I swear by my credit card reason being even if you just use a blod the size of a grain of rice,and finished applying it perfectly you can push the credit card flat against the processor and pull towards yourself and remove some more,and still retain the perfect finish,just my 2c :)
Originally Posted by Pygo You mean I've been doing it wrong all along? How do I remove paste from the cpu socket and the pins, as per image linked below??
Is this really all that bad? My friends are saying I'm stupid now... :( http://i87.photobucket.com/albums/k151/cburn25/42prosessor1.jpg
Put the motherboard and CPU into a dishwasher with white spirit or meths. :)
1. Use a carburator degreaser spray from your local auto shop.
2. Select preferred TIM based on your trusted reviewers on the internets.
3. ...
4. Profit
Originally Posted by Pookeyhead I've ALWAYS done it this way. This put a small blob in the middle and then apply the HSF is a stupid way of doing it. Whenever I've done that, then removed the heatsink, there are vast areas of the chip without thermal paste on it.
Grain of rice sized amount... spread to an even, thin film with finger. Works everytime.
I disagree... I'm sure spreading it thin works fine but I have an artic cooling freezer 7 pro cooling my oc'd c2d to 30*C right now... applied with the drop of TIM + apply heatsink method.
Quote:
Rubbish.
Prove it.
Whenever I've used the "accepted" method as you state, after removing the HSF, even after months, there are huge areas of the chip uncovered.
How the hell can spreading it with your finger "trap" air under the compound? LOL
Say what you like, but while that article may not win a pulitzer any time soon, it's spot on technically. If you're putting a grain of rice size blob down, then just clamping the HSF on, you're an idiot :)
Thing is... you said above you've always done it the way by spreading it with your finger, so how do you know the clamping method doesn't work? It obviously works for some (myself included). I don't know the merits of the finger spreading method :p but I think it is trying to be too anal over a little task in my eyes. As long as the TIM is there, and there isn't a stupid amount, then it will work. The best application in the world will be within 4-5*C of the shoddiest application, so there's no need to get stressed about it ;). (think in perspective, that's the same as a hot day vs a cold day lol)
Great article, the way I have done it in the past was to put 5 really small dots, one in the center and four just out from the center towards the corners, on the cpu and rub the heatsink against it in a circular motion to evenly cover both surfaces.
It all depends on the type of TIM used. Some TIM's just won't spread and you have to use the clamping method. But if your willing to go the extra mile on those hard to spread TIM's try this. Put the tube of TIM in a sealed ziplock bag and let it soak in HOT water for 5 minutes. The TIM should spread if it's going to :D
Talk about the HSF surface being concave/convex started and a thought occurred to me:
Using the "blob and clamp" method would likely fill those gaps as the heatsink is clamped down, and the TIM forced towards the edges. Pre-spreading would still force the TIM into the gaps, but to a much lesser extent, and in the case of a concave HSF base (ie. the curve is inwards) would likely result in air being trapped (unless you've spread the TIM a little thickly, or spread it on the HSF base as well).
As far as performance testing, why would the tests necessarily have to be CPU based? surely there would be some way of having a heating element on one side of a piece of metal, another piece that can be used to read temperatures and TIM in between. in this case, higher temperatures would be suggesting the best thermal transfer. This way, you simply replace the metal being used (using pieces cut from the one sheet for all the tests should ensure that the pieces have the same thermal transfer properties)
Since going to the "two grains of rice" method on IHS covered heatsinks, not only have my temps improved, but the resulting layer of TIM has been much thinner and more consistent than I could ever achieve with the razor/ finger/ credit card method.
Folks, the heat generating die is in the center of the IHS, not at the edges. You are cooling the die, not the heatspreader. Look at the stock Intel cooler. What do you see? a center mounted slug with NO contact at the outer IHS edges. Look at the design of current-gen waterblocks. What do you see? Bowed bases, stepped bases and round bases to increase pressure and reduce the thickness of the TIM layer, improving thermal efficiency where it counts- the CENTER of the IHS (over the die). Do you really think companies marketing to guys running $400+ cooling systems are doing it wrong?
Food for thought: the vertical edges of the heatspreder do NOT compress. What you are doing with any decent mounting apparatus at the center of the IHS (over the die) is creating pressure to:
1. compress the IHS (by extension the internal and external TIM layers),thereby reducing distance between die and the heatsink
2. cooling that which generates heat- the die.
Metal is a much better heat conductor than any TIM. You want as little TIM as you can manage, not as much as is possible.
This article is state-of-the industry-2002 stuff; it has very little bearing on today's cutting edge cooling methods.
Originally Posted by Pookeyhead This put a small blob in the middle and then apply the HSF is a stupid way of doing it...
Those idiots at Arctic Silver are doing it all wrong then? I'd sooner listen to them than some stupid kid writing a magazine article. ;)
You beat me to it - I was going to say exactly the same thing. Manufacturer of the most recognized TIM on the planet versus a blog? Hmm...nope, no contest there, I'm sticking to their directions (which I've been following for nearly 15 years with zero problems).
Originally Posted by Veles Or you're just following the advice of Arctic Silver, since they you know, make the stuff, I would imagine they know their stuff.
I'm just following the laws of physics. The greater the area of contact between two surfaces, the greater the thermal transfer between them will be. Therefore ensuring the entire surface has a coating of the stuff at an appropriate thickness BEFORE you apply the heatsink has got to be better than applying a blob and HOPING it covers the entire surface AFTER you apply the heatsink. Think about it. You can either apply a grain of rice sized blob and hope the heatsink pressure spreads it evenly, or you can apply the same size blob and make sure it covers the entire surface manually before you apply the heatsink. The absolute worse that can happen is there will be no difference whatsoever... the best that can happen is you will get better heat transfer.
I challenge anyone to argue with that logic.
I couldn't give a monkey's what arctic silver say.
Quote:
Originally Posted by Veles My CPU hasn't burnt out yet, if you think about it, the majority of the heat would be directly over the chip, the periphery of the heat spreader which wouldn't get covered would be at a much lower temperature than the centre unless the heat spreader is crazy efficient.
Like you say.. the majority will be directly over the chip, yes... the majority.. not all. Ensuring the entire heat spreader is coated will ensure ALL of the heat is transferred, especially as many third party HFS have a surface as large, or larger than the chip.
Again... challenge the logic if you must.
That might sound like common sense...this article certainly shows that the author thought along the same lines. "More paste covering the IHS must be better right?" Well, while tests like the one I linked to before simply test for total TIM coverage, the most absolute thing to do is to do temperature tests. And that's what Madshimps did back in 2007. While CPUs and Heatsinks might have changed since then, Laws of Physics don't. And the graphs show a good 5°C difference in the least. I used to fall under the spread-TIM group, but since I read that article, I just use the rice-grain/clamp method.
When I first started using 3rd party heat sinks, about 6 months ago, I was using way too much. Once I realized that, I looked at the method Arctic Silver recommend since I'm using the Arctic Silver 5. That is, a horizonal line for Quads and vertical line for Duos.
Now I'm wondering if that's still too much, but things seem to be working OK.
Ok first off to all the assholes complaining about this article: I myself have built several systems, done some basic to mid-level modding, and consider myself a knowledgeable person. However i've never really been sure on the best way to apply TIM. I know that less is more and that you practically can't have too little, the danger is too much.
But like many people who built their systems in late '06 I snagged one of those nifty LGA775 Pentium 4 HT's with a Prescott core. Now these things will fry more than eggs. They'll pan-roast pork tenderloin in a pinch.
I've had to settle with seeing 50C idle after trying a copper cooler from scythe, arctic silver, ceramic ****, damn near anything and i've been through several applications of TIM on the goddamn thing. I read the article and it made sense, as I always wondered why one of the brands of TIM i bought came with a credit card labeled "Compound spreader"
I tried the method out, and I used to see as high as 65C load (which is still within intel's specs for this chip), and after 30 minutes of pvp in WoW and some headshots in CS:S as well as an arena match or two in Guild Wars, I'm seeing 53C load. A difference of 12C just from switching up my application method.
So I am a convert to the saran-wrapped finger as opposed to the rice-grain and clamp method. It might make a bigger difference to me because the bottom of my hs isn't mirror smooth. I spread it evenly across the cpu die and due to the shite state of affairs that is my hs base, i used what little of the TIM was still clinging to the saran wrap on the copper base, an extremely thin layer, but that way i know the "little holes" were filled in.
a drop in 10C idle 12C load makes a convert out of me. Finally my prescott isn't so ****ing hot.
And I thought we'd discussed all of this to Effin death in the late nineties... :D
There used to be reviews (with Temp.measurements) of Arctic Silver (1...not 5) vs. pad vs. some silicone stuff. Should be around the Internet somewhere...look for a 2001 Timeframe :
This "imported" article isn't really bad but we're not the target audience. It just doesn't live up to the standards the "enthusiast" level reader expects.
Oh it does a fine, reassuring job for people that have never placed a heatsink themselves. (oohoohooh I'm a big modder now, I've replaced my stock cooler with an aftermarket one!) And to be honest, this is how many here (I guess) began down the "enthusiast" path. And before you know it, you're looking at rotary tools that never interested you before.
Nevertheless this article isn't up to your usual level, and if this is a peek of the future coming out of your accuisition by dennis Brrr!
Common sense tells me that the "rice grain in the middle" method is much more effective than spreading the TIM by hand.
Yes, spreading the TIM by hand WILL generally result in a larger coverage area than the rice grain method.
However, it's also guaranteed to create many air pockets due to the uneven surface of the spread TIM.
While with the rice grain, the TIM is forced out to by pressure between 2 metal plates, clearing the air pockets at the same time. It might have less contact area than spreading TIM by hand, but it should cover the vital area: the core(s).
Personally I've always used the grain rice method, and my CPUs got pretty good temp so far.
Thing is... you said above you've always done it the way by spreading it with your finger, so how do you know the clamping method doesn't work?
Because I just rebuilt my old rig for my wife, and to be frank, I couldn't be bothered taking all the same precautions I normally do, as it wasn't intended to be overclocked, or used heavily. I simply applied a small line of arctic silver about the width of a grain of rice and perhaps 10mm long, then applied the HSF. Later, because of something else unrelated, I had to remove it again, only to find that it was covering only half the chip. Besides, I say "I've always done it that way".. when really, what I should have typed was "I've been doing it that way for ages". Obviously, in the beginning, I did use the recommended method of putting a blob there and letting the pressure sort it out.
The problem, as I see it, arises from the pressure applied. The socket 939 clips on my wife's HSF were clearly just not cutting it.
I still maintain that it makes little practical difference. Perhaps pre-spreading may introduce some bubbles, maybe not... but is that any worse than having 30% of the chip with nothing on it? Remember, where there is no compound, there's probably also no contact between chip and HSF.
In all honesty, I can't say that I've noticed any difference in temps either way, but at least I have the piece of mind from knowing it's covered. The only way you can be sure the other way is to remove it and check, and then if you re-apply it you have air bubbles AND uneven coverage.... you end up getting in that situation where you become paranoid if the fridge light goes out when you close the door :).. if you know what I mean.
After cleaning the heatsink contact side, I often apply a small amount of TIM and using the finger and plastic bag, work it in so that hopefully any microscopic defects or holes are filled. I then wipe the majority of that TIM off using a clean cloth so it's clean - apart from the filling of the crevices/small defects, etc.
I then do the small rice sized blob on the chip and apply the heat sink and clamp done. Always been great for me. Done this for years...
I use basically the same method as in the article for years whit these different's.
1. I use half a grain drop but on both the CPU and the cooler
2. after spreading it around i try to remove as mouths as possible whit my finger (whipping the plastic bag on my finger on a cloth)
(golden rule whit TIM is, that less is more)
3. after installation i move the cooler around as far as it will go whit out forcing it for a bout 5~10min (twisting and sliding)
This would give me 2~3c lower temp then using the method as in the article
(I did do some testing whit H2O on a old OC rig before i installed the H2O set on my main system)
I will keep this as simple as possible. Thermal compound or thermal interface material only starts to do it's thing when in the presence of heat.
1. Put a drop ( or a line or a squiggle - it makes no difference ) in the middle of the surface you want to cool, be it cpu, chipset, mosfet e.t.c and then clamp the heatsink in place on top of the tim.
2. fire up the equipment and marvel in awe at how the paste forces itself into all the 'nooks and crannies' aided by the heat and pressure between the two surfaces.
3. Do not waste your time rubbing it around with a credit card, spatula, small woodland creature, plastic bag on the end of your finger, toes or other protuberance, because it isn't helping :)
4. And this is important. DO NOT USE ELECTRICALLY CONDUCTIVE MATERIAL AS TIM! ESPECIALLY NOT WHEN 'THERMAL PUMP OUT' MAY SPREAD IT ALL OVER YOUR SHINY NEW MOTHERBOARD WHEN THINGS HEAT UP!!
I tend to apply a bit more than you did. Then I place the cooler on top to spread out the TIM slightly. I then wipe the transfered TIM off the cooler base and spread the remaining TIM on the chip evenly. Works great for me.
I use AS5 and always tend to put just a little bit extra on just in case, especially if you are OCing.
Comments 76 to 100 of 109
ReplyPut the motherboard and CPU into a dishwasher with white spirit or meths. :)
I found this worst with some AS5 just make sure you throw away the first rice size grain out the syringe and its usually ok then.
Judging from other people's experiences, that seems to work fairly well.
2. Select preferred TIM based on your trusted reviewers on the internets.
3. ...
4. Profit
I disagree... I'm sure spreading it thin works fine but I have an artic cooling freezer 7 pro cooling my oc'd c2d to 30*C right now... applied with the drop of TIM + apply heatsink method.
Thing is... you said above you've always done it the way by spreading it with your finger, so how do you know the clamping method doesn't work? It obviously works for some (myself included). I don't know the merits of the finger spreading method :p but I think it is trying to be too anal over a little task in my eyes. As long as the TIM is there, and there isn't a stupid amount, then it will work. The best application in the world will be within 4-5*C of the shoddiest application, so there's no need to get stressed about it ;). (think in perspective, that's the same as a hot day vs a cold day lol)
http://www.anthsbespokecards.co.uk/daz/DSCF0060.jpg
Using the "blob and clamp" method would likely fill those gaps as the heatsink is clamped down, and the TIM forced towards the edges. Pre-spreading would still force the TIM into the gaps, but to a much lesser extent, and in the case of a concave HSF base (ie. the curve is inwards) would likely result in air being trapped (unless you've spread the TIM a little thickly, or spread it on the HSF base as well).
As far as performance testing, why would the tests necessarily have to be CPU based? surely there would be some way of having a heating element on one side of a piece of metal, another piece that can be used to read temperatures and TIM in between. in this case, higher temperatures would be suggesting the best thermal transfer. This way, you simply replace the metal being used (using pieces cut from the one sheet for all the tests should ensure that the pieces have the same thermal transfer properties)
Since going to the "two grains of rice" method on IHS covered heatsinks, not only have my temps improved, but the resulting layer of TIM has been much thinner and more consistent than I could ever achieve with the razor/ finger/ credit card method.
Folks, the heat generating die is in the center of the IHS, not at the edges. You are cooling the die, not the heatspreader. Look at the stock Intel cooler. What do you see? a center mounted slug with NO contact at the outer IHS edges. Look at the design of current-gen waterblocks. What do you see? Bowed bases, stepped bases and round bases to increase pressure and reduce the thickness of the TIM layer, improving thermal efficiency where it counts- the CENTER of the IHS (over the die). Do you really think companies marketing to guys running $400+ cooling systems are doing it wrong?
Food for thought: the vertical edges of the heatspreder do NOT compress. What you are doing with any decent mounting apparatus at the center of the IHS (over the die) is creating pressure to:
1. compress the IHS (by extension the internal and external TIM layers),thereby reducing distance between die and the heatsink
2. cooling that which generates heat- the die.
Metal is a much better heat conductor than any TIM. You want as little TIM as you can manage, not as much as is possible.
This article is state-of-the industry-2002 stuff; it has very little bearing on today's cutting edge cooling methods.
You beat me to it - I was going to say exactly the same thing. Manufacturer of the most recognized TIM on the planet versus a blog? Hmm...nope, no contest there, I'm sticking to their directions (which I've been following for nearly 15 years with zero problems).
That might sound like common sense...this article certainly shows that the author thought along the same lines. "More paste covering the IHS must be better right?" Well, while tests like the one I linked to before simply test for total TIM coverage, the most absolute thing to do is to do temperature tests. And that's what Madshimps did back in 2007. While CPUs and Heatsinks might have changed since then, Laws of Physics don't. And the graphs show a good 5°C difference in the least. I used to fall under the spread-TIM group, but since I read that article, I just use the rice-grain/clamp method.
http://www.youtube.com/watch?v=EyXLu1Ms-q4
When I first started using 3rd party heat sinks, about 6 months ago, I was using way too much. Once I realized that, I looked at the method Arctic Silver recommend since I'm using the Arctic Silver 5. That is, a horizonal line for Quads and vertical line for Duos.
Now I'm wondering if that's still too much, but things seem to be working OK.
But like many people who built their systems in late '06 I snagged one of those nifty LGA775 Pentium 4 HT's with a Prescott core. Now these things will fry more than eggs. They'll pan-roast pork tenderloin in a pinch.
I've had to settle with seeing 50C idle after trying a copper cooler from scythe, arctic silver, ceramic ****, damn near anything and i've been through several applications of TIM on the goddamn thing. I read the article and it made sense, as I always wondered why one of the brands of TIM i bought came with a credit card labeled "Compound spreader"
I tried the method out, and I used to see as high as 65C load (which is still within intel's specs for this chip), and after 30 minutes of pvp in WoW and some headshots in CS:S as well as an arena match or two in Guild Wars, I'm seeing 53C load. A difference of 12C just from switching up my application method.
So I am a convert to the saran-wrapped finger as opposed to the rice-grain and clamp method. It might make a bigger difference to me because the bottom of my hs isn't mirror smooth. I spread it evenly across the cpu die and due to the shite state of affairs that is my hs base, i used what little of the TIM was still clinging to the saran wrap on the copper base, an extremely thin layer, but that way i know the "little holes" were filled in.
a drop in 10C idle 12C load makes a convert out of me. Finally my prescott isn't so ****ing hot.
And I thought we'd discussed all of this to Effin death in the late nineties... :D
There used to be reviews (with Temp.measurements) of Arctic Silver (1...not 5) vs. pad vs. some silicone stuff. Should be around the Internet somewhere...look for a 2001 Timeframe :
Xir
Oh it does a fine, reassuring job for people that have never placed a heatsink themselves. (oohoohooh I'm a big modder now, I've replaced my stock cooler with an aftermarket one!) And to be honest, this is how many here (I guess) began down the "enthusiast" path. And before you know it, you're looking at rotary tools that never interested you before.
Nevertheless this article isn't up to your usual level, and if this is a peek of the future coming out of your accuisition by dennis Brrr!
Xir
Yes, spreading the TIM by hand WILL generally result in a larger coverage area than the rice grain method.
However, it's also guaranteed to create many air pockets due to the uneven surface of the spread TIM.
While with the rice grain, the TIM is forced out to by pressure between 2 metal plates, clearing the air pockets at the same time. It might have less contact area than spreading TIM by hand, but it should cover the vital area: the core(s).
Personally I've always used the grain rice method, and my CPUs got pretty good temp so far.
Because I just rebuilt my old rig for my wife, and to be frank, I couldn't be bothered taking all the same precautions I normally do, as it wasn't intended to be overclocked, or used heavily. I simply applied a small line of arctic silver about the width of a grain of rice and perhaps 10mm long, then applied the HSF. Later, because of something else unrelated, I had to remove it again, only to find that it was covering only half the chip. Besides, I say "I've always done it that way".. when really, what I should have typed was "I've been doing it that way for ages". Obviously, in the beginning, I did use the recommended method of putting a blob there and letting the pressure sort it out.
The problem, as I see it, arises from the pressure applied. The socket 939 clips on my wife's HSF were clearly just not cutting it.
I still maintain that it makes little practical difference. Perhaps pre-spreading may introduce some bubbles, maybe not... but is that any worse than having 30% of the chip with nothing on it? Remember, where there is no compound, there's probably also no contact between chip and HSF.
In all honesty, I can't say that I've noticed any difference in temps either way, but at least I have the piece of mind from knowing it's covered. The only way you can be sure the other way is to remove it and check, and then if you re-apply it you have air bubbles AND uneven coverage.... you end up getting in that situation where you become paranoid if the fridge light goes out when you close the door :).. if you know what I mean.
I then do the small rice sized blob on the chip and apply the heat sink and clamp done. Always been great for me. Done this for years...
1. I use half a grain drop but on both the CPU and the cooler
2. after spreading it around i try to remove as mouths as possible whit my finger (whipping the plastic bag on my finger on a cloth)
(golden rule whit TIM is, that less is more)
3. after installation i move the cooler around as far as it will go whit out forcing it for a bout 5~10min (twisting and sliding)
This would give me 2~3c lower temp then using the method as in the article
(I did do some testing whit H2O on a old OC rig before i installed the H2O set on my main system)
1. Put a drop ( or a line or a squiggle - it makes no difference ) in the middle of the surface you want to cool, be it cpu, chipset, mosfet e.t.c and then clamp the heatsink in place on top of the tim.
2. fire up the equipment and marvel in awe at how the paste forces itself into all the 'nooks and crannies' aided by the heat and pressure between the two surfaces.
3. Do not waste your time rubbing it around with a credit card, spatula, small woodland creature, plastic bag on the end of your finger, toes or other protuberance, because it isn't helping :)
4. And this is important. DO NOT USE ELECTRICALLY CONDUCTIVE MATERIAL AS TIM! ESPECIALLY NOT WHEN 'THERMAL PUMP OUT' MAY SPREAD IT ALL OVER YOUR SHINY NEW MOTHERBOARD WHEN THINGS HEAT UP!!
Apologies for the caps but, wow, just wow!
I use AS5 and always tend to put just a little bit extra on just in case, especially if you are OCing.
-
« Previous
-
1
-
2
-
3
-
4
-
5
-
Next »
Discuss in the forums