ICE Tower Raspberry Pi 4 Overkill Cooler

I know I’ve written a number of posts relating to heat issues with the Raspberry Pi 4, just check back through previous posts. I’ve also reviewed a number of different cases evaluating their cooling performance, and noise levels. However, I don’t think things are so bad as to need this substantial piece of hardware, but it exists nonetheless so let’s look to see if it does solve a problem I’m not yet aware of. Or if this is just a large piece of colourful bling mounted atop our awesome Raspberry Pi 4.

What is it?

The Heatsink is designed by 52Pi, their website is a bit incomplete, I guess they are busy creating crazy products link this. Mine was supplied by Seeed with an RGB fan, as the “Blink Blink ICE Tower“. There is another RGB fan equipped version, this time with a black heatsink and fan frame, the “Black Warrior“. Finally, there is a version with a mono colour fan (blue I think) “ICE Tower“.

The ICE Tower cooler includes the large heatsink and fan, mounting hardware and a plastic base which appears to be optional to use. The heatsink is comprised of a multi-layer base which interfaces with the ARM processor on the Raspberry Pi, heat is transferred to the fin stack via a single U shaped copper heat pipe. Which in turn can be cooled via a 40mm fan.

It’s compatible with the Raspberry Pi 4, and 3 B+. The installed fan can run at 5V or 3V, the latter running at a lower speed and therefore slightly quieter.

Who’s it for?

Well based on my previous testing there are a number of cases which can provide sufficient passive or active cooling to handle even sustained stress loads. So my initial guess this is aimed at someone who just wants to ensure their Raspberry Pi is ALWAYS cool regardless of load or ambient temperature, or for someone that wants a “statement” heatsink. The kind of person that is likely to go for the RGB version and be slightly disappointed there isn’t a way to control the exact colours of the RGB fan.

I fully expect with a large amount of metal, a large surface area provided by the fin stack, its ability to radiate heat without a fan should be significant, such that it will handle anything I can throw at it. Let’s see!

Testing

Having struggled to make the household quiet enough to be able to take noise level measurements, I’m not keen on attempting that again, especially as noise from the ICE Cooler isn’t an issue. It’s quiet at both 5V and 3.3V.

Instead, I’m focusing the testing on the cooling capacity under maximum stress load. For this, I’m once again using my slightly modified version of Stressberry.

Stress Test Results

As seen from the previous testing, anything open with active cooling can easily avoid thermal throttling. As tests are performed in a non-temperature controlled environment, it is necessary to measure the ambient temperature throughout the test and subtract that from the core CPU temperature measurement. Where results are of specific tests, and not comparisons, the absolute CPU temperature is used to show how close to thermal throttling the results are.

In order to evaluate the ICE Tower Cooler fully, I wanted to compare the results against another common cooling solution for the Raspberry Pi 4, the Pimoroni Fan Shim with a small heatsink. In order to push the limits, I also took the opportunity to overlock the Raspberry Pi 4.

Without triggering the warranty bit, I was able to Overclock my 2GB Raspberry Pi4 to 2.147GHz, a 43% increase from the stock maximum core clock frequency of 1.5GHz. The GPU frequency was set to 600MHz. This was achieved with the following settings in /boot/config.txt

[pi4]
dtoverlay=vc4-fkms-v3d
max_framebuffers=2
over_voltage=6
arm_freq=2147
gpu_freq=600

ICE Tower

With the Raspberry Pi4 in its stock configuration, the ICE tower is capable of avoiding thermal throttling even without the fan running.

In contrast to the stock performance results, the ICE tower needs to have the fan running in order to avoid thermal throttling from the 2.147GHz down to 1.073GHz. The gains seen from running at 5V instead of 3.3V are small. At least at this level of load and temperature, running at 3.3V is sufficient, and also a little bit quieter.

Pimoroni Fan Shim (+small heatsink)

Using the Pimoroni Fan Shim with a small heatsink sees the CPU temperature increase from 48°C to 68°C when overclocked to 2.147GHz.

ICE Tower vs Pimoroni Fan Shim

The overclocked Raspberry Pi 4 with the ICE Tower cooler manages to outperform the Pimoroni Fan Shim by ~10°C.

The Raspberry Pi 4 thermally throttles at around 80°C, the Pimoroni Fan Shim sees a safe margin under full load of 32°C in stock configuration and 12°C when overclocked to the maximum level possible without trigging the warranty bit. The ICE Tower cooler extends these margins to 40°C and 22°C respectively.

There appears to be sufficient cooling headroom to increase the overclock further whilst keeping the system cool for those willing to risk taking the Pi past known safe limits.

Phoronix Performance Test Results

Whilst this testing has focused on stress and temperatures under maximum load, the reason for overclocking and needing more extensive cooling is often to increase performance. In order to quantify the gains in performance, I ran a suite of tests using Phoronix Test Suite, I’ve published the results over on OpenBenchmarking.org – https://openbenchmarking.org/result/1911164-AS-1911122MA86

A future post will go into further details on these performance tests. The OC3 values represent the 2.147GHz / 600MHz test configuration detailed above.

Parts Tested

Note, Amazon UK links are affiliate links, meaning I may make a small amount of money in return from you following the link and buying the product.

• Raspberry Pi 4 (Amazon) (Pimoroni) (The Pi Hut) (Seeed)
• Raspberry Pi 4 official PSU (Amazon-UK) (Amazon) (Pimoroni) (The Pi Hut)
• Blink Blink ICE Tower Cooler (Seeed) (Amazon- UK) (Amazon) (The Pi Hut)
• Pimoroni Fan Shim (Amazon-UK) (Amazon) (Pimoroni)