Should I replace the thermal paste on my mini?

I pulled apart my mini tonight and re-applied Arctic Silver 5. I have only been up for about 2.5 hours and have yet to produce any thermal cycles, but the temperature is already down a good 6 C from before.

The process was utterly painless. For anyone that has taken apart a MBP1,1, you'll know what a joy this model is to work on.

For anyone with a modicum of skill, I would say dive in and remove the standard thermal compound. After AC5 has time to reach its full efficacy, I wouldn't be surprised to see an overall 10 C drop across the board. Even know, the 6 C is very much welcome.

As I write this, it is sitting at steady 46 C. Before AC5, it ran anywhere between 51-54C. I must admit I hesitated, but ultimately, I'm glad I did it.

Preliminary Data

I've had time to gather some preliminary data and run several heat cycles. It's now been about 120 hours since the application of AC5. I've also kept the ambient temperature outside the unit constant (increased at times actually, which made little difference in all honesty), at ~21 C.

Firstly, I applied the paste using the "rice drop" method. I placed a dab in the centre of each diode and squished it using the heatsink pads. I did not spread it (as is suggested by AC) as that leaves room for thinner points and often results in thermal paste leaking out the sides and acting as an insulator (the opposite of what we want).

The temperatures now generally remain at about 44 to 47 C during general use (surfing the web, playing songs in iTunes, mail open, and a few other general miscellanea running). This of course is with the mini outputting to a 23" monitor running at resolution of 2048x1152 pixels and running a secondary HDD (5,400 RPM 500 GB) on top of the primary SSD (so it's tight in there). When the monitor is put to sleep and the system allowed to idle (not sleep, but idle), the temperatures drop to around 41 (after several hours). Depending on the draw to the GPU, your mileage may vary.

During load, the temperature rise is much more progressive. The stock paste saw my mini spike in heat compared to the AC5, which sees a much smoother and gradual progression. The fan behaves the same in both cases.

One thing to note, when the chip is stressed, the CPU remains hot (about 50-51) for a considerable while, eventually coming back down to it's resting temp of ~45 C. I suspect this has to do with the extreme constraints of the components inside. Creating that much heat must bake the system and while the fan can cool the CPU, the other components will eventually heat up as well. Since a 51 C is actually the running temperature, the fans will not remain at a higher RPM to move the heat away, hence the system remains in that state longer). Again, this is temporary and after about 30 minutes of light use, it will subside.

At peak stress, the unit has not pushed past 70 C. The stock compound saw temperatures substantially higher (sometimes in almost 80 C). But I have no yet kept it under sustained load (will update when I reach AC5 thermal efficiency).

Final Stages of Reaching Thermal Efficacy

I have hit the 200 hour mark and have managed 4 full heat cycles (as outlined by AC5). The temperatures remain relatively stable from what was initially recorded. The system remains at ~47 C during light use. It rises to ~51 C with moderate use. I doubt there will be much of a reduction from this point on.

One thing to note. The application of AC5 doesn't necessarily "drop" the temperatures, rather than make the cooling more efficient. It's a bit misleading to claim that the CPU runs cooler because that rests entirely on the cooling system and that has not changed. AC5 certainly smoothens the temperature gradient and allows the CPU to rest at lower temperature, but outside of that, there won't be a miraculous change.

Additional Cooling Strategies

I should also mention that I've heard a lot about people raising their minis to allow more airflow. I put this to the test. I ran the unit raise 30 mm off the desk. No changes to temperature. So I went a step further and pulled the base off. That saw a 2 C reduction across the board. I went another step further and placed a small fan (blowing around 40 M3/H) aimed upwards. Surprisingly that saw no change in temperatures (above the 2 C from having the base removed). My theory would be that the components are so tightly packed and hidden away from the external fan, that blowing a bit of air on them was insufficient. It might be the case that a large fan would see a reduction to temperatures, but I'm more inclined to think that it would require a replacement of the existing fan with something more powerful (I want to look into this actually).

OK,

I did it!

I popped out the system board and removed the heat-sink and as expected, big gobs of grey paste. I cleaned up both the CPU, GPU and heat-sink with IPA (alcohol not the beer) then applied the Arctic Silver.

There isn't a specific guide for the Mini Server Mid 2011, but this one was pretty close and I used it to guide me.

The only differences I found were the position of some of the screws.

Things I learned

• Removing the wired connectors, you must pull UP not in the direction of the wires or the gold contacts pull out!
• The power connector is pretty stiff but comes off with little or no issue
• The T8 sprung screws holding the heat-sink down have pretty shallow heads, so make sure your screwdriver is in good condition and the tip is not mullered
• I don't have a 'special' tool for removing the board. I tried with 2x screwdrivers, as I had read, that didn't work for me. I found that some gentle pressure on the front edge of the board pressing against the round opening was enough to get it to pop out, but I had to remove the top (bottom) drive to get access to the board edge.

As previously mentioned, this, in theory, will void your warranty. But if you do it right, no one will ever know, unless of course you balls it up!

My mini is on the desk and 'old' style mini hard drives are stacked on top of it. I was worried about pressing the air intake closed and so have propped the front of the mini server with 2x rubber feet, infact each of the devices are propped up with rubber feet to allow air flow around the devices.

Lastly I added Fan Control pref pane to jack up the fan speed a bit, still got it set slow to keep the noise down, but its slightly faster.

Screen

replacing the heatsink's interface reduced my mini cpu temp to 55 degrees c idle (down from 65~70 C) and it rarely tops over 85 degrees C with plex running. I used Arctic Silver 5 and it was well worth the effort. My mini is not nearly as well vented as yours, in fact it pulls in a fair amount of heat from my receiver. Considering the environment, the AS5 dropped the operating temperature more than 10 degrees. Hope that helps.

Macmini 2011:

• 8GB ram
• 60GB SSD
• 1 TB WD HD

:::::::::::::::::::::::::::::: idle load :::::::::::::::::::::::::::::::::::::::::::::::::::

Treatment I:

condition 1: flat placement on class surface. (horizontal)

• idle temp (out of the box) 69.3 degrees C (on average over 10,000 samples taken over 2 days)
• Ambient air temperature 22.6 degrees on average over testing period.
• Case temperature: 43.4 degrees measured with ON-409 and ON-909 Series Surface Thermistor Temperature Sensors placed at positions 3 cm from each corner on the top of the mini. 5th sensor placed just below the apple. Hotspots verified using thermograph. 1,000,000 readings taken over 2 days.

condition 2: Mini placed (vertical) in a NewerTech NuStand Alloy

http://www.geeky-gadgets.com/wp-content/...

repeated tests in treatment 1

results:

• ambient air temperatre 21.9 degrees C
• idle temp 64.1 degrees
• surface temperature 38.9
• ambient air temp found to be insignificant (statistically)
• Data fed into matlab and processed in all cases

Conclusion: mini orientation has a significant impact on operating temperature possibly due to the flat surface area and the pattern of air movement created when operating. Eddies form on the surface producing a boundary layer impeding removal of heat. In the upright postion, formation of eddies are minimized allowing for convective currents to dissipate heat more effectively.

Treatment II

• removing and replacing thermal interface with Arctic Silver 5 via the line method.
• 200 hour break in period with 4 cool down cycles each 12 hours in duration.
• repeat procedures in Treatment 1.

condition 1: (horizontal)

• Ambient temperature 23.1 degrees
• Idle temperature 47.2 degrees average
• surface temperature 37.4 degrees average

condition 2: (vertical)

• Ambient temperature 22.6 degrees
• Idle temperature 45.9 degrees
• Surface temperature 38.1 degrees

Conclusion: The internal cooling system's efficiency was able to keep the mini's temperature stable in both the vertical and flat orientations. Although we saw 1.3 degrees difference between the positions, the orientation was not significant at idle processor loads, thus it made no difference. Temperatures did not reach high enough levels to allow convective passive cooling to have an effect.

::::::::::::::::::::::::::::::::: under load ::::::::::::::::::::::::::::::::::::::::

Treatment I (Stock thermal interface) 4 instances of yes > /dev/null

condition 1: (horizontal)

• ambient temperature 23.2 degrees
• processor temperature 98.1 degrees
• case temp 67.9 degrees

Condition 2: (vertical)

• ambient temperature 22.7 degrees
• processor temperature 92.1 degrees
• case temp 61.9 degrees

Treatment II (AC5 replacement interface) 4 instances of yes > /dev/null

Condition 1: (horizontal)

• ambient temperature 23.7 degrees
• processor temperature 88.1 degrees
• case temp 55.8 degrees
• fan speed 5430 (average)

Condition 1: (Vertical)

• ambient temperature 22.5 degrees
• processor temperature 86.1 degrees
• case temp 35.2 degrees
• fan speed 4450 rpm (average)

Conclusion: Mini orientation had a small effect on processor temperature between the two position as the efficiency of the active cooling system was able to keep the mini within 2 degrees between the two conditions after AC5 was applied, indicating that active cooling had a significantly greater impact (statically) over passive cooling. The remarkable drop in case temperature seen in both idle load and max load demonstrate that orientation of the mini reduces the work load on the active cooling system as seen in fan speeds resulting in quieter operation. A combination of orientation and thermal interface placement appears to have the greatest effect. Temperature spikes where all seen initially before temperatures normalized.

Running plex with sample reference 1080p mkv video saw fan speeds average out at 2297rpm and cpu temperature average 81 degrees C. More testing is required to verify these numbers, but since I'm perfectly happy with the results, I'm not going to go through any more trouble with this.

Before you comment that I have way too much time on my hands, note that I'm finishing up my masters in engineering and this was simply an exercise in curiosity. The experimental design could be more controlled, but its only a macmini.

Errata

I don't see any issue with the cooling system once the thermal interface has been replaced. Drilling holes in the case does little to improve airflow IMHO and only compromises the integrity of the case. Properly applying AC5 should be sufficient. Measuring the airflow volume with and without the bottom attached yielded minimal flow rate increases which where measured and barely detectable with a research quality flow meter. Perhaps (if one was inclined), adding a thermoelectric peltier cooler would be a more viable option if you're intention was to drop the temp to sub 40degree C levels.

Search on macrumors for my posts: ( primarily http://forums.macrumors.com/showthread.p... )

Drilling the ventilation holes hardly made a difference at lower speed however I am under the impression that it makes a difference at higher speeds. (used to be involved in electronics and have a drill template to create evenly spaced holes)

Since the 2011 has more often higher fan speeds I consider air filtering a welcome addition (see my other posts on this subject). With overhead lighting (ceiling lights) the filter is not even noticeable. Less dirt buildup internally = better cooling. No dismanteling = no warranty invalidation.

I've also a post on references to safe operating temperatures and running this machine above 80C is not advisable, it will impact reliability. ( http://forums.macrumors.com/showthread.p... )

Increasing the idle fan speed to 2200 rpm gives it about the same idle temperature as the 2010 mini. I am only running Windows 7 (I know) and am using MacFan0_65 utility to set the fan speed after startup. Wrote a few *.bat files and have shortcuts on the desktop to set the speed to 1900 / 2100 / 2300 plus one for checking the temperature. Initially I could not live with the 2011 upgrade but now I can (need the improved data throughput from Sandy Bridge)

I love the Mac mini (especially the 2010) since there is no equivalent quiet, same small form factor PC.

I am running Windows 7 and on both the 2010 and the 2011 Mac Mini I find that the fan kicks in at a higher temperature than under OS X. This makes the situation worse for those running Windows.

Initially I drilled a series of 1/8" holes in the bottom cover plate in the area above the grill surrounding the WiFi antenna. I've have enlarged these now to 11/64" but this makes the placement of some rubber feet critical since the bottom flexes a lot more which was not the case earlier on.

I get the impression that at 2300 rpm the front of the Mac Mini is cooler than at the back. This leads me to believe that the body plays a large part in the (passive part of the) cooling.

The 2010 Mac mini uses approx 8 watts at idle and the 2011 i5-2530M about 15 watts and imho the cooling system is not sufficient enough (at idling at least) for the 2011 mini. Something must be wrong with the fan control as well under BootCamp.

I am now seriously considering adding two heatsinks on the outside of the 2011 Mac Mini ( http://www.aavidthermalloy.com/cgi-bin/e...).

I intend fastening one on each side with doublesided adhesive thermal tape. This should aid in channeling away an additional 10 - 15 watts of heat from the body.