apply thermal paste

Cbshahji

[H]ard|Gawd
Joined
Jul 3, 2010
Messages
1,975
i know the title isn't much, but here it goes

how wise is it to apply the thermal paste on heat sink instead of on CPU? i'm planning on putting the thermal paste on H50's surface if it's same as applying on CPU, if not well i'll you tube some more videos to get it right.

thanks
CB
 
it doesnt really matter, imo. personally i apply it on the heatsink to minimize waste (paste that does not touche both cpu and heatsink), because it just attracts dust and stuff.

get a flat piece of plastic and apply a very thin even layer of paste on the heatsink (so you dont get air bubbles).
 
Or, you know, just plop a small dab in the middle (of either) and let pressure sort it out. :p It's how I do it as any other way has never shown any difference for me, period!

No fuss, no muss, no mess! (If you do it right.)
 
Pea-sized blob in the middle or line drawn over the actual CPU die works just fine, and saves a lot of muss and fuss. Pressure takes care of spreading it, and TIM is most important where the most heat is transferred - right between the CPU die and the base of the heatsink.

And no, no real benefit from putting the TIM on the heatsink instead of the CPU. It ends up in the same place regardless, and putting it on the CPU gives you more certainty of where the TIM will ultimately spread out to. Heatsink mounting might not perfectly align over the CPU, leaving your TIM off-centered and the potential for a hotspot. Low chance, but possible.

The heat spreader over the CPU die helps a bit, but it's still not going to yield even distribution of heat.

Depending on your CPU and where/how the CPU die itself is laid out under the heat spreader, either the blob or line method will work slightly better.

Something like a laptop which has an on-package graphics works better with the blob approach since there's more than one hotspot, which a decently sized blob will take care of.

A quad core desktop chip with an oblong CPU die (e.g. Sandy Bridge) tends to work best with the line method to make sure that TIM is between the heat spreader and heatsink for the entire length of the die.

There's no real improvement to be had from spreading a thin, uniform layer, and it takes a bunch of extra time (and mess!). If the base of your heatsink isn't flat, it can even be counterproductive. The whole point of TIM is to fill air gaps between the heat spreader and heatsink, and a thin layer can leave air gaps if the base is sufficiently warped.

Note: I'm saying "heatsink" here, but water block also applies (since a water system is still fundamentally a heatsink).
 
thanks for the razer blade tip i sure picked up one from work today will finish up tonight.
 
+1 for the small dab in the center


I've tried other methods and I ain't seen d*ck all difference from any of em' so that being the small dab in the middle and pressure sorting out for the win (and easy).
 
I've been spreading the TIM with a credit card onto the cpu heatspreader for a while. It works very well with MX-2 but is a pita since it's time consuming, messy, and doesn't work well with some pastes.
 
Well, I just swapped out my MX-2 which I spread a thin layer onto the cpu with a credit card like I have been doing for the longest time with some Shin Etsu which is supposed to be a better thermal paste. I used the pea method which people seem to swear by and I'm actually seeing slightly higher temps than I saw with the MX-2.
 
There are the 3 factors that affect proper application, spread, pressure, and contact. My guess is you didn't use enough and actually Shin Etsu is not forgiving and is very thin. IC Diamond is a lot more forgiving and so is MX-2 and will get contact where Shin Etsu won't on uneven surfaces. Take it back apart and look at what the cause was.
 
the pea method works awesome... with pastes with relatively low viscosity

shin etsu is pretty thick /high viscosity and as such doesn't work as well with the pea method.
 
Thin? Shin Etsu is a very thick paste especially compared to MX-2. I spread a super thin layer onto the cpu and saw better temps with MX-2. I think that mean that I'm dealing with a pretty even surface. This isn't the first time that I've tried this and spreading with a credit card has always given me the best results.

the pea method works awesome... with pastes with relatively low viscosity

shin etsu is pretty thick /high viscosity and as such doesn't work as well with the pea method.

I can't spread with a credit card either. It's too thick. I'm not too sure what people see in Shin Etsu. I have some MX-4, I may break it open and give it a shot.
 
My bad I was thinking of a different paste yeah Shin Etsu is very thick and hard to work with for sure.
 
I don't think it matters that much, more personal preference really
, like I don't see how it can make much of a difference at all
 
No, that video that Falcon linked to shows exactly why it should make a difference. I just don't know that I am seeing it. I wish that I had done a direct comparison with the two methods with MX-2.

Well I did end up trying MX-4. When removing the H50 I saw that I had used too much Shin Etsu. I ended up trying MX-4 and using a bit less than I had used with shin etsu. My temps were higher than the Shin even.

I actually removed the H50 again to see the mount and it was perfect. So I have my doubts about MX-4. One of these days I really should try some MX-2 with the pea method or MX-4 with the spread method.

OK, I actually reapplyed the Shin Etsu this time. I used the same amount that I did with MX-4 since that looked like that way to go. So far my temps are about the same that I saw with MX-2 using the spread method.

Ok so each test is the same. With Prime I had run small fft for 15 minutes and with Linx I ran the same test for just one pass which lasts about 5 minutes. Not exactly scientific but good enough for my purposes. Here is what I saw. This was the max temp reported on each core by Real Temp. This is a Core i5 760 at 4ghz and yes one of my cores runs hotter than the rest for some reason.

Broken in MX-2 spread

Prime
67 62 63 62

Linx
71 66 66 66

Shin Etsu too big of a dot

Prime
69 64 64 64

Linx
73 68 68 68

MX-4

Prime
70 65 65 65

Linx
74 69 69 70

Shin Etsu
Prime
68 63 64 63

Linx
71 66 66 66
 
Last edited:
i apply a rice-grain size of as5 onto the cpu plating, just as the directions say. for some reason i dont think it's enough as my temps are always about 8-10 degrees higher than expected with my setup, but w/e.
 
ya just some in the center should do it imho and the stuff usually need hours to cure before you really get the true results of the paste or what not you youngins are putting on there nowadays
 
Any excess TIM will just squish out you can use to little but not to much if you tighten the heatsink properly only so much can stay which would be the rtight amount any less and it won't work properly. Especiall if your TIM iss non-conductive. Now if you use conductive TIM you do not want excess squishing out to much but there is plenty of room for play with the sockets and locks because most of the excess squishes onto the edge of trhe CPU and the top of the lock on the socket.

And before you tell me you can use too much I was told this by a TIM manufacturer.
 
[H] is wrong the pea size drop is the best method. It works perfect with IC Diamond every time. Anybody wanna try it out for their GPUs or CPUs PM me your name, email, and address for a free sample.

HardOCP said:
When it comes to applying thermal paste we saw that different pastes respond differently to the various methods. A good rule of thumb when using a paste is if the paste is thick spread it out and if it’s thin in consistency let the heat sink do it for you.

Eh, pretty much more or less accurate. I just wouldn't fully spread it out.
 
[H] is wrong the pea size drop is the best method. It works perfect with IC Diamond every time. Anybody wanna try it out for their GPUs or CPUs PM me your name, email, and address for a free sample.

Yes, [H] is "wrong" because they didn't test your beloved ICD and that's the only one that matters.

And whatever your ICD mfg buddy says is truth for ALL TIM... I guess I should listen to "ACME CO." when they say they're the "World's largest sporting goods store" and vehemently insist that nobody else that claims the same is...

You're a tool that's trying to blanket ICD applications for all TIM no matter wat the viscosity when there's [H]ard testing numbers right in front of you. Seriously, get the ICD tube out of your ass lol.
 
Yes, [H] is "wrong" because they didn't test your beloved ICD and that's the only one that matters.

And whatever your ICD mfg buddy says is truth for ALL TIM... I guess I should listen to "ACME CO." when they say they're the "World's largest sporting goods store" and vehemently insist that nobody else that claims the same is...

You're a tool that's trying to blanket ICD applications for all TIM no matter wat the viscosity when there's [H]ard testing numbers right in front of you. Seriously, get the ICD tube out of your ass lol.

Dude you continue to attack me in ths forum trying to attack my credibility. You feable attempts continue to show your low self esteem.

This video has nothing to do with ICD and backs up all TIMs should be done pea size drop method to keep out the air bubbles: http://www.youtube.com/watch?v=EyXLu1Ms-q4&feature=fvw

Keep up your BS posts against me just shows what a bad [H] member you are and you need banned.
 
This video has nothing to do with ICD and backs up all TIMs should be done pea size drop method to keep out the air bubbles: http://www.youtube.com/watch?v=EyXLu1Ms-q4&feature=fvw

Yes... because that video tested ALL TIM's listed in [H]'s article. Once again, you can't blanket everything based on thee different pastes. So what do you say about [H]'s article then? Oh right, you think [H] is bullshitting us because you saw some random video on youtube that didn't measure actual temperature performance, much less with a CPU with a temp probe drilled into it's IHS... So far you've provided no quantitiative evidence that "pea sized" (more like half a pea or a grain of rice) method is better. Ya know what, I talked to a TIM mfg too, and they said the "line" method works best with their paste. Says so right in the manual, so OMG that must ALWAYS be the best method for every paste!

You feable attempts continue to show your low self esteem.

I'd have lower self esteem if I spent all day posting 60 times/day like you... That doesn't even include your other forums! If you're going to try to act smart at least spell the words right!
 
Last edited:
Yes... because that video tested ALL TIM's listed in [H]'s article. Once again, you can't blanket everything based on thee different pastes. So what do you say about [H]'s article then? Oh right, you think [H] is bullshitting us because you saw some random video on youtube that didn't measure actual temperature performance, much less with a CPU with a temp probe drilled into it's IHS... So far you've provided no quantitiative evidence that "pea sized" (more like half a pea or a grain of rice) method is better. Ya know what, I talked to a TIM mfg too, and they said the "line" method works best with their paste. Says so right in the manual, so OMG that must ALWAYS be the best method for every paste!



I'd have lower self esteem if I spent all day posting 60 times/day like you... That doesn't even include your other forums! If you're going to try to act smart at least spell the words right!

Actually your rudeness being unable to debate anything in a intellectual manner shows your lack of credibility and character.

Just like in any field new ways are dervived that are better than old ways period. That video is empirical data showing what happens to TIM based on the variety of installation techniques. The material used is represenative of TIMs base with whatever different materials added to the compound minus Indego Extreme.

Placing a Temp probe in an IHS has jack to do with good TIM application and is totally unecessary. If one uses the same tool to measure the temp no matter the tool like lets say RealTemp as long as the same tool is used with the different TIMs and the corresponding applications the results can be compared with scientific accuracy relative to the +-minor calibration the tool would need to be 100% accurate.

People who try to use spelling as a intellectual measurement are just grasping at straws to make someone look bad. Who cares. Are you the TIM and spelling police?
 
free samples ftw

Oh yeah for everyone just PM me your name, email, and address and post your results.

It seems some people cannot stand people offering [H] members free TIM and TIM application education. It just ruffles their feathers but they have nothing to offer but rude comments.
 
Back
Top