Thermal compound replacement, RESULTS and PICTURES
YOU CAN HAVE A MACBOOK PRO THAT ISN'T SCORCHING HOT!!
(note that this will NOT fix any whine or moo; they are unrelated problems)
Here's how.
My Mother's Macbook Pro (Or Cookbook, if you will) Arrived a week ago today. I played around with it to make sure everything is alright because my mother doesn't know how to check for things like dead pixels, bad ram, or cough improperly applied thermal compound.
One of the first things I noticed after turning on the new MBP is how totally sweet it is! The second thing I noticed is how freaking hot it is: I recoiled in pain from the Fn key bar the first time I touched it. Disappointed, I started searching the web and sure enough, pretty much everyone who owns one is complaining that it's too hot for comfortable use.
Apple calling this a "notebook" instead of a "laptop" is a total semantic cop-out. It's a PORTABLE COMPUTER and I must be able to trust it around my dogs, children, valuables without the MagSafe burning up or the battery swelling and bursting.
I don't have objective figures for just how hot it was, but it was right about at my pain threshold above and below the belt, and sometimes over it. I couldn't hold my hand to it for more than a few seconds. In particular the area to the left of the touchpad was of concern. I do have before-and-after figures of the CPU and HDD; I invite yourselves to look at them:
Before After (Temps in degrees C, ambient 25 C)
50-60 26-35 CPU (idle)
76-85 56-65 CPU (load)
41 33 HDD
In particular the HDD figures are a great relief. HDDs are notoriously sensitive to temperature and even a few degrees C can cut their lifetimes significantly. Furthermore the area is now cool to the touch and I can once again rest my left palm on it without discomfort. The ranges are due to the fact that the sensor inside the core duo is flaky. In 5 seconds it can run anywhere from +-5 to +-10. Nevertheless it is accurate enough for our purposes. Below are photos, procedure, and a rough outline of test methodologies:
http://www.pbase.com/silentplummet/image/62641589
Here's the idle scenario before the procedure. The computer has been on for hours (days really) and I'm doing the work I normally do on it. TextEdit is open (to a little project I'm working on) with firefox and the temperature monitor. Alt-tab is to show that those are the only programs running. CPU temp is dead at 50 C. This is INSANE for an idle figure on ANY computer; desktop, laptop, "notebook" or otherwise.
http://www.pbase.com/silentplummet/image/62641590
The operation area and stress test. To stress the ATI chip I've jacked up the resolution, run a couple of quartz programs, SNES9x (a hardware emulator; the software shown is called "Energy Breaker") with a brutal multitasking OpenGL hardware renderer, and Google Earth. To get the CPU going I have Adobe's Lightroom processing thumbnails, and again SNES9x. For various I/O I have two shells executing yes > /dev/null, a USB mouse plugged in, and all the HDD access from lightroom. I figure it paints a pretty good picture of the "worst case scenario" of hardware stress for a laptop like this.
http://www.pbase.com/silentplummet/image/62641591
The CPU core(s) is at 76 C. I should note that at this point, as hot as the CPU is, overall the case is really not much hotter than it was before. In other words, it's just as unacceptable.
At this point I turned her off and dug in. I used a howto from Ifixit to serve as my guide. The procedure went without surprises until I got all the way to the logic board.
http://www.pbase.com/silentplummet/image/62641592
Dear lord!! That is an obscene amount of compound. It's casting a shadow over the rest of the board!! This gray gak is piled on so thick, it's no wonder the cooling system couldn't work effectively. It had even gotten all over components nowhere near the dies. That definitely cannot be good for their lifespan. Here's a shot of the heatpipe:
http://www.pbase.com/silentplummet/image/62641593
Terrible. Thermal compound has been squished out all over the place, including the chassis itself. This explains why it was getting so hot. A photo from the MBP service manual has been floating around the net, illustrating that this gross amount of compound is actually according to procedure.
http://www.pbase.com/silentplummet/image/62641594
Clean as a whistle. After removing the bulk of the compound with q-tips I used ArctiClean 2-step process to emulsify the rest and remove it with paper towels. You can see it's not perfect but it's close enough for me. I'm not overclocking this thing; I just want to run it "in spec" and have it not burn me. At this point, the CPU dies and the heat pipe interfaces should have mirror finishes. It's an overclocker's dream, and Apple already did the work for me.
http://www.pbase.com/silentplummet/image/62641595
This is how I applied Arctic Silver 5 to replace the compound I removed. Squeeze the tiniest little bit out of the syringe directly on to the die, and scrape it across with a flat edge (they recommend a razor blade but I just used a plastic ID card). Take the amount of compound you see on the Core Duo (on the right) and make a flat, even layer like the one you see on the ATI (left).
http://www.pbase.com/silentplummet/image/62641596
Turned it on and went straight for the hard stress test, after making sure everything was OK of course =) Wow! It reads 58 C in the screenshot, and doesn't go above 65C!! Moreover, there isn't even a bit of warmth above the Fn keys, and the HDD area is cool to the touch. I'd call this one a complete success. I'm idling right now and the temperature reads between 26 and 31C. Even the bottom is just slightly warm to the touch. Now I have a real laptop again!
So why did this happen?
There's a lot of confusion about the way the Macbook Pro cools itself. I admit it's confusing. Basically, Apple is shipping Macbook Pros with one cooling system, and replacing the thermal compound changes it into a very different system. Let me try to explain what I learned from digging around the hardware.
1. The built-in thermometer in the CPU is flaky. That's why you have to access it with a kernel extension and all kinds of hacks, and why Apple circumvented it completely in the cooling system. That's right: the MBP cooling system ignores the Core Duo temperature entirely.
2. The cooling system consists of a convective (my guess, I don't think anyone really knows what kind of) heatpipe which is in the base, directing heat out to two heatsinks which are then to be cooled by two fans if need be.
3. There are two temperature sensors. One is on the heatpipe itself, and the other one is on the chassis just next to the right fan. Probably the hardware monitors these temperatures and the differential between them to decide when to activate the fans and how long.
4. Behavior before the replacement procedure: The CPU core would get hot, hotter than I've ever seen a CPU go, at 80-85C. Most people confirm their MBPs also exhibit this. Where was this heat going? Well the fans didn't turn on until I put it at full load. Even when the fans did turn on, there wasn't much warm exhaust coming out of the vents at the back. The chassis heated up until it was unbearable, and most of the excess heat was being radiated away from it.
To sum up, the ineffective thermal interface between the CPU dies and the heat pipe was inhibiting heat from tripping the fan sensors. This explains why the fans didn't turn on until drastic temperatures were attained, and why the chassis got so hot. Essentially, the chassis was serving as a big heatsink for the CPU, which is the only reason it didn't overheat and shut down.
Effective, Apple, but not quite appropriate.
5. Behavior after the replacement procedure: The first thing you notice is that the fans scream from the second you turn the thing on. They aren't going full blast but pretty close to it. An effective thermal interface using an APPROPRIATE AMOUNT of AS5 (anything would do but I figured if I'm applying thermal compound, why not go for the authority) allows the heat to go straight from the cores into the heatpipe, tripping the sensor early and fast. The fans come on, I can feel hot air coming out the back, and the chassis now removed from the thermal equation is cool and comfortable again.
Of course, the thermal equation is different from before, and from the way Apple has tuned the fans to work from the factory configuration. This is more cooling than we probably need, and I foresee an update to Tiger allowing us to choose the thermal/noise tradeoff for ourselves.
Well, I hope that explains it, and I hope that those of you still suffering the abuses of your "in spec" MBP can take some hope from my findings, or are emboldened to go ahead and repeat the procedure yourself. I will post informative links here.
http://www.ifixit.com/Guide/85.1.0.html
http://www.arcticsilver.com/
http://forums.somethingawful.com/showthread.php?s=&threadid=1864582 (This is where I learned of the thermal paste issue)
Remember if you ever open up your MBP to NOT BREAK ANYTHING and please, always read the instructions before you reach in. =)
(note that this will NOT fix any whine or moo; they are unrelated problems)
Here's how.
My Mother's Macbook Pro (Or Cookbook, if you will) Arrived a week ago today. I played around with it to make sure everything is alright because my mother doesn't know how to check for things like dead pixels, bad ram, or cough improperly applied thermal compound.
One of the first things I noticed after turning on the new MBP is how totally sweet it is! The second thing I noticed is how freaking hot it is: I recoiled in pain from the Fn key bar the first time I touched it. Disappointed, I started searching the web and sure enough, pretty much everyone who owns one is complaining that it's too hot for comfortable use.
Apple calling this a "notebook" instead of a "laptop" is a total semantic cop-out. It's a PORTABLE COMPUTER and I must be able to trust it around my dogs, children, valuables without the MagSafe burning up or the battery swelling and bursting.
I don't have objective figures for just how hot it was, but it was right about at my pain threshold above and below the belt, and sometimes over it. I couldn't hold my hand to it for more than a few seconds. In particular the area to the left of the touchpad was of concern. I do have before-and-after figures of the CPU and HDD; I invite yourselves to look at them:
Before After (Temps in degrees C, ambient 25 C)
50-60 26-35 CPU (idle)
76-85 56-65 CPU (load)
41 33 HDD
In particular the HDD figures are a great relief. HDDs are notoriously sensitive to temperature and even a few degrees C can cut their lifetimes significantly. Furthermore the area is now cool to the touch and I can once again rest my left palm on it without discomfort. The ranges are due to the fact that the sensor inside the core duo is flaky. In 5 seconds it can run anywhere from +-5 to +-10. Nevertheless it is accurate enough for our purposes. Below are photos, procedure, and a rough outline of test methodologies:
http://www.pbase.com/silentplummet/image/62641589
Here's the idle scenario before the procedure. The computer has been on for hours (days really) and I'm doing the work I normally do on it. TextEdit is open (to a little project I'm working on) with firefox and the temperature monitor. Alt-tab is to show that those are the only programs running. CPU temp is dead at 50 C. This is INSANE for an idle figure on ANY computer; desktop, laptop, "notebook" or otherwise.
http://www.pbase.com/silentplummet/image/62641590
The operation area and stress test. To stress the ATI chip I've jacked up the resolution, run a couple of quartz programs, SNES9x (a hardware emulator; the software shown is called "Energy Breaker") with a brutal multitasking OpenGL hardware renderer, and Google Earth. To get the CPU going I have Adobe's Lightroom processing thumbnails, and again SNES9x. For various I/O I have two shells executing yes > /dev/null, a USB mouse plugged in, and all the HDD access from lightroom. I figure it paints a pretty good picture of the "worst case scenario" of hardware stress for a laptop like this.
http://www.pbase.com/silentplummet/image/62641591
The CPU core(s) is at 76 C. I should note that at this point, as hot as the CPU is, overall the case is really not much hotter than it was before. In other words, it's just as unacceptable.
At this point I turned her off and dug in. I used a howto from Ifixit to serve as my guide. The procedure went without surprises until I got all the way to the logic board.
http://www.pbase.com/silentplummet/image/62641592
Dear lord!! That is an obscene amount of compound. It's casting a shadow over the rest of the board!! This gray gak is piled on so thick, it's no wonder the cooling system couldn't work effectively. It had even gotten all over components nowhere near the dies. That definitely cannot be good for their lifespan. Here's a shot of the heatpipe:
http://www.pbase.com/silentplummet/image/62641593
Terrible. Thermal compound has been squished out all over the place, including the chassis itself. This explains why it was getting so hot. A photo from the MBP service manual has been floating around the net, illustrating that this gross amount of compound is actually according to procedure.
http://www.pbase.com/silentplummet/image/62641594
Clean as a whistle. After removing the bulk of the compound with q-tips I used ArctiClean 2-step process to emulsify the rest and remove it with paper towels. You can see it's not perfect but it's close enough for me. I'm not overclocking this thing; I just want to run it "in spec" and have it not burn me. At this point, the CPU dies and the heat pipe interfaces should have mirror finishes. It's an overclocker's dream, and Apple already did the work for me.
http://www.pbase.com/silentplummet/image/62641595
This is how I applied Arctic Silver 5 to replace the compound I removed. Squeeze the tiniest little bit out of the syringe directly on to the die, and scrape it across with a flat edge (they recommend a razor blade but I just used a plastic ID card). Take the amount of compound you see on the Core Duo (on the right) and make a flat, even layer like the one you see on the ATI (left).
http://www.pbase.com/silentplummet/image/62641596
Turned it on and went straight for the hard stress test, after making sure everything was OK of course =) Wow! It reads 58 C in the screenshot, and doesn't go above 65C!! Moreover, there isn't even a bit of warmth above the Fn keys, and the HDD area is cool to the touch. I'd call this one a complete success. I'm idling right now and the temperature reads between 26 and 31C. Even the bottom is just slightly warm to the touch. Now I have a real laptop again!
So why did this happen?
There's a lot of confusion about the way the Macbook Pro cools itself. I admit it's confusing. Basically, Apple is shipping Macbook Pros with one cooling system, and replacing the thermal compound changes it into a very different system. Let me try to explain what I learned from digging around the hardware.
1. The built-in thermometer in the CPU is flaky. That's why you have to access it with a kernel extension and all kinds of hacks, and why Apple circumvented it completely in the cooling system. That's right: the MBP cooling system ignores the Core Duo temperature entirely.
2. The cooling system consists of a convective (my guess, I don't think anyone really knows what kind of) heatpipe which is in the base, directing heat out to two heatsinks which are then to be cooled by two fans if need be.
3. There are two temperature sensors. One is on the heatpipe itself, and the other one is on the chassis just next to the right fan. Probably the hardware monitors these temperatures and the differential between them to decide when to activate the fans and how long.
4. Behavior before the replacement procedure: The CPU core would get hot, hotter than I've ever seen a CPU go, at 80-85C. Most people confirm their MBPs also exhibit this. Where was this heat going? Well the fans didn't turn on until I put it at full load. Even when the fans did turn on, there wasn't much warm exhaust coming out of the vents at the back. The chassis heated up until it was unbearable, and most of the excess heat was being radiated away from it.
To sum up, the ineffective thermal interface between the CPU dies and the heat pipe was inhibiting heat from tripping the fan sensors. This explains why the fans didn't turn on until drastic temperatures were attained, and why the chassis got so hot. Essentially, the chassis was serving as a big heatsink for the CPU, which is the only reason it didn't overheat and shut down.
Effective, Apple, but not quite appropriate.
5. Behavior after the replacement procedure: The first thing you notice is that the fans scream from the second you turn the thing on. They aren't going full blast but pretty close to it. An effective thermal interface using an APPROPRIATE AMOUNT of AS5 (anything would do but I figured if I'm applying thermal compound, why not go for the authority) allows the heat to go straight from the cores into the heatpipe, tripping the sensor early and fast. The fans come on, I can feel hot air coming out the back, and the chassis now removed from the thermal equation is cool and comfortable again.
Of course, the thermal equation is different from before, and from the way Apple has tuned the fans to work from the factory configuration. This is more cooling than we probably need, and I foresee an update to Tiger allowing us to choose the thermal/noise tradeoff for ourselves.
Well, I hope that explains it, and I hope that those of you still suffering the abuses of your "in spec" MBP can take some hope from my findings, or are emboldened to go ahead and repeat the procedure yourself. I will post informative links here.
http://www.ifixit.com/Guide/85.1.0.html
http://www.arcticsilver.com/
http://forums.somethingawful.com/showthread.php?s=&threadid=1864582 (This is where I learned of the thermal paste issue)
Remember if you ever open up your MBP to NOT BREAK ANYTHING and please, always read the instructions before you reach in. =)
Mac mini, Mac OS X (10.4.6)
