40mm Fan Shootout: Part 2 – Cooling

Introduction:

At this point, you’ve likely already seen the noise analysis done in part 1. Many customers tell me that Noctua fans are “silent”. I suppose they mean that the fan noise is inaudible at their seating distance. This of course would indicate that the sound is at or below ambient levels. The data shows that it is not, in fact silent, but how well does it cool? Welcome to part 2 of the four part shootout:

  • Part 1 – Noise
  • Part 2 – Cooling
  • Part 3 – Airflow
  • Part 4 – Conclusion

Test Methodology

Electronic devices are characterized by their ability to operate in various environments. From a military perspective, systems need to be able to withstand the environment in which they operate. This may mean anything from rigorous EMI testing, to climatic testing.

Image courtesy of Wikipedia

The McKinley Climatic Laboratory at Eglin Air Force Base is famous for its climatic test chamber. The laboratory consists of various rooms that can replicate almost any weather environment. Temperatures can range from -62C to 77 C. Rainfall can reach 15 inches per hour and wind nearly 60 knots.

Now, I may be mistaken but I assume that most users will use their MiSTer in an indoor, climate controlled, environment. We can cross off the snow, wind, and most of the temperature requirements.

Measurement Equipment:

While I don’t have access to any such chamber for testing a 107mm x 70mm circuit board, I do have access to a FLIR TG167 thermal camera, and an ETEKCITY Lasergrip 774 infrared (IR) thermometer. That should work well enough for our purposes.

FLIR TG167 (left) and Etekcity Lasergrip 774 (right)

Cooling configurations:

Setup:

Before starting the test, the Etekcity IR thermometer to measure the temperature of the table, next to where the DE10-nano was sitting. This established an ambient temperature for the test. If any fan is able to cool below this temperature, without some kind of refrigerant, it will automatically win because it has broken the first law of thermodynamics.

The DE10-Nano idled at the MiSTer menu for 5 minutes before a temperature was taken at the corner of the heatsink, which was previously found to be the hottest).

After this measurement, Yoshi’s Island was loaded and allowed to run for another 5 minutes before a second temperature measurement was taken. This should have allowed sufficient time for temperatures to stabilize.

The FLIR camera was initially planned to be used for all measurements, but its limitations quickly became apparent. The IR camera on the device has very low resolution (80×60). See EEVBlog’s FLIR TG165 Teardown (previous generation model) if you want to to know more. As such, it’s difficult to determine the temperature of small objects, like a 20mm heatsink. As such, the FLIR will be used for visual references and relative temperature, and the IR thermometer for more accurate temperatures. The following images were captured after 5 minutes of Yoshi’s Island.

Thermal measurements captured with the FLIR TG167

Some observations of the above image:

  • I didn’t have a working generic 12V fan at the time of this recording. Maybe I’ll update the image in the future.
  • The heatsink seemingly only decreases the temperature 3.5 C before thermal saturation. Adding thermal mass is still better than nothing, but active cooling offers much better performance (to no one’s surprise).
  • The Noctua and generic 5V fan have similar cooling performance (remember: independent of noise)
  • The Scythe 12V is better than nothing, but not good at all.

A Note About Chip Temperature Specifications

Chip specifications usually include a few thermal parameters:

  • Minimum and maximum storage temperatures
  • Minimum and maximum operating temperatures
  • Thermal junction maximum (Tjmax, not TJ Maxx)

For the purposes of this discussion, we’re most concerned with Tjmax, which Intel specifies to be 100 C. There are three operating temperature ranges for different applications: Commercial = 0 to 85C, Industrial = -40 to 100 C, and Automotive = -45 to 125 C.

Thermal Resistance Diagram from allaboutcircuits.com

What is thermal junction temperature? Essentially, it’s the chip’s internal temperature. This is the temperature modern CPUs report. This is NOT the temperature measured at the heatsink. The chip is typically wrapped in a casing, sometimes with a thermal spreader on top of it that interfaces with the heatsink through some kind of thermal interface material, like heatsink compute or thermal tape. But enough about that. Just know that, without access to an internal temperature, the surface temperature is the best we can do. It should be assumed that the junction temperature will never be lower than the heatsink temperature (Ths).

Results:

The results are fairly self-explanatory. At least now the Noctua fans (pun-intended) will have some data to back up their unquenchable love for their brown fans. For me, personally, I’ll continue to offer their fans as upgrade options through my site. But I’ll sit here wiping my brow with a crisp $12 bill (price difference between the generic 12V and Noctua fans). Now it’s up to each user to find the right combination of sound, thermals, and price, knowing that “good enough” is a subjective term left to the consumer.

End note: I don’t know if I’ve mentioned this yet, but @Tinkerplunk was the person that originally tried to convince me to use the generic 12V fans. I tip my hat to him.

End note 2: It turns out that the FLIR and the IR thermometer measure within a few degrees Celcius of each other.