Best way to apply thermal compound?

[CitizenSnips]

Going to be building my first rig pretty soon, got the last of the parts in and I just need some time off work (and my 64 bit Vista disc to be shipped) and I'll be ready to go. But I'm a bit confused about the application of the thermal compound to the CPU/heatsink.

I have the Arctic Silver 5 and the Zalman CNPS9500LED for the record. On Arctic Silver's website they advocate using the "use BB-sized dollop then press the heatsink down" method, but the instructions of the Zalman show the "apply compound then spread evenly" method.

Any input on which one ensures better and even coverage between the CPU and heatsink? Online build guides I've seen mostly seem to use the spread method.

 

[Idontcare]

See the thermal past review at Madshrimps. The data are pretty convincing that spreading your paste is not the way to go if low-temps are your objective.

http://www.madshrimps.be/?action=getarticle&articID=635

 

[JEDIYoda]

Originally posted by: Idontcare
See the thermal past review at Madshrimps. The data are pretty convincing that spreading your paste is not the way to go if low-temps are your objective.

http://www.madshrimps.be/?action=getarticle&articID=635

You follow the manufacturers instructions to the letter!!

 

[JEDIYoda]

Originally posted by: CitizenSnips
Going to be building my first rig pretty soon, got the last of the parts in and I just need some time off work (and my 64 bit Vista disc to be shipped) and I'll be ready to go. But I'm a bit confused about the application of the thermal compound to the CPU/heatsink.

I have the Arctic Silver 5 and the Zalman CNPS9500LED for the record. On Arctic Silver's website they advocate using the "use BB-sized dollop then press the heatsink down" method, but the instructions of the Zalman show the "apply compound then spread evenly" method.

Any input on which one ensures better and even coverage between the CPU and heatsink? Online build guides I've seen mostly seem to use the spread method.

The themal paste that shipped with the Zalman 9500 provided you bought it new comes with thermal paste that is as good or better than the AR5!!

 

[tofumonster]

Ah yes. The age old debate.

I, personally, do the following:

-clean both the heatsink and CPU with Alcohol and lint free cloth
-place a blob of AS5/MX-2 the size of half a grain of rice on the CPU
-plop down my ACFreezer
-lightly rotate the heatsink to get all the air bubbles out
-secure the pins.

For the amount of TIM to use (if this is your first time), the general rule I follow is this:

If you think you think you've pushed out enough thermal paste, its already too much; if you think you put too little, you've put enough.

Kinda abstract...what I'm trying to get across is that too much creates bad thermal transfer (but also 'not enough' is bad too, however you're more likely to put too much than too little on your first go)

 

[CitizenSnips]

Thanks for the input guys, especially that little tidbit you gave me Tofu. I'm generally the kind of guy that ends up going a little overboard, so I'll be sure to reign it in and remember to keep the amount small.

 

[Synomenon]

Arctic Silver has new instructions for the Intel's Core 2s and Core 2 Quads. You put a thin line of AS5 down the middle of the CPU.

 

[CitizenSnips]

Originally posted by: IsLNdbOi
Arctic Silver has new instructions for the Intel's Core 2s and Core 2 Quads. You put a thin line of AS5 down the middle of the CPU.

Thanks a lot for the heads up, just checked AS's website, that makes a lot of difference!

Well that settles that, press method it is. Thanks for all the input guys

 

[CH3VYMAN]

I have always heard that you should clean both the cpu and the heatsink then apply a small portion in the center of the cpu. Then use a plastic bag or plastic glove to cover your index finger then spread the compound evenly to fill all the microscopic holes on the surface. Make sure it is a super thin amount then mount the heatsink.

 

[WaTaGuMp]

I look at it this way, even if you use the spread method once you attach the HS its going to press the paste anyhow.

 

[PolymerTim]

I have seen all three methods (dollop or line and press, spread with baggie over your finger, spread thin with a flat edge like a credit card or razor blade) described by various groups with varying results. I think in the end, what you want is the thinnest layer of TIM that fills all gaps between the CPU and the HS. If you think about it though, there are a lot of variables and possible pitfalls to consider.

The dollop (or line depending on your proc) and press is the safest way to ensure you don't trap bubbles. Air is going to be far worse than just a thick layer of TIM, so I think that is why this is the preferred method from vendors. The problem is that it is difficult to get a very thin layer of TIM while pressing, especially with the thicker TIMs.

I think the spread-it-yourself method could work very well (speaking from logic and not personal experience) if it is done properly. But I think it may be harder to do properly than most people realize. Let's think about some crucial points/pitfalls.

Spread it thin - I read the article posted above a few weeks ago and the picture with the TX2 just amazed me. If you're going to spread the TIM that thick, you might as well use the dollop method and avoid the risk of trapping bubbles. I think the whole advantage of spreading your own is that you start off with a layer thinner than you could press down to, otherwise its pointless. You can easily get full coverage with the dollop method.

Spread it flat - OK, you've just spread that goop with your bagged finger and now it has little hills and valleys all over it. Once you put you flat HS down on it, where do you think the air in those valleys is going to go? Seems like a flat edge is really important for smoothing the TIM as flat as possible before assembly.

Flat HS base? - I've heard a lot of people testing their HS and finding the bas eis concave before lapping. Even if you have a perfectly flat layer of TIM on your CPU, it seems like a concave HS base is going to cause problems trapping air. If you know you have a concave surface, I think you have to either lap it or go with the dollop method.

Not too shiny - I can't find it now, but I recently read an article where I guy actually varied the level of polish on a HS (for a laser in his lab, but the same kind as for procs) to see the effect on thermal transport using the same TIM as computers use. To their surprise, they found an optimal roughness (using about 800-1000 grit sand paper). Putting a mirror finish actually degraded performance a little. The hypothesis is that the micro grooves created by the 800 grit sandpaper were able to push through the thin TIM layer and make direct metal-metal contact.

After all these pitfalls/requirements I've thought up, and reading a lot of reviews on the matter, I have trouble finding a comprehensive test that takes them all into consideration. I don't really trust a lot of the results out there comparing methods because they rarely take all these things into account. All that being said, you might notice that it is much simpler to go with the dollop method. Its no wonder that most manufacturers recommend it.

Tim

 

[DerwenArtos12]

I to this day use the spread method with no problems to date. I usually use a plastic card, safeway membership card actually) to do the spreading and am very careful to go both directions and use some real force to make it a good THIN layer. I also like this method when first installing a new heatsink because I can remove the heatsink and check the contact pattern. I have no intention of changing my method any time soon but, the dallop/line method seems more I.D.10.T proof.

 

[PolymerTim]

If it ain't broke, don't fix it! I agree completely. There are a lot of things that might go wrong no matter what technique you use. In the end, I say if it gives you the temps you want, then I don't care if you use your elbow to spread it!

All this talk about TIMs and mounting reminds me of another popular philosophy from my childhood:
"If it ain't s'posed to be movin' and it is, then use Duck tape. If its s'posed to be movin' and it ain't, then use WD-40."

 

[DerwenArtos12]

Originally posted by: PolymerTim
If it ain't broke, don't fix it! I agree completely. There are a lot of things that might go wrong no matter what technique you use. In the end, I say if it gives you the temps you want, then I don't care if you use your elbow to spread it!

All this talk about TIMs and mounting reminds me of another popular philosophy from my childhood:
"If it ain't s'posed to be movin' and it is, then use Duck tape. If its s'posed to be movin' and it ain't, then use WD-40."

I shudder at the thought of WD40 anywhere near my computer. Most people don't realize what WD-40 was really designed for. WD-40 was designed to displace water and it's the 40th formula they tested. Water Displacement formula 40. People just decided to use it as a lubricant which was a horrible idea as it collects dirt and dust really well.

 

[zorrt]

Normally I would go with spreading. Spread a thin layer on both surfaces then bring the 2 together.

Recently I read some guy's review on whether the spread or dallop method was better, and his result showed that the dallop yeild better results. So I decided to give it ago and remove the heatsink on my video card and reapplied AS5 using the dallop method. Unlike his result my result remained the same with no increase or decrease in temps. HOWEVER, spread method takes time as you need to apply on both surfaces and spread it out. Dallop method just takes 1 second and you're done. So in future just for convenience I'd most probably stick to the dallop method.

 

[Jhhnn]

Meh. Back in the day, some pastes had pretty thick consistency, and the heatspreaders large, so it was important to not use too much- the spring tension might not squeeze it out properly, leaving too thick a layer for best results.

AS5, in comparison, has a very thin consistency, so just putting a small dallop in the middle and depending on pressure to spread it is a good strategy.

 

[PolymerTim]

Originally posted by: DerwenArtos12

Originally posted by: PolymerTim
If it ain't broke, don't fix it! I agree completely. There are a lot of things that might go wrong no matter what technique you use. In the end, I say if it gives you the temps you want, then I don't care if you use your elbow to spread it!

All this talk about TIMs and mounting reminds me of another popular philosophy from my childhood:
"If it ain't s'posed to be movin' and it is, then use Duck tape. If its s'posed to be movin' and it ain't, then use WD-40."

I shudder at the thought of WD40 anywhere near my computer. Most people don't realize what WD-40 was really designed for. WD-40 was designed to displace water and it's the 40th formula they tested. Water Displacement formula 40. People just decided to use it as a lubricant which was a horrible idea as it collects dirt and dust really well.

Hehe, yeah. That quote kind of pre-dates computers. I don't see to many people using duct tape in computers either (although electrical tape and zip ties seem commonplace).

The only thing I actually use WD-40 for these days are locks. I always amaze people when they tell me about how their lock is starting to freeze-up and they have trouble operating it. I a little squirt of WD-40 and its like brand new. Yeah, it wasn't designed as a lubricant but it is a decent one. I also use it occasionally on squeeky door hinges when I'm too lazy to take it apart and grease it.

-Tim

 

[DerwenArtos12]

Originally posted by: PolymerTim

Originally posted by: DerwenArtos12

Originally posted by: PolymerTim
If it ain't broke, don't fix it! I agree completely. There are a lot of things that might go wrong no matter what technique you use. In the end, I say if it gives you the temps you want, then I don't care if you use your elbow to spread it!

All this talk about TIMs and mounting reminds me of another popular philosophy from my childhood:
"If it ain't s'posed to be movin' and it is, then use Duck tape. If its s'posed to be movin' and it ain't, then use WD-40."

I shudder at the thought of WD40 anywhere near my computer. Most people don't realize what WD-40 was really designed for. WD-40 was designed to displace water and it's the 40th formula they tested. Water Displacement formula 40. People just decided to use it as a lubricant which was a horrible idea as it collects dirt and dust really well.

Hehe, yeah. That quote kind of pre-dates computers. I don't see to many people using duct tape in computers either (although electrical tape and zip ties seem commonplace).

The only thing I actually use WD-40 for these days are locks. I always amaze people when they tell me about how their lock is starting to freeze-up and they have trouble operating it. I a little squirt of WD-40 and its like brand new. Yeah, it wasn't designed as a lubricant but it is a decent one. I also use it occasionally on squeeky door hinges when I'm too lazy to take it apart and grease it.

-Tim

For locks you'll be much better off using a small squirt of a penetrating oil followded by a silicon lubricant. Should break the lock up just as well if not better and the silicon won't collect dust the way WD-40 does. Graphite is actually the best lubricant for door hinges.

Ah the things you learn working at home depot.

 

[DerwenArtos12]

Oh, and i just re-cabled my HTPC/file server last night and decided to try this out. Before re-cabling my CPU was idling around 32C with 18C ambients. I removed all the old ceramique and replaced it with a dab of AS5 and pressed the HSF on and twisted slightly to distribute the TIM. Fired her up and she's idling around 30C with 18C ambients, as far as i can tell thats just the difference between the ceramique and the AS5. I gotta say if you take your time and are careful most anyone should be able to adequately spread the stuff.

 

[Martimus]

Not too shiny - I can't find it now, but I recently read an article where I guy actually varied the level of polish on a HS (for a laser in his lab, but the same kind as for procs) to see the effect on thermal transport using the same TIM as computers use. To their surprise, they found an optimal roughness (using about 800-1000 grit sand paper). Putting a mirror finish actually degraded performance a little. The hypothesis is that the micro grooves created by the 800 grit sandpaper were able to push through the thin TIM layer and make direct metal-metal contact.

If the two surfaces are completely flat (The heat spreader and heatsink), then it would be best not ot use thermal compound at all. Any TIM that they applied would actually make theheat conduct worse than the actual metal would, since there should be no trapped air between the two. The TIM is only used to displace poor thermally conducting matter (in this case air) with something that conducts better. TIM still conducts heat worse than the heatsink and heatspreader, so it would really just get in the way in that example, as there is no air for it to displace.

 

[CitizenSnips]

Originally posted by: Martimus
If the two surfaces are completely flat (The heat spreader and heatsink), then it would be best not ot use thermal compound at all. Any TIM that they applied would actually make theheat conduct worse than the actual metal would, since there should be no trapped air between the two. The TIM is only used to displace poor thermally conducting matter (in this case air) with something that conducts better. TIM still conducts heat worse than the heatsink and heatspreader, so it would really just get in the way in that example, as there is no air for it to displace.

I'm not experienced with this stuff at all, but everything I've read about it has said the whole point of thermal compound is to fill in the small gaps and imperfections that you can't see between the CPU and heat spreader.

 

[Martimus]

Originally posted by: CitizenSnips

Originally posted by: Martimus
If the two surfaces are completely flat (The heat spreader and heatsink), then it would be best not ot use thermal compound at all. Any TIM that they applied would actually make theheat conduct worse than the actual metal would, since there should be no trapped air between the two. The TIM is only used to displace poor thermally conducting matter (in this case air) with something that conducts better. TIM still conducts heat worse than the heatsink and heatspreader, so it would really just get in the way in that example, as there is no air for it to displace.

I'm not experienced with this stuff at all, but everything I've read about it has said the whole point of thermal compound is to fill in the small gaps and imperfections that you can't see between the CPU and heat spreader.

Yeah, but a highly polished and flat Heatsink and Heatspreader wouldn't have many airgaps; unseen or otherwise. Although, under heat they may warp differently if they are not the same material. I know that I have used heatsinks in the past without using any thermal paste, and the temperatures stayed pretty low. On the other hand, when I have added too much thermal paste, the temperatures raised very quickly. This leads me to believe that using no thermal paste is much better than using too much (Which makes sense logically as well), so you should always err on the side of too little versus too much.

 

[PolymerTim]

Originally posted by: Martimus

Originally posted by: CitizenSnips

Originally posted by: Martimus
If the two surfaces are completely flat (The heat spreader and heatsink), then it would be best not ot use thermal compound at all. Any TIM that they applied would actually make theheat conduct worse than the actual metal would, since there should be no trapped air between the two. The TIM is only used to displace poor thermally conducting matter (in this case air) with something that conducts better. TIM still conducts heat worse than the heatsink and heatspreader, so it would really just get in the way in that example, as there is no air for it to displace.

I'm not experienced with this stuff at all, but everything I've read about it has said the whole point of thermal compound is to fill in the small gaps and imperfections that you can't see between the CPU and heat spreader.

Yeah, but a highly polished and flat Heatsink and Heatspreader wouldn't have many airgaps; unseen or otherwise. Although, under heat they may warp differently if they are not the same material. I know that I have used heatsinks in the past without using any thermal paste, and the temperatures stayed pretty low. On the other hand, when I have added too much thermal paste, the temperatures raised very quickly. This leads me to believe that using no thermal paste is much better than using too much (Which makes sense logically as well), so you should always err on the side of too little versus too much.

That's true, but then I guess the next question is how flat is flat. I don't know the answer to that one, but I would guess that to get surfaces flat enough that they do better without TIM (properly applied) you would probably wouldn't be lapping by hand.

Originally posted by: DerwenArtos12
For locks you'll be much better off using a small squirt of a penetrating oil followded by a silicon lubricant. Should break the lock up just as well if not better and the silicon won't collect dust the way WD-40 does. Graphite is actually the best lubricant for door hinges.

Thanks for the tip. You never know what you're going to pick up around here. But I still wonder if silicon oil picks up less dust than WD-40. It probably sticks around longer though.

 

[Martimus]

That's true, but then I guess the next question is how flat is flat. I don't know the answer to that one, but I would guess that to get surfaces flat enough that they do better without TIM (properly applied) you would probably wouldn't be lapping by hand.

They would need to be identical to have the best conductivity. So if one was concave, the other would need to be convex at the same proportions that the first is concave. The whole point is to have the heat transfer from the heat spreader to the heat sink, so you would want as much surface area touching each other as possible. The best way of doing it would be to manufacture the Heatsink onto the chip, instead of using a user interface at all. That way you could reduce the possibility of error. I was just trying ot say that the group may have had better results if they had used no thermal paste on the highly polished part than with thermal paste. Of course, when you do polish things like that by hand, you are likely to make the part convex, which would reduce its effectiveness. The tiny imperfections caused by using a lower grit sandpaper would help offset that tendency, so that may be what caused their issues.

 

[DerwenArtos12]

Originally posted by: PolymerTim

Originally posted by: Martimus

Originally posted by: CitizenSnips

Originally posted by: Martimus
If the two surfaces are completely flat (The heat spreader and heatsink), then it would be best not ot use thermal compound at all. Any TIM that they applied would actually make theheat conduct worse than the actual metal would, since there should be no trapped air between the two. The TIM is only used to displace poor thermally conducting matter (in this case air) with something that conducts better. TIM still conducts heat worse than the heatsink and heatspreader, so it would really just get in the way in that example, as there is no air for it to displace.

I'm not experienced with this stuff at all, but everything I've read about it has said the whole point of thermal compound is to fill in the small gaps and imperfections that you can't see between the CPU and heat spreader.

Yeah, but a highly polished and flat Heatsink and Heatspreader wouldn't have many airgaps; unseen or otherwise. Although, under heat they may warp differently if they are not the same material. I know that I have used heatsinks in the past without using any thermal paste, and the temperatures stayed pretty low. On the other hand, when I have added too much thermal paste, the temperatures raised very quickly. This leads me to believe that using no thermal paste is much better than using too much (Which makes sense logically as well), so you should always err on the side of too little versus too much.

That's true, but then I guess the next question is how flat is flat. I don't know the answer to that one, but I would guess that to get surfaces flat enough that they do better without TIM (properly applied) you would probably wouldn't be lapping by hand.

Originally posted by: DerwenArtos12
For locks you'll be much better off using a small squirt of a penetrating oil followded by a silicon lubricant. Should break the lock up just as well if not better and the silicon won't collect dust the way WD-40 does. Graphite is actually the best lubricant for door hinges.

Thanks for the tip. You never know what you're going to pick up around here. But I still wonder if silicon oil picks up less dust than WD-40. It probably sticks around longer though.

Not silicon oil, silicon oil has silicon suspended in oil and will attract and absorb dust and dirt, home depot I know for sure carries a "dry" silicon lubricant that won't attract or absorb any dust/dirt.