The new Compute Module 5 (“CM5”), like the commercially available Raspberry Pi 5, has a significant performance boost.
Therefore, we tried benchmark tests with Compute Module 4 (“CM4”) and CM5.
We used Geenbench and PiBenchmarks.
The PL-R4 and PL-R5M were used for the trial.
The operating system for the “PL-R4” with CM4 is bullseye and the “PL-R5M” with CM5 is bookworm.
Each is activated by eMMC.
Geekbench
Geekbench, the famous benchmark test, can be run on the Raspberry Pi. However, only the preview version, which is treated as a beta version, is supported.
This is because the CPU is an Arm.

Geekbench preview version: https://www.geekbench.com/preview/
For Linux, I found that it is compatible with Ubuntu 18.04 or later. It worked without trouble on the Raspberry Pi OS, probably because it is also derived from Debian.
The version at the time of writing is Geekbench-6.4.0-LinuxARMPreview.
execution method
The preview version of Geekbench works just by running it with a command after downloading. It’s easy.
Download
wget https://cdn.geekbench.com/Geekbench-6.4.0-LinuxARMPreview.tar.gz
tar -xf Geekbench-6.4.0-LinuxARMPreview.tar.gz
cd Geekbench-6.4.0-LinuxARMPreview
execution (e.g. program)
./geekbench6
The measured results are automatically uploaded to Geekbench and the URL link of the measured results is output to the terminal screen. (e.g. https://browser.geekbench.com/v6/cpu/<id>)
If you are new to Geekbench, you can create an account on the page you visit from the URL. We recommend that you create an account because once you do, you will be able to save all of your results in one place and view comparisons.
Execution time will take a fair amount of time on Raspi. Wait a little.
Geekbench results
Geekbench is expressed as a score. You can also see specific measurements as details.
As a result, the CM5 score seemed incomparably different from the CM4.
It is an order of magnitude difference. I am a little surprised.
This is a performance difference that can be felt even at the speed of experience.

CM5 wins hands down in both single-core and multi-core. In fact, the difference in score is 3.4 times.

URL of the results of this test
The measured results also show detailed specifications. For example, file compression is 126.6 MB/sec.
There is a function that displays the two graphs so that they can be compared.
The URLs of the compared graphs are as follows
Comparison graph based on CM5
: https://browser.geekbench.com/v6/cpu/compare/10676765?baseline=10638221
Comparison graph based on CM4
: https://browser.geekbench.com/v6/cpu/compare/10638221?baseline=10676765
See where it has significantly increased.
PiBenchmarks
Next I measured it with PiBenchmarks.
It is similar to Geekbench. This one is also expressed as a score.
I have been using Raspida to measure NVMe drives for some time. (Reference: https://raspida.com/pi5-nvme-ssd-like-a-pc/ )
If the OS is running on eMMC, the following command will do the job.
sudo curl https://raw.githubusercontent.com/TheRemote/PiBenchmarks/master/Storage.sh | sudo bash
You can leave it as is this time.
If you want to measure a microSD card with eMMC activated, you must specify the target device.
Optionally, specify the path to the device.
findmntcommand, it is convenient to display both the source path and the mount location.
In this example, /dev/mmcblk0 is the eMMC and /dev/mmcblk2 is the microSD card, but you can also see where they are mounted.
├─/boot/firmware /dev/mmcblk0p1 vfat rw,relatime,fmask=0022,dmask=0022,codepage=437,iocharset=ascii
└─/media/pi/SDCARD /dev/mmcblk2p1 vfat rw,nosuid,nodev,relatime,uid=1000,gid=1000,fmask=0022,dmask=00
Measurements for another drive are downloaded and then run.
With the industrial raspberry pi used in this project, another drive (e.g. microSD card) can be used, so if you want to measure other than bootfs, you can download the sh file once and specify the path to it.
wget https://raw.githubusercontent.com/TheRemote/PiBenchmarks/master/Storage.sh
chmod +x Storage.sh
Just run the sh file with the path at the download location.
sudo ./Storage.sh /media/pi/SDCARD
In this case, I wanted to know the score of eMMC, the bootfs, so I ran it without specifying the path.
If you want to measure a drive other than the one you are running from, remember to download it once and specify the path.
CM4 eMMC read/write speed
Like Geekbench, the PiBenchmarks website also allows you to see a list of your own measurements and compare them with other test data. (Comparison tables are not available.)
The web can be used to filter by Raspberry Pi model, such as which manufacturer’s drives are most commonly used, and as a reference for purchasing NVMe-connected SSD drives.
As before, let’s compare it to CM5 based on CM4.
This is the score of CM4. The image is what is displayed in the terminal.

CM4 eMMC results URL: https://pibenchmarks.com/benchmark/104107/
Other test data showed that scores were about the same, around 6,000.
CM5 eMMC read/write speed
How about the new CM5? This is what I mean by an order of magnitude difference.

CM5 eMMC results URL: https://pibenchmarks.com/benchmark/104123/
I looked at the scores uploaded by others and they were about the same, around 13,100.
Some of the overclocked measurements showed scores in the 14400 range.

Not to mention the score, not to mention the eMMC of the CM4, which outperforms the CM4 in all read/write categories.
After all, we can see that the performance has improved a lot with CM5.
Other measurement methods
There are other tool apps that can measure as well, such as GeekBench and PiBenchmarks.
This is also performed by specifying an eMMC or microSD card.
hdparm command
You may simply use the hdparm command to measure only the read speed.
For accuracy, execute the same command four or five times and take an average.
For read speed, it is the -t option.
hdparm -t /dev/mmcblk0
gnome-disks for GUI
A typical example that can be measured with GUI is gnome-disks.

You can use it by installing gnome-disk-utility.
It worked on the Raspberry Pi.
sudo apt install gnome-disk-utility
The settings may be left as default.
CM5 read/write speeds have roughly doubled?
Both Geekbench and PiBenchmarks were similar in their usability.
First, try Geekbench, which is well-known for its bar graphs.
This time, considering the Geekbench measurements, the read speed was approximately 80 MB/s for CM4 and 309 MB/s for CM5, which is really more than 3.8 times faster.
The result is also an increase in write speed of approximately 1.7 times.
In total, as the scores indicate, CM4 scores in the 6,000 range and CM5 scores in the 13,000 range, suggesting that read/write speeds have more than doubled in performance with CM5.
Since there is nothing better than fast processing, it seems that CM5 will be the one to introduce Raspberry Pi for industrial use in the future.
It is important to note that CM5 is only available from bookworm, so if you want to use modules or applications that are faulty depending on the OS version, you still have to discard the CM4 option.
*The PL-R4 series of industrial Raspberry Pi is configured with CM4, but can be easily modified to be compatible with CM5.

Processing specs are not the only thing that matters, but projects that were previously too burdensome may work.
It is better to be as fast as possible, especially when you want to handle heavy data.
In the future, Raspberry Pi Compute Module 5 is also likely to be seen more often in the industrial world in Japan.
Article contributed by Raspida
Raspberry Pi information site that even non-engineers can enjoy using raspida.com a Raspberry Pi information site that even non-engineers can enjoy and handle. He also contributes technical blog articles to the PiLink site on the Raspberry Pi for industrial use.

