Beware thermal paste

Mark Rose · 2015-11-01, 02:36

So the SSD in my desktop at home recently died, and I wanted to test out thermal loading my new HTPC case anyway, so I decided to move it to the new case. However, thing went bad when I tried to remove the CPU cooler. I had used a thin dab of Arctic Silver (the thermal paste, not the glue) and the Hyper 212 was basically welded to the CPU. While trying to work it loose, I pulled the AMD Athlon X4 640 out of the socket. So far not so bad.

I tried everything to remove the CPU. I started with isopropyl alcohol. No luck. I tried dental floss, but there is basically no space between the CPU and cooler, since I used very little thermal paste. I tried heating it to 70°C with a hair dryer (the heat pipes work quickly in reverse, and I watched the temperature with my infrared thermometer). I tried freezing it to -20°. Nothing worked.

Due to the orientation of the cooler, I wasn't able to action the CPU socket lever. Since the Socket AM3 motherboard is five years old, I wasn't going to buy a new CPU for it. I figured since I pulled the CPU without levering the socket, maybe it will go back in... and it did, but now the fit is loose. So I broke the motherboard.

I've got some other things to take care of before buying a new Skylake system, so to tide myself over, and I'm picking up a used ASRock H81 PRO BTC Motherboard for $40 and a new i3-3170 to go in it. I guess it was time to buy a new system, since this "cheap fix" will double the speed of my system.

chris2be8 · 2015-11-01, 16:42

Been there, done that. I managed to get the CPU off the cooler by soaking the paste with the solvent that came with my supply of Arctic Silver (basically white spirit with an additive that works as an antioxidant) round the edge of the CPU until the thermal paste softened enough to get the CPU off. But it took over 2 hours (I was working on my patio since the bottle of solvert said to use it in a well ventilated place).

Chris

EdH · 2015-11-01, 20:08

I had a friend many years ago that used to test the physical properties of IC packages. One of the tests he performed was to see how much stress it took to "twist" the top off an IC, before looking for degradation of the internal connections. That got me thinking about breaking the bond between heat sinks and ICs by a slight twisting motion, which is what I commonly use when removing the CPU coolers in older machines. I haven't run across any which needed a lot of force, however I have had a couple which, due to the type of mounting, didn't offer any twisting room. I wonder if a small amount of twisting force would work for a CPU that is bonded as well as Mark Rose's.

I also wonder if Acetone would be detrimental to a CPU, since it does a real good job deforming some plastics. I suppose my preferred solvent would be 99% alcohol, which I have on hand and use for electronics.

As to the original motherboard - Are you confident it is destroyed? Many types of IC sockets that close the contacts on the pins, have quite a bit of space behind the closing portion if no pin is between them. This could account for sloppiness if the contacts were in the closed position and you placed the CPU back into the socket.

Mark Rose · 2015-11-01, 21:26

I'm also concerned about twisting the circuit board off the bottom of the CPU IC. I almost used and adjustable wrench but my biggest one was slightly too small.

From what I've read, acetone is a no-no.

LaurV · 2015-11-02, 04:33

I somehow missed this thread yesterday.
Some thermal pastes are "curing" type, this means that you have to heat them to ~120°C (depending on paste) and it will "cure", i.e. it can really melt like tin, and "weld" the CPU to the cooler. They will also "cure" at lower temperature, in longer time, like for example 90 hours at 85°C, but when cured like that, they will only reach about 96% of the thermal transfer capacity. Those types are much better for thermal transfer, after curing they have the "metal" transmissibility for heat, but they are not so commonly used, exactly from the reason what happened to you - they are designed for systems that you don't intend to take apart. We use thingies like that in our industrial embedded touch-terminals here in production. Make sure you didn't use such a paste, before attempting to take it of. If you used "arctic silver 5", for example, and your CPU "accidentally" got over 105°C for few minutes, or you had more than 300-400 hot/cold cycles, then you may be just an unlucky son of a gun. If you did, you will end up damaging everything long before being able to take the cooler off the CPU. Some of them still have specific solvents, which will soften it, even after curing, but some not. Google for "curing thermal paste" for clarifications, but be careful that a lot of people talk about "curing" without having idea what it is. The best way is to take the datasheet of the paste - provided by the manufacturer, I can help you if you tell me exactly the type - and look how it behaves.

The alternative is to put the whole assembly (cooler and CPU) in a vice, take a piece of wood and a hammer, and delid the toy. You only have to look carefully of the hitting direction, put it on the long-side of the die, and add a pillow in the back, for this setup (some other videos add a pillow, but they are not so funny as the CPU doesn't fly away

).

DON'T TWIST the CPU! You will damage the die. And you may damage the die if you hit it laterally too. You are only allowed to hit it longitudinally, due to how the golden balls/bumps are fixed on the PCB.

Mark Rose · 2015-11-02, 05:15

The chip was run hot for a while a year ago, when the Hyper 212 Evo fan failed. I don't know how hot it got exactly, but it was thermal throttling. It's max operating temperature was 71°.

Now that I think about it, I probably just used the stock paste on that cooler. I wasn't overclocking or anything, I simply wanted something quieter than the stock cooler I had been using (whose fan was getting noisy).

I don't think hitting it would work as the circuit board protrudes around the die on all four sides. It looks like the attached picture.

I only tried twisting it with my fingers, but with enough force to leave dents in my thumbs.

Anyway, I'm about to power on the replacement system.

LaurV · 2015-11-02, 06:36

That only looks like a silicone rubber or epoxy resin ring to me. The process to "dig sculptures" into the FR4 PCBs is not easy, I don't believe that the metal cap is "buried". It would be the first time I see it. You can try some of what this guy does here, (for about the same CPU).

OTOH, talking with someone here, they say the Arcticlean solution is xylene based, and you can use any carburetor cleaner solution, or the additive that is added to the gasoline on your car to make it more powerful (here you can buy small bottles of it in any superstore) to dissolve that paste. Put it in and let it for few hours.

If it disappears completely, CPU, cooler and all, don't blame me.

VBCurtis · 2015-11-02, 06:40

Quote:

Originally Posted by LaurV

If it disappears completely, CPU, cooler and all, don't blame me.

kladner · 2015-11-02, 06:48

Would it be possible to insert a very thin blade to apply pressure right at the junction of the heatsink base and the heat spreader? Here are some possibly useful Xacto models. Maybe get the junction wet with alcohol or Arctic Silver cleaner, and then either scrape along the edge, or judiciously tap the knife handle. Getting enough purchase to twist or pry with the blade might get non-destructive results.

Mark Rose · 2015-11-02, 06:57

Quote:

Originally Posted by LaurV

That only looks like a silicone rubber or epoxy resin ring to me. The process to "dig sculptures" into the FR4 PCBs is not easy, I don't believe that the metal cap is "buried". It would be the first time I see it. You can try some of what this guy does here, (for about the same CPU).

OTOH, talking with someone here, they say the Arcticlean solution is xylene based, and you can use any carburetor cleaner solution, or the additive that is added to the gasoline on your car to make it more powerful (here you can buy small bottles of it in any superstore) to dissolve that paste. Put it in and let it for few hours.

If it disappears completely, CPU, cooler and all, don't blame me.

I guess I could do what that guy does. Even if everything is broken, it would be neat to see the inside. Still wouldn't help remove the lid from the heatsink though.

The motherboard/CPU would still be useful to me if I can get them working. I may just try reinstalling everything and leave the heatsink attached and see if it's stable.

The new machine is running. The heatsink included with the i3-4170 is junk. Just one core has it hitting 90°. That's not so bad though, since the hand-me-down RAM is running at 1333.

Mark Rose · 2015-11-02, 07:00

Quote:

Originally Posted by kladner

Would it be possible to insert a very thin blade to apply pressure right at the junction of the heatsink base and the heat spreader? Here are some possibly useful Xacto models. Maybe get the junction wet with alcohol or Arctic Silver cleaner, and then either scrape along the edge, or judiciously tap the knife handle. Getting enough purchase to twist or pry with the blade might get non-destructive results.

I tried forcing an exacto blade in between and broke the holder. The blade is wider than any gap by far.

2015-11-01, 02:36	#1
Mark Rose "/X\(‘-‘)/X\" Jan 2013 2·1,553 Posts	Beware thermal paste So the SSD in my desktop at home recently died, and I wanted to test out thermal loading my new HTPC case anyway, so I decided to move it to the new case. However, thing went bad when I tried to remove the CPU cooler. I had used a thin dab of Arctic Silver (the thermal paste, not the glue) and the Hyper 212 was basically welded to the CPU. While trying to work it loose, I pulled the AMD Athlon X4 640 out of the socket. So far not so bad. I tried everything to remove the CPU. I started with isopropyl alcohol. No luck. I tried dental floss, but there is basically no space between the CPU and cooler, since I used very little thermal paste. I tried heating it to 70°C with a hair dryer (the heat pipes work quickly in reverse, and I watched the temperature with my infrared thermometer). I tried freezing it to -20°. Nothing worked. Due to the orientation of the cooler, I wasn't able to action the CPU socket lever. Since the Socket AM3 motherboard is five years old, I wasn't going to buy a new CPU for it. I figured since I pulled the CPU without levering the socket, maybe it will go back in... and it did, but now the fit is loose. So I broke the motherboard. I've got some other things to take care of before buying a new Skylake system, so to tide myself over, and I'm picking up a used ASRock H81 PRO BTC Motherboard for $40 and a new i3-3170 to go in it. I guess it was time to buy a new system, since this "cheap fix" will double the speed of my system.

2015-11-02, 04:33	#5
LaurV Romulan Interpreter "name field" Jun 2011 Thailand 3×23×149 Posts	I somehow missed this thread yesterday. Some thermal pastes are "curing" type, this means that you have to heat them to ~120°C (depending on paste) and it will "cure", i.e. it can really melt like tin, and "weld" the CPU to the cooler. They will also "cure" at lower temperature, in longer time, like for example 90 hours at 85°C, but when cured like that, they will only reach about 96% of the thermal transfer capacity. Those types are much better for thermal transfer, after curing they have the "metal" transmissibility for heat, but they are not so commonly used, exactly from the reason what happened to you - they are designed for systems that you don't intend to take apart. We use thingies like that in our industrial embedded touch-terminals here in production. Make sure you didn't use such a paste, before attempting to take it of. If you used "arctic silver 5", for example, and your CPU "accidentally" got over 105°C for few minutes, or you had more than 300-400 hot/cold cycles, then you may be just an unlucky son of a gun. If you did, you will end up damaging everything long before being able to take the cooler off the CPU. Some of them still have specific solvents, which will soften it, even after curing, but some not. Google for "curing thermal paste" for clarifications, but be careful that a lot of people talk about "curing" without having idea what it is. The best way is to take the datasheet of the paste - provided by the manufacturer, I can help you if you tell me exactly the type - and look how it behaves. The alternative is to put the whole assembly (cooler and CPU) in a vice, take a piece of wood and a hammer, and delid the toy. You only have to look carefully of the hitting direction, put it on the long-side of the die, and add a pillow in the back, for this setup (some other videos add a pillow, but they are not so funny as the CPU doesn't fly away ). DON'T TWIST the CPU! You will damage the die. And you may damage the die if you hit it laterally too. You are only allowed to hit it longitudinally, due to how the golden balls/bumps are fixed on the PCB. Last fiddled with by LaurV on 2015-11-02 at 04:47

2015-11-02, 05:15	#6
Mark Rose "/X\(‘-‘)/X\" Jan 2013 C22₁₆ Posts	The chip was run hot for a while a year ago, when the Hyper 212 Evo fan failed. I don't know how hot it got exactly, but it was thermal throttling. It's max operating temperature was 71°. Now that I think about it, I probably just used the stock paste on that cooler. I wasn't overclocking or anything, I simply wanted something quieter than the stock cooler I had been using (whose fan was getting noisy). I don't think hitting it would work as the circuit board protrudes around the die on all four sides. It looks like the attached picture. I only tried twisting it with my fingers, but with enough force to leave dents in my thumbs. Anyway, I'm about to power on the replacement system. Attached Thumbnails Last fiddled with by Mark Rose on 2015-11-02 at 05:16

2015-11-02, 06:48	#9
kladner "Kieren" Jul 2011 In My Own Galaxy! 2·3·1,693 Posts	Would it be possible to insert a very thin blade to apply pressure right at the junction of the heatsink base and the heat spreader? Here are some possibly useful Xacto models. Maybe get the junction wet with alcohol or Arctic Silver cleaner, and then either scrape along the edge, or judiciously tap the knife handle. Getting enough purchase to twist or pry with the blade might get non-destructive results. Last fiddled with by kladner on 2015-11-02 at 06:49

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
conductivity of thermal paste	wildrabbitt	Hardware	5	2017-05-10 13:07
Best Thermal paste	kracker	Hardware	50	2015-02-23 23:41
XP Pro x64 Cut & Paste	storm5510	Software	5	2013-10-29 12:34
i7 thermal throttling?	fivemack	Hardware	20	2009-03-20 23:49
BitTorrent Users Beware!	ewmayer	Lounge	4	2005-06-19 02:21

2015-11-01, 16:42	#2
chris2be8 Sep 2009 2²·607 Posts	Been there, done that. I managed to get the CPU off the cooler by soaking the paste with the solvent that came with my supply of Arctic Silver (basically white spirit with an additive that works as an antioxidant) round the edge of the CPU until the thermal paste softened enough to get the CPU off. But it took over 2 hours (I was working on my patio since the bottle of solvert said to use it in a well ventilated place). Chris

2015-11-01, 20:08	#3
EdH "Ed Hall" Dec 2009 Adirondack Mtns 5,261 Posts	I had a friend many years ago that used to test the physical properties of IC packages. One of the tests he performed was to see how much stress it took to "twist" the top off an IC, before looking for degradation of the internal connections. That got me thinking about breaking the bond between heat sinks and ICs by a slight twisting motion, which is what I commonly use when removing the CPU coolers in older machines. I haven't run across any which needed a lot of force, however I have had a couple which, due to the type of mounting, didn't offer any twisting room. I wonder if a small amount of twisting force would work for a CPU that is bonded as well as Mark Rose's. I also wonder if Acetone would be detrimental to a CPU, since it does a real good job deforming some plastics. I suppose my preferred solvent would be 99% alcohol, which I have on hand and use for electronics. As to the original motherboard - Are you confident it is destroyed? Many types of IC sockets that close the contacts on the pins, have quite a bit of space behind the closing portion if no pin is between them. This could account for sloppiness if the contacts were in the closed position and you placed the CPU back into the socket.

2015-11-01, 21:26	#4
Mark Rose "/X\(‘-‘)/X\" Jan 2013 2×1,553 Posts	I'm also concerned about twisting the circuit board off the bottom of the CPU IC. I almost used and adjustable wrench but my biggest one was slightly too small. From what I've read, acetone is a no-no.

2015-11-02, 06:36	#7
LaurV Romulan Interpreter "name field" Jun 2011 Thailand 3·23·149 Posts	That only looks like a silicone rubber or epoxy resin ring to me. The process to "dig sculptures" into the FR4 PCBs is not easy, I don't believe that the metal cap is "buried". It would be the first time I see it. You can try some of what this guy does here, (for about the same CPU). OTOH, talking with someone here, they say the Arcticlean solution is xylene based, and you can use any carburetor cleaner solution, or the additive that is added to the gasoline on your car to make it more powerful (here you can buy small bottles of it in any superstore) to dissolve that paste. Put it in and let it for few hours. If it disappears completely, CPU, cooler and all, don't blame me.