The Raspberry Pi Foundation just launched the Raspberry Pi Compute Module 3+ for $25 and up with Broadcom BCM2837B0 processor offering better thermals than BCM2837 processor. As is often the case, this post generated some insightful comments from the smart readers of CNX Software, and they pointed out some of the shortcomings of the module such as the lack of Ethernet support that would be required for cluster use cases for example. You can also add a USB to Ethernet chip to the carrier board, but that’s not ideal, and instead someone linked to Qihua CQH6 system-on-module powered by Allwinner H6 processor and sold for 158 RMB ($23.5 USD) or 199 RMB ($29.65) for respectively 512MB RAM/4GB flash and 1GB RAM/8GB flash configurations.
- SoC – Allwinner H6 quad-core Cortex A53 processor @ 1.8 GHz with Arm Mali-T720MP GPU
- System Memory – 512MB, 1GB or 2GB DDR3L RAM
- Storage – 4GB, 8GB, 16GB, or 32GB eMMC 5.1 flash
- Edge Connector – 204-pin SO-DIMM3 connector with USB 3.0, Gigabit Ethernet, HDMI 2.0a with 4K HDR support, (flawed) PCIe interfaces
- Power – AXP805 PMIC
- Dimensions – 6.7 x 3.4 cm
The module is said to support Linux with Qt 5.0, Ubuntu 18.04 and Android 7.0. The 1GB/8GB board is actually about the same price as than RPi 3+/8GB compute module but offers much better features including Gigabit Ethernet, USB 3.0, 4K video output, and PCIe interfaces.
Raspberry Pi Compute Module 3+ will have better documentation and software support, but it might be less important for such type of module, since it’s designed for companies making their own custom board, so they already have some know-how in house. I struggled to find documentation for the SoM until I discovered a thread on Armbian that points to some resources on Baidu with password 7u3q. Sadly the link does not work for me, and I don’t know if that’s a temporary issue, or the link is wrong. I don’t see any mention of a development kit either, So it’s still a little early to find out if Qihua CQH6 module has a bright future, but at least it has potential. Alternatively, there’s always Pine64 SOPINE64 system-on-module.