Thanh Le / bench notes
EMBEDDED · BOARD HUB · IN PROGRESS

Intel Edison: an x86 module the size of a stamp

Two Atom cores at 500 MHz on a Mini Breakout: its boot measured from the USB side, then Zephyr run from RAM

Last change2026-11-22: hub page and two STAR write-ups

Intel Edison

The Edison boot, measured on my board: reset to login in 59.52 s

The Edison is Intel’s 2014 IoT module: a Merrifield/Tangier SoC with two Atom (Silvermont) cores at 500 MHz, 1 GB of RAM, 4 GB of eMMC and Broadcom Wi-Fi, on a module smaller than a matchbox. Mine sits on the Mini Breakout, which gives it an FTDI console (J3) and a USB OTG port (J16). It still runs ubilinux (Debian 7.8), kernel 3.10.17 and U-Boot 2014.04.

Why look at it in 2026? It’s x86, and every other board on my shelf is ARM. Its kernel has no device tree; it describes the hardware with SFI tables and PCI instead, which I’d never dealt with. And Intel dropped it years ago, so whatever I want on it now, I do myself.

An Intel Edison Mini Breakout with numbered parts A Mini Breakout like mine (not my board): FTDI console on J3, USB OTG on J16, the 70-pin module connector J7. Photo: Mwilde2, CC BY-SA 4.0, Wikimedia Commons; numbers added by me.

What I got it to do

I started with read-only probes on the console: two Atom cores at 500 MHz, 983 516 kB of RAM for Linux, 11 SFI tables and 36 PCI functions. One timed reboot took 59.52 s from reset to login, and 24.16 s of that was a DHCP wait. The boot ROM prints nothing, so I timed it from the USB side instead, and that gave me a first answer to what happens when the eMMC is empty. Then I built Zephyr for an out-of-tree “edison” board and ran it from RAM through the old U-Boot.

The write-ups

Each one is Situation / Task / Action / Result, with the numbers I measured on my module.

#StoryResultStatus
1 Timing the Edison boot ROM from the outside, 0.1 s at a time
The ROM prints nothing for 13 s. Polling lsusb during one reboot showed what it does in that time
Reset to login 59.52 s; the ROM waits 8.92 s on USB for a flasher at every boot Done
2 Zephyr on the Edison: a 15 KB image over YMODEM and one wrong reg-shift
An out-of-tree Zephyr board for an x86 module, loaded into RAM by the old U-Boot, nothing written to eMMC
Hello World 0.05 s after go; thread period 0.6038 s vs 0.6039 s expected Done

Still open

Zephyr only uses one core and has no shell yet (UART receive isn’t wired up). I haven’t tried an empty or broken eMMC for real, or built the xFSTK recovery tool, so the recovery path is still on paper.