The Official MSI GT73VR Owners and Discussions Lounge

Remember use 2000+grit or finer sandpaper.
Use non-permament marker, draw a grid on the heatsink with ink.
Then sand the surface down to a 'polish'.

As you saw, there is probably a 'bump' in the middle of the heatsink. Probably closer to the colder cores. So you want a flat heatsink. You can test flatness with a GLASS TRANSPARENT base or platform. You can buy a small square of glass at any hardware store. Doesn't have to be large. Just not brittle or fragile. like a 6 by 6 inch square of glass. It's VERY easy to test heatsink flatnesss with that. You can use a tiny 'dab' (1 drop rice size) of toothpaste, thermal compound, anything in the middle of the heatsink, then press on the glass and watch for spread pattern.

then you can do what @Shinigami4444 did and test for flatness by measurement.

 

@Ivan994 @Shinigami4444 @GENOCID @Papusan @ZKST @J4ck974

Notice the difference between your CPU's and @Shehary's temps.

Notice that on his CPU, the hottest core is no longer core #2?
on his, core 2 and 3 are the same temp now. And in fact, core #1 is just hotter than core #0.
So he will never have the Core #0+#2 hotter than core #1+#3 temp issue you guys are having.
Because he lapped his heatsink so the pressure is now even on all the cores.

 

Why haven't people tried to stretch the pads right before they put on and screw down HS ? Worked well for graphics in Clevo (Be sure you have/use pads who survive some stretching).

 

Hi Master Papusan
Please read my post in BGA Venting thread.

Not a pad problem.
If there's a bump in the heatsink, it will cause uneven pressure. New paste=fills all the gaps. low temps. Then heat expansion and contraction+low gaps=core #0 and #2 temps explode again.
Stretching pads can help **after** you deal with the cancer BGA jokesink

 

Advanced modders (e.g. people who are used to sanding pentium 2/3 slugs back in the day before IHS's) may also want to try measuring flatness of the BGA chip. If there is any unevenness on the CPU slug itself, even if you make the heatsink perfectly flat, that could also cause uneven core temps. the CPU surface CAN be sanded, or rather, shined up a bit, if it looks rough, but you must be 10x careful with that than with a replaceable heatsink. there is also not much room to measure flatness, unless you had some very small square piece of glass that you could use, to apply a drop of toothpaste or basic thermal compound on the CPU, and then press down and see how it spreads through the glass. This should only be checked after you sand the heatsink down to a shiny smooth finish.

Typically, "polishing" the CPU surface should only be done to get rid of scuff marks, like if you had a mark left by uneven pressure from a heatsink (very obvious). I had a mark like that so I took polishing sandpaper and carefully sanded the top of the CPU until that mark was completely removed. As a general rule, just make sure the slug surface is fully smooth and free of any scuff marks, and focus on the heatsink instead. A good CPU sample should already have a pristine surface.

 

@Ivan994 @GENOCID

In addition on to above

Google "Lap or Sand HS guide", mostly you will see other people experience and guide how to lap / sand heatsink but only for desktop, which in this case is 90% useless, but what you can learn is tips / tricks...so read couple of guides

1. must remove thermal pads before sand / lap otherwise thermal pad will be wasted or particles of copper on thermal pads can cause damage to MB or VRAM if they landed on wrong place while installing back the HS

2. Remove the 4 screws, can be done via removing the E-lock (use nose plier, press the head of screw and pull the e-lock) , you can buy the e-lock easily but in process of lap / sanding, if you sand or lap the edges of screws, one screw damage means whole heatsink useless ( you can ignore this step if you are sure you can lap / sand the HS without damaging the screws but if you remove the screws sanding will be more easier).

3.Do not wrap the sand paper on your finger and use it to sand, fingers are soft and you never know how much pressure you are applying when sanding, so you may end up with more un-even surface.

4. as @Falkentyne said, get a piece of glass, 15 to 25mm thickness, glass is easy to get and only surface which is 99.9% flat so get (or make) one glass piece as the equal or less size (length * width) of heatsink copper and wrap a sand paper on it, tape the sand paper from other side and you will have your own small sanding kit, small glass piece gives you enough room to lap easily.

5. Patience is a virtue, as it's laptop HS so you have to hold it in hand while sanding and you have to go slow and steady and watch how much pressure you are applying, each lap has steady hands and pressure.

If you are afraid you can sand more copper (which you cannot, you are shaving the copper the cutting it) give 5 laps and use this method suggested by @Falkentyne to check your progress.

Good luck.

 

I actually don't think I even 'sanded' my heatsink at all now. I think I only polished it. Maybe or maybe not a little uneven since I did use my finger (no glass anywhere. I should improvise), but it's an IMPROVEMENT on stock. And the smoothness is much improved now and it looks shinier. My heatsink doesn't look sanded at all. Just very nicely polished. I think that's because I always use alcohol which limits the sanding power and just helps me clean the old 'stains' from the surface. It does look much better though. The stock heatsink looks so dull and rough.

 

Ok guys,
@Ivan994 @GENOCID @Shehary listen up

Shin is 100% correct.
I decided to REPASTE AGAIN.
And this time I decided to actually use a flat surface to sand the heatsink.
There was silver residue from the liquid metal and i first wiped as much off as I could with a napkin and alcohol, but the silver stain cannot be removed like that.
so I started sanding by how Shehary told us to do it.

After awhile, the very first thing I noticed was that the copper started appearing through the silver stain in the VERY CENTER OF THE HEATSINK. Literally to a T, exactly how @Shinigami4444 explained it.
The silver stain was in the shape of the CPU slug, but the copper started appearing right in the very middle!
In fact, it wasn't perfectly in the middle.
It was actually more towards a millimeter towards the bottom, pretty much exactly where you would 'expect' the two "cooler" CPU cores (if 0, 1, 2 and 3), cores 1 and 3, to be at !

So that hump is what is causing all these problems.

In fact, this hump is also the EXACT reason why, when you use prime95 with AVX, why the core differences between (0,2) and (1,3) get so much larger! Because any defect in heatsink pressure gets amplified when an AVX load is put on the CPU ! so something that would have a 1C core temp difference without AVX, might have a 3C difference with AVX.

Anyway I spent a good 15 minutes sanding down with the 2000 grit sandpaper (or 3000, it's so "fine" that I can barely feel the texture), and I kept sanding until all of the "stain" was basically removed, except a faint outline that wouldn't come out. And then I repasted.

Results:
Now 1C core temp difference at 4.7 ghz, but the temps were about 2C lower than before, with the 'minimum' temp from before, now being the new temp.
Not only that, now the 'hotter' core was now core 1 and 3, when usually they are the cooler core. Just by 1C at most. So that full sanding definitely made a positive diffrence.

So then I got real curious and went back to 4.4 ghz so I could run AVX.

sure enough.....
1C core temp difference WITH AVX.

Very happy and grateful that @Shinigami4444 found about that 'bump' in the center of the heatsink. That is the cause of probably most of your guys problems.

Do a full flat sand as @Shehary said, and all your temp problems should be solved.

BTW this is AVX small FFT, not blend.

 

@Shehary @Falkentyne Thank you very much for share with us your own experience and for tips. I am going to done it as soon as I finish work and come home. I am very grateful for all Your help.

Thank you again!!!

 

Here's another AVX run with the heater on so ambients are slighter warmer. Just did this, quick 5 min test.

 

A couple weekends ago I lapped my GT72 heatsink, repasted with kryonaut, and now, it never goes over 65ºC(but I'm running a custom fan profile, but the fans are around 60% PWM), pretty quiet, but my core 4 is always 3-4ºC cooler than the other 3 that are within 1ºC of each other.

I dont really care, its cooler, not hotter, and the heatsink did have a concave shape for sure, took me almost 2 hours starting at 400 grit to get it flat.

 

@Ivan994 pls let us know how u get on im gonna lap my hs in the next two three weeks

btw
i still have 4-5c difference only this never happened day after repaste in previous repastes... i also polished hs as shiny as possible but i doubt its because of that.... im gonna see temps next week other than that i never had it like this next day.. always at least 6-8c temps rised. i still doubt it gonna last long but ill see

 

@Falkentyne yeah glad it wasn’t just me that noticed that “bump/hump” in which senso used the correct term and that s that our heatsinks have a concave shape and is the reason it’s not fully flat. I’m sure it’s not the same for everyone but I think that’s why the stock paste is a big huge “pigeon poop stamp” like you guys called it because they probably always have minor deformities when they are molded.

I checked the heatsink again and it does have a concave shape, repasted after lapping a bit and temp difference went down to about 5-6 degrees compared to 10 I got yesterday. Definitely an improvement but I don’t know if it will hold out or I have to lap it a bit more(was afraid of lapping too much, don’t want to mess up the heatsink. ). Anyways you guys have definitely been helpful and hopefully I can get it to even a 2-5c difference.

 

So you haven't done any sanding or alterations to the heatsink surface?

That means your improvements in core temperature differential are due to careful re-pasting, reassembly, tightening the screw points evenly to get constant even pressure across the heatsink to get even contact across the surface of the CPU / heatsink.

I think that is the key to getting good results, re-assembly that carefully accounts for the mechanical contact and pressure.

My own experimentation done many years ago to perfect my technique in getting the best thermal exchange was found by varying the thickness of paste coverage - until it was an extremely thin all over coverage.

Thermal paste filling in the low points, but not heaping paste on the high points, a thin all over surface, only enough to make even contact across the entire surface of the CPU / heatsink.

Trimming the edge of paste all around to make sure no paste was exposed to the air was the last technique that extended the life of a paste job, no re-pastes have been needed on those computers nearing 10 years.

Too much paste in one area over another will cause compression variation and failure to get even contact pressure across the CPU / heatsink, causing one screw point or another to not compress as far with the same perceived pressure, causing lifting on the thick paste side. Kinda like too thick thermal pads, too thick / uneven paste can cause uneven pressure and contact that will vary over time as paste compresses over time and oozes out the edges.

There will always be variance in surface, paste evenness of coverage, pressure variance per attachment point due to assembly variance, mitigated during production with fixed torque settings per screw point.

The solution used in production using paste that resists drying out even if exposed to the air, along with even coverage of thick paste, evens out the pressure using the paste as a cushion, this is easiest in production due to it not requiring this fiddling with assembly you all are finding.

Uneven paste (pea, X, 5 spot, etc) and not knowing or being able to measure the pressure variances you are causing is requiring lots of blind testing to get back to out of the box core temperature differentials.

Really, this fixation on getting core temperature differential below what is considered normal, is a waste of time. As long as the hottest core is below 93c to stop thermal throttling, possible with undervolting and fan control, there is nothing to be gained.

The fallacy that reducing temperatures is going to extend laptop life is just that, a fantasy of the mind. These laptop CPU's will be running long after you sell your laptop and will never fail in use or during it's serviceable lifetime.

You aren't reducing temperatures for yourself, or the next owner, or the next, you are doing it for helping the CPU survive on the junk pile

And, thinking these reassembly problems are due to bad heatsinks is missing @GENOCID 's experience - he is seeing improvement into the low differential measurement simply by careful reassembly.

If you have a bad core differential temperature (>8c) out of the box even after disabling AVX/FM3 extensions in Prime95/AIDA64, and if after undervolting the hottest core is causing thermal throttling @ 4/8 core/thread x 100% CPU usage then return it.

Don't waste weeks on a problem that isn't worth wasting your time on, return it.

If you must fix it yourself, pay attention to pressure through the heatsink screw torque. You might be able to screw a couple of heatsink screws tighter - then measure the difference again - and adjust those or the other set of screws - and tune it that way, rather than dis-assembly and re-pasting.

You guys are going on for hundreds of posts over something that won't affect performance, causing problems by not knowing how to re-paste, re-assemble, and having opened that Pandora's box, are stuck following through for weeks and months chasing your tails for no good reason.

It's far better in the short term and in the long run to never have opened up and re-pasted, really it is.

Just undervolt, tune the fan curve, lift the back of the laptop higher than the front to start convection cooling, and enjoy your laptop, it's not meant to be a DIY project, really it's not.

 

ive done test where i seen heatsink wasnt flat so i doubt temps stay like this anyway for a week i think its cos i spreaded paste very thin layer and left screws not much tightened. but ull see in a week what a core temp difference will be.

 

That's what the paste is for, to mate the uneven surfaces of the CPU and heatplate.

A thin layer shows there is enough in the depressions in the surface but not so much it piles on the high points. This also keeps the exposed gap between the CPU and heatsink thin after assembly so that very little air can reach the paste to oxidize it.

As you perfect your paste spreading and re-assembly you are figuring out how to get better and better results.

Again, living with it as it is out the box with an undervolt, as long as the hottest core isn't thermal throttling it's going to work fine for a very long time.

Often I see guys get overly OCD over thermals, when it's not affecting performance or life span of the laptop, instead of enjoying their laptop they are a nervous mess. It's sad, really it is.

Just enjoy your laptop, you don't need to take it apart, really you don't.

@GENOCID you've already opened Pandora's box, and you've put in a lot of time perfecting your technique, hopefully this last effort will be the last you need to get results you can live with.

 

I never spread my thermal paste, just a thin line ending 2-3mm before the edges of the cpu and a small ball/rice on the gpu.

All heatsinks are warped, its a die cut piece of cooper that was soldered to the heatpipes, there is bending all over the place, its not a massive deviation from what a heatsink needs in order to maintain proper thermal contact with the cpu/gpu die, and the stock thermal paste is not great, but there is a lot worse than what MSI uses, where I was, Artic-MX4 was what was allowed to be used, or keep the pre-applied when it was a new heatsink.

A good repaste, clean all the surfaces with IPA and dont use too much paste is the basic.

Making sure that no thermal pads are overlaping or a corner is folded and making the heatsink sit in a crocked matter will impact temperatures a lot more than anything that can be done to an heatsink.

 

@GENOCID @Falkentyne @Shehary
I have sanded my HS. This is results, there are not perfect there is still 5-6 C difference but If this stay like this I will be satisfied. Maybe if I want to lower more, I need better paste then NOCTUA. I used DOT method again.

 

If he gets temps 20C higher again however he will need to sand or RMA, and I suggest both. If he gets 2 heatsinks, great. I'd love 2 heatsinks since i could run wild on modding.
HMscott, I know you're trying to help, but the problem is, your posts are referring more to heatsinks with IHS's (Desktop CPUs) rather than delidded or bare die contact.
Everything you're saying is 100% completely correct, IF we were talking about desktop heatsinks and desktop cooling here. (or laptops with LGA cpu's).

The problem is, you are using *direct die* contact, rather than an IHS.
An IHS has a very large surface, which spreads the heat around, making weak contact points not matter as much because the main heat transfer point is from the die to the IHS first. Then the heatsink cools the entire surface, although the center is obviously the priority, But there is room for small defects this way.

However, when you have direct die contact, you have the heat concentrated in a VERY small area, which then spreads after it goes onto the heatsink and the pipes. But the heat is absorbed directly from the cores, right there. Meaning ANY slight imperfections will now be magnified massively due to the 1) much smaller surface area, 2) direct die contact.

Both cause extreme stress on traditional thermal pastes, and most of these pastes were made for applying on a CPU IHS, not for direct die contact, because the temp stress is much higher.

It's the same issue people who delidded LGA CPU's had when they used Kryonaut for the delid repaste rather than LM. Kryonaut worked better than the intel pigeon poop. For a little while, it started out good, but then the Kryonaut RAPIDLY degraded. But when used on the IHS to heatsink surface, it did not degrade due to far less thermal stress and all the heat not being concentrated in a tiny area.

So yes, a concave heatsink is not going to be fixed by a repaste like that, because the defect has to be fixed. IC diamond would clearly work, but not lesser stuff.
see this post from someone with LM, just to see how even a slightly imbalanced heatsink can degrade temps:

http://forum.notebookreview.com/thr...3-owners-lounge.797884/page-687#post-10681313

 

@Ivan
look at your sanding work.
You did a good job sanding except for the VERY first core.
Notice that core is far hotter than the other 3?
Your core #0 should be around 67-68C.

Make sure you carefully work on that. Remember: that core is closest to the VRM's (top left corner) when looking at the CPU like this:



Remember where core #0 is on the heatsink:


@Ivan994
If you are going to work on the heatsink again, can you please do me a big favor and take a high resolution picture of your surface work? Hold it a little at an angle. Then take a second picture of it from "flat" so we can see the flatness of the surface. (if that makes sense). try to make them high resolution.

So the problem is core 0 still has 'slightly' weak contact.
When you sanded, did you use a flat surface? or did you use your finger? Also what "grit" sandpaper did you use?

You can start with 400 grit to "roughen" up the surface, and then switch to 3000 grit or 2000 grit. This makes polishing easier also.

 

hmm I hope It will stay like this. If stays just core 0 6 C hotter than others it is not big deal for me because I am not going over 4 GHz. I will report changes

 

That's fine. it was probably your first sanding attempt so its never going to be perfect. That's why you go slowly. It's actually sort of exciting, when you think you may have a surprise waiting for you to reward you for your work.

 

haha yes it is my first I just need to hold finger crossed to stay like this This is test after Cinebench. I have very good score for 4 GHz

 

I'm talking about all kinds of CPU / heatsink mating, and I am covering all of it as part of describing the important focal points to be aware of when doing a re-pasting.

Applying paste thinly to even out the contact surface, the force balancing at the attachment points, the variance of paste thickness.

All of those points are considerations under both IHS and bare CPU dies. There are different sized IHS surfaces as well as larger and smaller bare die surfaces. With the surface area of the heatplate contact varying in relation to the CPU surface area as well.

Reducing variance in assembly pressure is important on all surface area contact's, a little too much or too little pressure on one point or more will tilt the contact and ruin the thermal exchange.

The thinness of LM, with added thickness causing liquid pooling that migrates, is it's downfall.

When you spread thick non-conductive paste over an area, if there is a depression on the CPU surface or the heatsink surface it will fill it and not migrate. LM can't do this as well as non-conductive thicker pastes that don't migrate.

You can apply an otherwise thin coating of non-conductive paste over the entire surface, and the "pooling" of it in the low parts won't run or migrate, and even if it does oozes out and some of it drops onto the motherboard it won't damage it, because it's non-conductive.

That's why LM isn't recommended if the surfaces aren't evenly level and flat. In this case, like with Clevo GPU heatplates, avoiding LM altogether is the recommendation.

It's not *necessary* to lap an uneven surface if the paste can level the surface (paste on both CPU and heatsink) if the pressure is correctly applied evenly, it will compress and provide good thermal conduction.

If you had 2 perfectly flat surfaces you could mate them without paste, but the imperfections in the surfaces need the non-conductive paste to fill in the gaps and make an even surface for contact.

The never ending journey to get perfect mating and lowest temperatures is a never ending game unto itself, but it's not necessary to get good performance in use for 99% of the GT73's out of the box.

As long as you aren't thermal throttling, you aren't going to get any gains by lowering temperatures further.

The extreme tests used to find core temperature differentials are not intended to push people to fight a never ending battle to get a 0c difference.

Doing this test is to find gross thermal throttling problems that require returning the laptop within the initial return period of 7-14 days, and to find it before that time expires so the owner isn't forced to RMA, or waste time trying to fix it themselves.

Stretching it out into a never ending journey with no reward, with no performance improvements to be gained, is why I call it a waste of time and effort.

There are a lot more rewarding and joyful things to do with your new laptop.

 

I doubt any variance is due to sanding, it's more likely due to variance in flat paste coverage - focusing on making it enough but not too thick, with better focus on the assembly - better focus on evening the pressure on the contact surface by balancing the screw pressure.

@GENOCID focused on those details for re-assembly and reduced the core temperature differential without sanding.

Instead of sanding around that core, try loosing the screw opposite a half turn and tightening the scew near the core a half turn. Varying the pressure to improve contact is where I would focus to get the best out of what you have right now, without further sanding.

Consider the CPU as a small surface with the large heatplate surface rocking on top of the CPU surface, you want to vary the screw pressure to balance the heatplate evenly across the CPU surface.

Good luck

 

Well some people have time for PUBG and videogames, and Dark souls 2
Some have time for that and modding laptops and still get to enjoy them.

It's the same reason why people do 'silly' work on cars. It's a hobby. You want every last bit of performance. I understand that isn't for you, but you do have to accept that not everyone thinks like you.
some people just want to live a little and get under stuff. I do know your goal is to have people enjoy their system and not butcher them. But some people do wind up having issues, sometimes.

Now yes, if Genocid no longer gets temps rising then yes, he should just leave it like that and enjoy it. But I feel it's unfair for you to say that he should enjoy his laptop with a 20C core temp difference.
And yes, that may not matter at 'stock', true, if he's under 95C. But what if he actually wants to overclock? He's either going to reach PROCHOT, or get WHEA Correctable cache errors because of those 2 extremely hot cores.

 

If this temps stay like this after one week or so. I am not going to sand anymore, just leave like this and finally have fun with laptop Do anyone have maybe problem with bluetooth? My bluetooth does not work after last windows security update. As soon as I delete update, everything working well. I tried new drivers from killers site, but still no luck. I hate windows 10, every ****ing update **** up everything Is there option to postponed windows update for 1 mount or so? I can not find it anywhere :S

 

Sounds good

@Killer_Networking does @Ivan994 's problem sound familiar due to the current Killer BT drivers + newest Windows update? Any suggestions?

 

sorry, I just saw your edited post. I used flat surface, alcohol and just grit 2000. If I do, I will take pictures for you. I was sanding for about 10-15 min, HS was very bright orange almost like a mirror. Maybe I just repaste wrongly? I always use DOT method because I am not very good with spreading, always spread it wrong With DOT method, I always had the best results.

 

Thank You !

 

Don't use dot method.

just use spread.
Spread it evenly.

try that
take a picture of your heatsink for me before you repaste.
I want to see your masterwork

and spread is easy.
very easy.
Just pretend it's honey on a sandwich
very easy.
Don't overthink it.

Dot method is best for IHS's because the CPU slug is in the middle of the IHS.
for direct die, best to spread!

 

I will wait just a little bit to see could my temps stay like this. If not, I will repaste again and take a picture of everything

 

@Falkentyne Look now, there is small change even after 2 hours

 

Try the spreading....

 

Yeah, as long your Cpu "aren't thermal throttling". Many people start use their notebooks with stock clocks. Some with a very low overclock. 3-4 weeks down the road they can/will see that other with similar laptops/processors run higher clocks. They will be tempted of this and finally will start raise their own Cpu clocks. But damn... They are screwed No much thermal headroom for the wanted higher clockspeed they have seen other can manage. Not all run overclocked speed once they start use their new laptops. No thermal headroom with stock or slighty oc'd cpu is a bad start if you later want a bit more from your chip!!

I prefer the 3/4 long line. As long the chips ain't 100% square this will work. 100% square = X-cross method. Even two 3/4 long thin lines will work. Not all is able/manage to spread a even coat thermal paste on the DIE/IHS.

 

This is crazy !!!! Just 3 hours later from 5-6 C difference to 11!!! I will sell this crap of laptop and buy deskop. I am tired of this.

 

Please don't give up, I know it's painful but if you look closer you'll still see the "current" temps are 5c-6c different.

The Core 0 usually runs hotter in normal use due to more single threaded things running - something else could have kicked in when you were testing.

What are you running to test? Prime95 / AIDA / Cinebench with AVX/FM3 off?

Those max temps don't look very high, so maybe you aren't 100% CPU load testing?

If you aren't 100% CPU testing, then normal usage will tag peak temps on each core that won't be representative of a core temp differential test, they will vary all over the place.

Only the pure 100% per core CPU load on all cores / threads can be used to show core temperature differential. And, make sure to clear the monitor readings just before you start the test, and grab an image just before the test ends so there aren't other things running polluting the readings.

Also, try using hwinfo64 (sensors only mode), it's got more detailed information, logging which I encourage people to use to see the second by second readings rather than rely only on Max single peak readings, and you can click the clock timer that shows run time just before starting a test to reset the readings - to clear previous max readings.
http://www.guru3d.com/files-details/hwinfo64-download.html

What were you using for testing?

 

I am just mad because I lost 2 hours for nothing I will try again tomorrow for sure. It was 100 %. This on the picture is cinebench, I was testing with prime 95 also it is the same, I can not see any difference. Prime 95 is with AVX/FM3 off. I will try also tomorrow not to use DOT method. Maybe I will try with line or spread method. I just want core difference under 10. Nothing more, do not need to be 1-2. Tomorrow rhis difference will rise again, I am sure about it. If it rised for 3 hours from 5 to 11 tomorrow it is going to be 20 again

 

That's wise, fresh eyes and a fresh outlook will help a lot

You are hitting your limit on enjoyment, now it's work, frustrations lead to wanting to give up, but that's just your mind saying it needs a break.

The Kryonaut paste itself should be stable at a wide range of operating temperatures. Rated stable use range is -250c / +350c so it's not pumping out due to temps, and oxidation shouldn't happen that quickly unless there is some huge open route for air to get in.

Something that helped a long time ago with someone that only had MX-4 and K5 Pro when doing laptop CPU re-pasting was to use a small ridge of K5 Pro all around the CPU, use your finger with K5 Pro on the tip - and use the edge of the CPU to scrap it off - creating a ridge around the outside perimeter edge of CPU.

When screwing the heatplate down that ridge of K5 Pro squishes in and out around the CPU, wipe off the external over squish, and the K5 Pro solidified providing a protective perimeter seal of K5 Pro around the MX-4 to keep it from drying out.

That solved their MX-4 drying out problem.

They had weeks between temperature changes though, not hours or days. MX-4 dries out the easiest I have seen. I find it hard to accept that Kryonaut is drying out faster.

What I do is use a business card or razor blade, or Xacto Knife to scrap away about 1/16" around the CPU after applying a thin layer of paste so no paste squishes out to act as a route for drying air to get in.

Another hint, after initial spreading paste over the CPU I use a business card or credit card, one with flex I am comfortable with, to gently scrape thin layers of paste off the CPU - until I am happy the coverage is as thin as I can get it will 100% coverage. It can take a few trys to get a happy coverage the first few times you try this, but over time I can do it with a couple of passes of the card the first time.

Anyway, you'll develop your own technique over time.

Oh, and the pea / X / 5 dot methods work better on high pressure fittings - like on desktops with huge heavy coolers or heavy spring loaded fittings. Laptops don't have that, and it's low pressure, so a mound of paste can act as a lump that won't flatten out completely, which is why I suggest spreading a thin layer instead.

Hope that helps

 

Thank you mate for support. I will try tomorrow to sand little bit more and to use your tips for repasting after. I will try to spread nice and easy with some card. I will report result for everyone like I did today and explain what I did. I hope, I will have good tems and cores under 10 C difference. I will be more than haopy with that and finish this repasting thing finally

 

Cool, remember it's supposed to be fun

Considering you are using Windows 10 that can start background tasks all the time, you might want to run TaskManager during the runs to see if another process kicks off and starts loading Core 0 or Core 1 during the run that can skew the results.

I turn off all automation, anti-virus, close browsers and other apps, I'll even go so far as to disable services I am not using while doing benchmarks. Anything starting off while you are benchmarking will cause skewing of readings.

I usually make several runs, with cool down periods in between - I let the fans cool the metal down to the idle temps between runs - and reset the hwinfo64 after the run starts, like 10 seconds to erase the initial spike readings, and using snipping tool take the reading a few seconds before the run completes.

You'll often get wild readings on occasion - Outlier data - that you'll want to toss before doing averages, the same for single runs like you are doing.

Again, as long as you aren't thermal throttling you aren't losing performance, try to keep that in mind. Your temperatures even on the hottest core isn't getting anywhere near 93c - the thermal throttling point.

Maybe game with it for a few days, and then get back to it? Let the paste and the mating settle, maybe it'll get better?

 

There is no sound for my GT73's speakers.
Several days ago, when I plugged in my headphone, the window that ask whether I plugged in a headphone will not automatically pop up. So I uninstalled the realtek high definition audio driver in device manager and the windows 10 automatically install it for me. But there is no sound in the speaker after that. If I plug my headphone to the 3.5mm port nearest to the USB port, it output sounds (but the system didnt detect I plug in a headphone and still shows using speaker). But the 3.5mm port nearest to the user(where I used before the reinstall) cant output sound.
So what is the problem? I have tried to uninstall the driver, reboot and reinstall it several times. I also tried using the one downloaded from the MSI GT73 official website.

 

Try to delete last win 10 update. I have similar problem like you just with bluetooth. System do not see my bluetooth at all. I deleted last update and everything work perfect. Just try to delete last update to see if is that or you need to search for something else.

 

Actually the last 2 updates were installed on 2-14, my speakers works fine after that day, and the windows10 only allow me to uninstall one of them.

 

And even I uninstall it, the windows 10 ask for a reboot, and it will be automatically reinstalled after reboot

 

• I think the problem is my computer detect the speakers as headphone(the nahimic show it as headphone).

 

In the Sound Control panel can you manually select the speakers vs the Headphones as output?

 

I really do not know how to help you. If you try to reinstall Nahimic and drivers, I do not know what other to suggest. Wait for somebody else here, they will help you for sure

 

In Sound panel, it show I am using speakers and the green bar vibrating, but there is no sound in speaker.

 

Try to look here, in sound playback and see if is speakers set as default, if is not set it.