Applying thermal compound to an exposed GPU core?

M

metallicaband

Limp Gawd
Joined
Nov 3, 2007
Messages
484
Hello all,

Was wondering which method do you guys use to apply thermal paste on an exposed GPU core, many people recommend the spread method because unlike CPU heatspreaders, the GPU core must be fully covered and the dot in the middle method doesn't do that..etc, then you hear all about the spread method causing air bubble and you'll get all mind fucked about what to do lol

Here's an interesting image from another forum (Source) about applying TIM to an exposed core:

thermalcompoundandgpus.png


But these videos makes it sound like it's a really bad idea to use the spread method with all those air bubbles:
http://www.youtube.com/watch?v=EyXLu1Ms-q4

http://www.youtube.com/watch?v=ffK7L0Qj13Q

From the looks of it seems like a nice line would be the way to go since it covers a good amount of space, but would like to know which method do you guys prefer.

I'll be using a non-conductive paste for this for sure since it's an exposed core, hell I'm paranoid enough to use non-conductive pastes even with CPU heatspreaders lol
 
I just blob a ton of it on the die in an X shape and let it ooze wherever it wants to go.

If you wanted to be anal about it, then the spread method would be the best.
 
I've found it hard to make a dot of grease so small that it doesn't cover the entire GPU with compound to spare. The GPU dies are so much smaller than a cpu die you'd have to intentionally put too little on using a needle or something. If in doubt, razorblade on one surface and a small dot in the middle of the other.
 
It sounds fairly similar to pasting a CPU when installing a heatsink. Personally I've gone with the spreading method with CPU's and used either an old 'credit' card or a gloved finger to spread it around in a nice even coat before setting the sink down.
 
When applying directly to the die, I always use the edge of a card to smooth it out. Done it a dozen times and have never had any sort of problem.

With a heat spreader, the dot method is best.
 
or0b3d.jpg

I did this to my GTX 280 after I baked it with a heatgun because it died. Got temps down from 98c at load to 81c.
 
NickJames said:
I did this to my GTX 280 after I baked it with a heatgun because it died. Got temps down from 98c at load to 81c.
Click to expand...

Yeah, but the OP is asking about GPUs without a heat spreader.

I say a dot in the middle. As was already said, it's going to be hard to not get enough thermal compound to cover the whole die. If you have a little bit of spillage, it's no big deal as long as you're using non-conductive thermal compound.
 
I apply the paste on the heatsink base rather than the CPU itself because the base of the cooler has the machining imperfections on them. I then use a credit card or a hard thin plastic object to spread the thermal paste evenly all around the bottom of the cooler.
 
i've always used the spread method. It is also the method i have achieved the best results with.

while those glass spreading videos are neat and all, all they do is apply the thermal paste to the cpu, and don't actually turn the computer on. Heat + constant pressure from the mounting system rids you of air bubbles.

Some of the best thermal compounds are so thick they must be spread.

but what ever works for you.
 
given the core is exposed you absolutely want to use the spread method.
 
The blob on the GPU always cover the entire GPU for me, with minor excess going around the corner of the GPU. It's also on the IC diamond website for CPU/GPU proper application. There are tons of youtube video showing how the spread method adds air bubbles. The blob in the center or the X method is so far the ones that I have seen that work well in eliminating air bubbles.
 
I use line method for CPUs and Spread Method for GPUs.
It really comes down to the person and if they know what they're doing to get the desired effect.
 
i spread it around, but i put the paste in the middle. yes this method may cause air bubbles, but i have always thought that the air bubbles are better than putting too much paste on. i have always been worried that putting too much on will infact insulate it, causing more heat to stay on the chip rather than transfer to the heatsink
 
Another vote for dot in the middle. For GPUs like AMD with exposed GPU, I'm a bit heavy handed with the size of the dot (relative to a CPU application). I also use non conductive paste like Shin Etsu X23-7783D.
 
4fingerninja said:
i spread it around, but i put the paste in the middle. yes this method may cause air bubbles, but i have always thought that the air bubbles are better than putting too much paste on. i have always been worried that putting too much on will infact insulate it, causing more heat to stay on the chip rather than transfer to the heatsink
Click to expand...

Actually that's not true. Air is an excellent insulator, even if you apply too much thermal paste it's a very efficient heat conductor.
 
all these talk of air bubbles is funny.

Its all based on a glass plate, and a cpu, and then pushing down with some thermal compound in the middle. These systems were never turned on, and constant pressure was never applied while they were turned of.

Heat changes the properties of all substances known to man. Also constant pressure from the heatsink will push any excess paste away. For all your air bubble fans, how come most factory heat sinks come with a Tim already applied to the surface which is a spread out method, rather than a single blob in the middle. Surely they must be worried about air bubbles :D

I am not trying to argue which method is better, cause really as long as your cpu makes contact with your heatsink / water block your fine.
 
ebduncan said:
all these talk of air bubbles is funny.

Its all based on a glass plate, and a cpu, and then pushing down with some thermal compound in the middle. These systems were never turned on, and constant pressure was never applied while they were turned of.

Heat changes the properties of all substances known to man. Also constant pressure from the heatsink will push any excess paste away. For all your air bubble fans, how come most factory heat sinks come with a Tim already applied to the surface which is a spread out method, rather than a single blob in the middle. Surely they must be worried about air bubbles :D

I am not trying to argue which method is better, cause really as long as your cpu makes contact with your heatsink / water block your fine.
Click to expand...

Why? Probably because a spread-out application is far cheaper and easier to mass-produce reliably. Emphasis on the cheaper part. Kinda like how lots of GPU coolers come with a minimum of hold down screws (sometimes as few as 4) when the reference PCBs have holes for a dozen.

Yea, so long as you get an acceptable result it is moot.
 
Why? Probably because a spread-out application is far cheaper and easier to mass-produce reliably. Emphasis on the cheaper part. Kinda like how lots of GPU coolers come with a minimum of hold down screws (sometimes as few as 4) when the reference PCBs have holes for a dozen.
Click to expand...

actually quite the opposite. You need less Tim with a blob in the middle, rather than a spread out square. Less Tim and well costs are cheaper.

In any case the flaws in the glass test method i listed above make it impossible to hold any merit on how you apply your thermal goo. This test doesn't met all the same operating conditions a REAL computer would undergo.

However your right, all that matters is that your application method is acceptable.
 
ebduncan said:
actually quite the opposite. You need less Tim with a blob in the middle, rather than a spread out square. Less Tim and well costs are cheaper.

In any case the flaws in the glass test method i listed above make it impossible to hold any merit on how you apply your thermal goo. This test doesn't met all the same operating conditions a REAL computer would undergo.

However your right, all that matters is that your application method is acceptable.
Click to expand...

Catch being, you presume that the manufacturers start off with a dot and spread it, which wastes TIM.
 
This might be in the "duh that's obvious category", but I'll mention it anyways: If you're curious to know what a specific thermal compound application technique looks like on a particular die, test it out with a flat piece of glass instead of the heat sink first.
 
This might be in the "duh that's obvious category", but I'll mention it anyways: If you're curious to know what a specific thermal compound application technique looks like on a particular die, test it out with a flat piece of glass instead of the heat sink first.
Click to expand...

yes you will see how your thermal compound gets spread on the die. You will not exactly know if it will stay that way when you turn the computer on, and after its been there for a bit.
 
ebduncan said:
yes you will see how your thermal compound gets spread on the die. You will not exactly know if it will stay that way when you turn the computer on, and after its been there for a bit.
Click to expand...

As long as there are no trapped air bubbles to expand when things heat up, it's pretty much going to stay where it is, between the heat sink and the die. A glass test will show bubbles pretty clearly :)
 
Kueller said:
As long as there are no trapped air bubbles to expand when things heat up, it's pretty much going to stay where it is, between the heat sink and the die. A glass test will show bubbles pretty clearly :)
Click to expand...

On a GPU die, that is completely exposed, i agree that in theory, the middle dot, if correctly applied (risk), makes more sense. But we 're talking about minimal differences, maybe 1C.

On a CPU though, the spread method gives the same temperatures if applied well and has less risks that the middle method.

The infamous "air bubbles", don't explain why several people that have tried repeatedly both methods, don't find temperature difference. On the other hand the dot method becomes riskier, because there is more that can go wrong. For example, with my very wide cooler, the dot method often results in eccentric spread pattern while you try to clamp down the cooler on the AMD bracket.

On the other hand, i do have a theory why the air bubbles don't penalize so much as the bubble theory would presume, even on those tests that do find a difference:

1) In a real system, the compound is heated up. Heat must be doing something to the compound's bond with the metal. Some compounds have break-in time. This implies, that that the nature of the bond with the metal changes. Even those that claim no such time, with time become stickier, some to the point that you risk ripping off the socket if you don't try to twist it first. This doesn't happen when its fresh, heat is what changes that. And when something becomes more attached, logic would suggest that the infamous "trapped air" (which is so much in the glass test, that would hardly allow it to stick) is moved out as the compound sticks more to the metal (vacuum effect).

2) High School Physics. Heat will travel better inside a metal and inside the same metal. This can compensate for the "air bubbles", because the spread method allows more metal to come to better contact with the heatsink. The claim that the "cores are in the middle, so the edges have no effect", is ignoring basic physics. Heat will NOT simply travel vertically towards the center of the compound and in the heatsink. It will also travel towards the edges of the heatspreader and to the heatsink. Metal vs compound, metal always wins and even metal vs different metal also wins, because the thermal conductivity is reduced between the passage from one metal to the other. This is the principle of use of direct heatpipes too. Instead of putting a metal plate and the heatpipes afterwards, based on the plate, they put the heatpipes directly on the base of the heatsink, so to eliminate the "extra passage" between a metal layer. So, the spreading method, even if we accept the "air bubbles" defect, has more surface for heat exchange. That's why it doesn't fare so badly after all despite the "huge amount of highly insulating air" seen in the youtube videos.

With the spreading method, the difficulty is to make a really flat, even, thin layer. But it takes patience. Once done, there is nothing that can go wrong. No spilling over to the motherboard, no incomplete coverage of one core, no bad mount.

The dot method, is more "guesstimate". You must put the right amount of TIM (right amount changes from grease to grease, as they don't spread the same), you must seat the heatsink perfectly (or it will spread more towards one direction) and pray you got it right. As as i said, with my cooler, it's very easy to go wrong with the dot method, because the cooler is so wide and big that you 're not seeing what you 're doing when you try to clamp it in the case and there is high probability that you will not seat the heatsink perfectly on the center of the dot and with equal force, causing the TIM to move to the side and spread unevenly.

But, these are matters of taste. I wanted to believe in the dot, exactly because i was glad to be able to be lazy and just put "wham-bam" and over with it. But no matter how times i tried, with 2 coolers and 4 different TIMs, i wasn't able to get a single degree C better temperature with the dot, no matter how many times i watched the video with the lady and her plexyglass on youtube... So i have to believe my eyes, on real mount...

Oh i also tried a hybrid method. Spread plus a tiny dot in the middle for "reinforcing" (moving away the air in the middle). That works fine too, but with no temperature benefit whatsoever. But it's fun.
 
Faethon said:
On a GPU die, that is completely exposed, i agree that in theory, the middle dot, if correctly applied (risk), makes more sense. But we 're talking about minimal differences, maybe 1C. ...
Click to expand...
Glad you got that off your chest...but I never advocated any particular way of applying the compound, just that bubbles are bad. Though not necessarily because they directly interfere with surface contact. More because with repeated heating/cooling expansion cycles they can create contact problems.
 
Kueller said:
Glad you got that off your chest...but I never advocated any particular way of applying the compound, just that bubbles are bad. Though not necessarily because they directly interfere with surface contact. More because with repeated heating/cooling expansion cycles they can create contact problems.
Click to expand...

I only took advantage of the thread to say what i think about that test, you happened to be the last that posted...So don't worry, i didn't mean that you advocate the method.
 
You must log in or register to reply here.
Back
Top