's regular packing foam for fragile items. Any hardware store or even a hobby store should have it. Not sure if the word is Polyurthane or not however.
Here is one type of foam that is compressible and will work.

https://www.amazon.com/gp/product/B01MXDJFCB/ref=ox_sc_sfl_title_2?ie=UTF8&psc=1&smid=AEF8AWXR95D0

Here is how the fabric of the foam should look like:

https://www.amazon.com/Reticulated-...rd_wg=NuLsv&psc=1&refRID=JJ0JNJBEB3R9854HF6YN

Here is a picture of BAD (too thick) foam on the left and good (highly compressible) foam on the right.


Here is more "bad" foam that is far too thick and not compressible enough:
https://www.amazon.com/Length-Neoprene-Cosplay-Projects-Gaskets/dp/B01JSWJTNY/

 

Here's two types of foam dams that will also work(i have the black one on my GPU right now actually).
The black version is a little denser than the grey version, so the starting thickness should be thinner on it before you "cut out" the shape of the CPU or GPU inside of it (like surgery). The main difficulty is, if your starting sheet is already thick, trimming it without destroying the cutout is pretty hard and annoying. Probably best to make sure you get the right thickness (thinner is better) when buying. On cancer heatsinks, the less resistance you have, the better.

 

Here is how a completed foam dam would look like if done for a BGA CPU. You "could" put this in a delidded LGA CPU too, between core and IHS, but there's absolutely no point, since foam dams are NOT to be used in place of nail polish on the SMD resistors, but in combination with it (if LM touches exposed SMD's, it can dissolve the solder and make the SMD fall off, or short out SMDs and damage or destroy stuff; cellulose based nail polish (transparent) protects things really well), and the re-sealed IHS acts as its own dam to stop any stray conductive balls of doom from getting on the mainboard. However for nickel plated or aluminum external heatsinks, you would use "square" external foam dams like shown in @Mr. Fox 's picture above. picture.

BGA foam dam (i just cut this out this morning so it looks nicer than the first one, and the foam is a bit denser so its easier to trim):

 

I would go for ICD. Follow up how it works if you go for this paste. It's a reason some OEM's go for this thicker thermal paste.

Edit. Too much talk about Liquid metal and foam dam become more as derailment for OP if he is not very experienced regarding thermal paste. Start with the safe and see how it works first.

Some Info so it don't make it too confusing for others. ICD7 and ICD24 is same paste but in different package size. No point in buying a more expencive package ICD24 if you don't repaste many machines or do it very often. If the price differrance is huge... Buy the cheapest package.

 

i have had like .... 3 slips off and the mobo make 2 beeps sound. i had to disassembly, clean and repaste. but my mobo is not dead

 

That result is luck of the draw. Some shorts could just be where a resistor, etc., is no longer functional and the device fails to initialize and post other can be catastrophic shorts taking out the main board and other components. Glad to hear yours was not the later.

 

How bad can the cooler mounting be for MX-4 to fail quickly.

I've had MX-4 installed in my T61, since I installed the T9300 in there... 3 years ago, and I run the machine regularly, and I load it. I run an aggressive fan curve on it, but it still has never cracked 80'C. It idles in the 30's, and under usual use, it's bouncing around in the 40-50 range. The little NV GPU keeps it running warm.

T420s and it's NVS4200M, has had MX-4 for a year, same deal. Qosmio G30, with it's T7200 and it's 7600GT, same deal, that MX-4 has been in there for a long time, since I got the Mobo replaced by Toshiba for the NV solder debacle, and then I had to get the cooler mount bits from a supplier since Toshiba, reused the faulty mount stuff. The WD 320GB blacks died with 6 months left in their warranty... So almost 5 years.

Wife's desktop, won't crack 50'C with it's X3350. Her little Sony, super-netbook style of machine, pins the clock speed on that little AMD dual core, and the cooler spits hot air, with no signs of thermal throttling, and it's been 4 years since I swapped the stock garbage out.

I've had great luck with MX-4, and I've installed it in many many places. Parents desktop, their HP Elite, friends systems, many GPUs, chipset coolers... Things I'm around so I'd know if they were cooking themselves.

I'm just shocked to hear people are having issues with it.

 

MX-4, and other TIM's, will usually do well where the temps are not ever getting that high. It is the systems that breach the 95c or constantly hit temps creating thermal throttling that usually have an issue. Needless to say it has to get pretty hot for boil out to occur.

 

On desktops MX4 will do. On laptops MX4 or similar will give you worse performance than stock paste after 2 weeks or less. You are free to try it.

 

Not sure but for me MX-4 it is ok, for laptops too, I used in last 2-3 years...

 

I used MX-4 for a while in my old Acer 5930G... and I have to say that I had much better success with Gelid GC Extreme as the temperatures were better and I didn't have to repaste since using it.

 

I just used my T61 to do my taxes on. It's had MX-4 installed in it for over three years, when I swapped out the T7300 and upgraded it to a T9300. Without any helping and just using the stock curve in TP-fan control, under Windows 10 Pro 1709 x64, running windows updates, and edge with lots of tabs, it never got over 55'C. The NVS140M also means I can't idle temps low worth a damn on it without the undervolt bios, but for fairness I was running stock volts.

T420s is only a year. The T7200 in my Qosmio G30 is actually I think 4 or 5 years old, and the fans on it almost never rev up. I also have the factory throttling on the CPU disabled. Silly Toshiba setting the bios to throttle 50% all the time...

Yeah, me too, which is shocking. I mean the only time I had a bad result was on a Sony TB33, since the northbridge chip needs a thermal pad and MX-4 can't make up .5mm of a gap, its been in there for... 3 years too, and the fan still blows hot air, and that thing has a really lame cooler. Though I guess for the thousands of installs, a few are bound to get a bad tube, or have a bad mount.

 

I notice a lot of people who have had success with MX-4 are using it with old CPUs such as Celeron/Pentiums with high TDPs of 30-40W. I believe people with modern Core U, HQ, HK, (4th gen and above) will have a lot less success with MX-4. With modern chips, idling CPU power can be lower than 1W, but max load package powers can be much more. Whilst the TDP of these processors, 15 for U series and 45W for HQ and HK have been lower, so has cooling standards from laptop manufacturers. When OEMs design a laptop with a specific CPU, they will design the cooling for the TDP limit ONLY. For older CPUs, the operating power range might be

There are a lot of enthusiasts here who like to improve the performance. I've seen 7700HQs unlocked from 45W to 60+ Watts. My 15W 8250U can be unlocked to 40W. The issue arises when the petty cooling solution in laptops is overloaded by enthusiasts, and MX-4 is no good when temperatures rise high. I once had an AMD E2-1800 which idled at 50C and never rose above 80C, even though it was a 18W CPU. I also once had a Core i7-6500U, which had a TDP of 15W but when unlocking the TDP the CPU would rise to 99C. All the older CPUs don't appear to have TurboBoost. TurboBoost is basically automatic overclocking. If I ran my 6500U at its base speed, the temperature wouldn't go over 70C. If I can my 8250U at base speed, it wouldn't go above 60C. TurboBoost is SUPPOSED to only work for short periods to give a kick in performance (such as opening Word, Photoshop, or browsers) but a lot of enthusiasts here raise the TDP of their chips, allowing them to Turbo indefinitely, as long as the cooling allows. With TurboBoost, TDP doesn't mean anything anymore.

 

Just a note: Turboboost in notebooks has been around since the first gen Arrandale/Clarksfield CPU's - which was eight years ago now. While some people might be running older CPU's than that, I'd say that it's safe to assume that the sizable majority of users are running CPU's with TB. There was also a form of dynamic overclocking in a previous generation CPU, which had a different name, IIR. EDIT: It was called Dynamic Acceleration (IDA).

 

I was using ceramique 2 paste on my vaio with t7100 and core 2 duo desktop which ran great and had lower temps. Fast forward now, those paste(s) perform poorly since heatsink is flimsy and temps worsen after each day.