Raspberry Pi Zero 2 W mini review – Benchmarks and thermal performance

The Raspberry Pi Foundation launched the Raspberry Pi Zero 2 W board yesterday with the main difference against Raspberry Pi Zero W board being the much faster Raspberry Pi RP3A0 SiP with a Broadcom quad-core Cortex-A53 processor clocked at 1.0 GHz and overclockable to 1.2 GHz.

I received my sample shortly after publishing the announcement, and I had time to test it. Since the main difference is the processor, I’ll focus this review on benchmarks and whether additional cooling is required for the board.

Raspberry Pi Zero 2 W kit unboxing

If you purchase the board for $15, that’s all you’ll get, but Raspberry Pi Trading sent me a kit with Raspberry Pi Zero 2 W SBC, a USB OTG adapter, a mini HDMI to HDMI adapter, the CSI camera cable, and four rubber pad for the enclosure that comes with three covers: full, hole for 40-pin GPIO header, or hole for the camera.

Raspberry Pi Zero 2 W Kit

The board is exactly the same size as the Raspberry Pi Zero W, and so are the ports’ arrangements. Raspberry Pi Zero 2 W boardI have a Raspberry Pi Zero somewhere, but I’m unable to find it…

First boot and system information

I downloaded the Raspberry Pi OS May 2021 image from the official website and flashed it to a MicroSD card with USB Imager. After inserting the microSD card, an HDMI cable, and Logitech USB dongle for a wireless keyboard and mouse combo, I connected the power supply. I specifically used MINIX NEO P2 100W USB adapter so no joking around!

Raspberry Pi Zero 2 W Review

It would boot to the desktop, but I was unable to use the keyboard or mouse for that matter. I switched to a USB keyboard still no luck. I replaced the USB OTG adapter with mine, but no luck. I tried the microSD card in a Raspberry Pi 4 just to make sure there were no issues with the image itself and tried again in Raspberry Pi Zero 2 W. It worked somehow and I could carry on with the setup wizard.

Raspberry Pi OS setup Pi0 2 W

I also enabled SSH to have easier access to the command line, and made sure I had the latest version of the packages:


Some system info:


It’s detected as the same Broadcom BCM2835 as found in Raspberry Pi Zero, instead of BCM2710 / BCM2710A0, but it does not matter as four Cortex A53 cores are detected.


The CPU frequency will vary between 600 and 1000 MHz, the idle temperature is around 47°C, and out of 493 MB total memory, 181.4 MB is used with nothing running. Needless to say, the Raspberry Pi Zero 2 W is not designed to be used as a desktop machine, but for some specific tasks or even headless projects, it should be great value.

I could confirm idle temperature and CPU frequency (600 MHz) with vcgencmd utility:

Raspberry Pi Zero 2 W Benchmarks with Phoronix

Let’s install Phoronix


I’m using the older Phoronix Suite 8.8.1 to match the version I used on Raspberry Pi 4 review.

Let’s start the benchmark for a comparison with others:


The full results can be found on OpenBenchmarking, but let’s check some of the specific results.

Raspberry Pi Zero 2 W Benchmark-Phoronix John The Ripper

John the Ripper is a multi-threaded benchmark, and here the Raspberry Pi Zero 2 W performs surprisingly well compared to a Raspberry Pi 3. Software/firmware may have improved since then, as in theory, the latter should be 40% faster with a score of about 729. You’ll also notice Raspberry Pi 4 is only marginally better than Pi Zero 2 W, and there’s an easy explanation the Pi 4 was naked at the time, with the firmware lacking optimizations released later on. That just means the Raspberry Pi Zero 2 W performs well without heatsink even in a room at about 28°C.

C-Ray Benchmark Raspberry Pi Orange Pi

C-Ray should the Raspberry Pi Zero 2 W getting very close to Pi 3 performance.

Raspberry Pi Zero 2 W FLAC Audio EncodingFLAC audio encoding is no different, except we’ll notice the superior performance on Raspberry Pi 4 which must have some specific instructions that accelerate encoding.

Raspberry Pi Zero 2 W temperature chartFor reference, this is what the temperature chart looked like during testing, so never above 75°C.

Raspberry Pi Zero 2 W benchmarks with SBC Bench

We can have a closer look at the performance and temperature under by various loads installing SBC Bench scripts:


Let’s run the benchmark:


No throttling was detected, and the temperature never went over 63°C. The room temperature was about 26 to 27°C at the time of the benchmark.

Raspberry Pi Zero 2 W SBC-Bench temperature chart

This is how results compare to other Raspberry Pi boards at stock frequency.

7-zip benchmarks

AES benchmarks
AES-256 with 16KB blocks

Let’s try to overclock the board to 1.2 GHz by editing /boot/config.txt:


then reboot and check and check the frequency:


Let’s run SBC Bench again:


Raspberry Pi Zero 2 W Overclock Temperature

SBC Bench detects “Silent throttling” but I’m not sure what that means with the temperature not exceeding 70C, and no apparent drops in frequencies in the log:


An overclocked Raspberry Pi Zero 2 W now performs just as well as Raspberry Pi 3 B+ in the 7-zip test, at least with the numbers we’ve got.

7-zip Raspberry Pi Zero 2 W Overclocked
7-zip results

It’s too easy, so let’s put the board inside its enclosure to warm things up a bit…

Raspberry Pi Zero 2 W case

SBC bench:


I got a few problems with RPI Monitor going down during the review, so no chart this time, but we can check the temperature in the log:


Still always under 70°C. I find the average load a bit low at 2.88, so I repeated the test to double-check, and there are indeed there are four p7zip processes running at the same time. It’s just the benchmark may not run long enough to get close to a load average of 4.

7zip load average

Although it will not be needed in most cases, I still decided to test the board with a heatsink to see how much the temperature would lower.

Raspberry Pi Zero 2 W heatsinksBut the heatsinks I used with Raspberry Pi 4 were not quite suitable for the smaller board, especially the “ICE Tower CPU cooling fan“… So I went with a smaller heatsink that could also be used within the enclosure.Raspberry Pi Zero 2 W HeatsinkLet’s run SBC Bench script one last time for today:


Same silent throttling warning, but 7-zip results were marginally better (up to 3,648 points), and the temperature only went up to about 64°C.

Pi Zero 2 W Heatsink overclocked temperatureThat’s about 6°C cooler than our overclocked test without a heatsink.

Conclusion

The Raspberry Pi Zero 2 W performs nearly as well as a Raspberry Pi 3 B+ especially when overclocked, and despite its small size has no problem staying cool. That means most people will not need to add a heatsink to the board, unless possibly for full loads for an extended time, or higher room temperature (35°C+). However, the smaller memory capacity (512MB) and available connectors will probably make it suitable for different use cases.

I’d like to thank Raspberry Pi Trading and Eben Upton for sending a review sample. I may continue this review a little later by checking out the power consumption in different configurations using Qoitech Otii Arc power supply.

Share this:

Support CNX Software! Donate via PayPal or cryptocurrencies, become a Patron on Patreon, or buy review samples

Subscribe
Notify of
guest
The comment form collects your name, email and content to allow us keep track of the comments placed on the website. Please read and accept our website Terms and Privacy Policy to post a comment.
19 Comments
oldest
newest
tkaiser
tkaiser
1 month ago

Thank you! So single-threaded performance at 600 MHz is identical or even slightly better than original Zero (ARM11 at 1000 MHz). Multi-threaded performance at least 4 times better.

Also interesting the DVFS opp at the different clockspeeds (slight reduction to what they used with RPi 3B/3B+):

tkaiser
tkaiser
1 month ago

Just realised you did upgrade both OS while testing: you started with Thread X from Apr 30 2021 and kernel 5.10.17 and were later on ThreadX from Sep 30 2021 and kernel 5.10.63.

Probably important settings changed when upgrading ThreadX: dvfs=2 -> dvfs=3, desired_osc_freq=0x325aa0 -> desired_osc_freq=0x331df0

over_voltage_avs has been adjusted automagically by ThreadX based on arm_freq. I guess there’s some room for improvements / experiments to bring temperatures / consumption further down. At least it’s not an ARM11 fed with 1.35V any more…

tkaiser
tkaiser
1 month ago

It shouldn’t matter but most probably with most recent settings the CPU cores at 1000 MHz run at a slightly lower voltage than 1.2125V and as such consume a little less.

Would also be interesting whether disabling CPU cores really results in lower max consumption. 2016 when I did some tests with RPi 3B this was not the case (ThreadX being the boss and doing whatever the RTOS wants 😉 )

bezeria
bezeria
1 month ago

Thank you for this review. Can we have a benchmark comparison of rpi zero2 against the radxa zero?

tkaiser
tkaiser
1 month ago

> Radxa Zero. It should be at least twice as fast

IMO more like 70% to 80% faster (both single-thread and multi-threaded). And the Radxa board should consume a little less power regardless of idle or full load.

tkaiser
tkaiser
1 month ago

In fact someone just now sent a PR with Radxa Zero results: https://github.com/ThomasKaiser/sbc-bench/pull/23

While memory (and AES of course) performs much better, integer performance isn’t that great (just a bit above Zero 2 level) but the testing was affected my massive swapping.

Also I wonder why cpufreq is limited to 1.8 GHz…

Salvador
Salvador
1 month ago

Tony told me than the amlogic blob clearly reduce performance when loading all cores on the radxa zero. That’s common on amlogic socs I guess.

tkaiser
tkaiser
1 month ago

Not sure why he thinks so since the 1.8GHz limitation is a DT setting (that’s why sbc-bench is walking through all cpufreq OPP).

When you look in results list at Radxa Zero and then above you can compare 3 boring SoCs all at 1.8GHz: S905Y2, RK3566 and H6. They all perform more or less the same (though the A55s in RK3566 accessing RAM much faster). BTW: If you would ‘overclock’ the ‘Renegade’ entry below to 1.8 GHz then RK3328 would also look the same.

tkaiser
tkaiser
1 month ago

The 7-zip results should be taken with a huge grain of salt since due to low memory 7-zip’s internal benchmark did something different on Zero 2 compared to devices with 1GB RAM or more: only 2 dictionaries were tested on the Zero 2 compared to 4 on devices with more RAM.

It’s a bit unlikely that Zero 2 at 1200 MHz outperforms the 3B+ at 1400 MHz… but Zero 2 shows slightly faster memory access (both bandwidth and latency).

Theguyuk
Theguyuk
1 month ago

On YouTube a tester used a much bigger cooler and did some retro gaming on this device, for those interested in such gaming. The heatsink used covered some of the gpio holes.

and.elf
and.elf
1 month ago

I wish I’d waited for this instead of buying an rpi3 a few months ago, since I only use it for libreelec I’d saved a few bucks 🙁

Anonymous
Anonymous
1 month ago

I just enhanced my LibreELEC Pi4 by connecting to a Wi-Fi extender with the ethernet port.

Sometimes it’s nice to have real ports.

Eric
Eric
1 month ago

The orange pi one held awesome value for a very long run, too bad it didn’t have the same love and support.

David Willmore
David Willmore
1 month ago

There is Armbian and the linux-sunxi folks doing a lot of work to support the allwinner Opi boards.

Advertisement