all test were conducted using the official RPi power adaptor (2.5A)

the Rpi was running at 1.366GHz and 500MHz with 6 volts over voltage.

there was a slight change in ambient temperature over the duration of the tests (it became 1 degrees warmer)

I apologize for my bad pictures, just remember its all in the name.

MR-B

Step 1: Test 1

the first test was with no cooling and it ran, damn hot!

the following code was used throughout all testing to stress the CPU;

sysbench --test=cpu --cpu-max-prime=20000 --num-threads=4 run >/dev/null 2>&1
vcgencmd measure_temp sysbench --test=cpu --cpu-max-prime=20000 --num-threads=4 run >/dev/null 2>&1 vcgencmd measure_temp && sudo cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq mp && sudo cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq

vcgencmd measure_temp sysbench --test=cpu --cpu-max-prime=20000 --num-threads=4 run >/dev/null 2>&1 vcgencmd measure_temp sysbench --test=cpu --cpu-max-prime=20000 --num-threads=4 run >/dev/null 2>&1 vcgencmd measure_temp && sudo cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq mp && sudo cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq

Step 2: Test 2

for the first cooling experiment i used a Noctua NF-A6x25 5V PWM 60mm Low Noise Fan.

the fan is powered from the USB port and blows across the board as per image.

Step 3: Test 3

to determine if the position of the fan made a difference I dropped" the fan on top for this test to check for variations.

Step 4: Test 4

to see if the distance from the fan to the cpu made a difference i raised it a little over the board with toothpicks ;)

Step 5: Test 5

For this test i used a 22 x 22 x 2.5 mm stick on ceramic heat-sink

Step 6: Test 6

this test was a combination of the best fan arrangement and the ceramic heat-sink

Step 7: Results and Conclusions

On its own the RPI3 does get hot under heavy load. around 75 degrees C stock temperatures. I am sure it would start to throttle down if run continuously.

the fan i used was a high quality fan! the fan flows up to 17.2 CFM and produces upto 19.3 dB(A). The max draw of the fan is 1.3W. Unfortunately I do not know how how much the fan is being utilized through the RPI3 USB port but it is certainly sufficient as seen in my results.

Fan comparisons:

the fan on top (test 5) was not as effective as the fan (blowing through) by around 5 degrees C. Raising the fan off of the RPI3 also decreased effectiveness further by around 4 Degrees C.

The ceramic heat sink was very effective for its size! the temperature stayed around 66 Degrees C throughout the test. This was during an intense sysbench routine. this make a great passive heatsink especially at stock or lightly raised clock speeds! (Remember I am running mine at 1.366GHz.

For casual users: A Ceramic heatsink is sufficient for cooling and keeping throttling at bay.

For Speed chasers: Adding a fan made a big difference (albeit a really decent fan)

Also for speed Chasers: My RPI3B would not boot over 1.396 GHz, with 500 MHz ram and 6v overvolt.