We are evaluating and working with multiple architectures and manufacturers at this moment. We are taking into account several factors including but not limited to:
- Security ( Trustzone, HAB support, Secure Memory )
- Logistics ( Market availability, Lead Time )
- Support ( Existing BSPs, Technical Support, Documentation )
- Robustness ( Temperature range, Power supply range, SBC/Carrier Format)
- Peripheral ( Ethernet connection, USB connection )
We have contacted each manufacturer, evaluated their support process, received technical documentation. In many cases, we have acquired the product to run internal tests.
As a reference we can provide a list of some of the devices we are currently working with:
|STM32MP157C-DK2🔗||STM32MP157C||ST Micro||✅||✅||✅||✅ |
|MCiMX8-evkb 🔗||i.MX8M Quad||NXP||✅||✅||✅||✅ |
|ODYSSEY STM32MP157C 🔗||STM32MP157C||Seeed||✅||✅||⛔ availability|
|phyCORE-STM32MP1 🔗||STM32MP157C||Phytec||✅||✅||⛔ Secure boot|
|byteDEVKIT 🔗||STM32MP157C||BytesAtWork||✅||✅||✅||⛔ TEE issues|
|emSBC-Argon 🔗||STM32MP157C||emtrion||✅||✅||⛔ availability|
|USB Armory Mk-II 🔗||i.MX6UL||F-secure||✅||✅||✅||⛔ OPTEE|
|WISE-710 🔗||i.MX6||Advantech||✅||✅||✅||⛔ BSP quality|
|phyBOARD-Segin 🔗||i.MX6||Phytec||✅||✅||⛔ availability|
|LEC-iMX8M 🔗||i.MX8 Quad||Adlink||⛔ No replies|
|phyBOARD-Polaris 🔗||i.MX8M Quad||Phytec||✅||✅||⛔ OPTEE|
|Apalis iMX8 QuadMax 🔗||i.MX8 Quad Max||Toradex||✅||✅||✅||⛔ TEE issues|
|SBC-iMX8M-Mini 🔗||i.MX8M Mini Quad||Compulab||✅||✅||✅||⛔ cost + bugs|
|PICO-PI-IMX8M 🔗||i.MX8M Mini Quad||TechNexion||✅||✅||✅||⛔ tech support|
|Verdin iMX8M 🔗||i.MX8M Mini Quad||Toradex||✅||✅||✅||⛔ availability|
[A] contacted manufacturer [B] we adquired a device [C] evaluated custom image [X] not available yet
Why some devices are prioritized?
After analyzing and testing a wide range of devices, we noticed a few common themes:
- BSP quality issues
- Inadequate form factor for datacenter applications: We believe ethernet support is a critical requirement.
- Market availability problems: devices are still not available, stocks were low and supply chain is not clear.
With respect to software layers, we had a look at different design alternatives. Taking a pragmatic long-term support perspective, we decided to use a extremely minimal Buildroot+OPTEE image for the initial implementation.
While there are interesting opportunities in the bare metal / RTOS space, we consider them too experimental. Moreover, we prefer to rely on projects with a substantial community to ensure long term maintenance. This means, relying on extensive testing, CVE availability, strong community, etc.