BeagleY-AI SBC features TI AM67A vision processor with 4 TOPS AI accelerators

The BeagleBoard.org Foundation’s BeagleY-AI is an open-source hardware, credit card-sized SBC powered by a Texas Instruments AM67A quad-core Cortex-A53 vision processor with various programmable blocks capable of delivering up to 4 TOPS for AI algorithms.

The board ships with 4GB RAM, relies on a microSD card slot for storage and OS booting, implements gigabit Ethernet, WiFi 6, and Bluetooth 5.4 connectivity, and can drive up to three displays via micro HDMI, OLDI (LVDS), and MIPI DSI interfaces. The BeagleY-AI also comes with two MIPI CSI camera interfaces, four USB 3.0 ports, a USB Type-C port, and a 40-pin GPIO header for expansion. We can also see a 16-pin PCIe FPC connector that looks somewhat similar to the 20-pin PCIe connector on the Raspberry Pi 5 but officially supports PCIe Gen3 x1.

Texas Instruments AM67A single board computer

BeagleY-AI specifications:

  • SoC – Texas Instruments AM67A (J722S) “vision processor”
    • CPU
      • Quad-core 64-bit Arm Cortex-A53 processor at 1.4GHz
      • Arm Cortex-R5 subsystem for low-latency I/O and control
    • GPU and VPU – “GPU, video and vision accelerators, and other specialized processing capability”
    • DSP/AI accelerator – Dual general-purpose C7x DSP with Matrix Multiply Accelerator (MMA) capable of 4 TOPS
    • High-speed interfaces – PCI-Express Gen3 single-lane controller,  USB 3.1 Gen1 port, Gigabit Ethernet
  • System Memory –  4GB LPDDR4
  • Storage – MicroSD card socket, board identifier EEPROM (4Kbit)
  • Display I/F
    • 1x micro HDMI port
    • 40-pin OLDI (LVDS) connector with touchscreen support
    • 22-pin MIPI-DSI with touchscreen support (muxed with MIPI-CSI)
    • Up to 3x simultaneous displays
  • Camera I/F – 2x 22-pin MIPI CSI camera connectors
  • Networking
    • Gigabit Ethernet RJ45 jack with Power-over-Ethernet (PoE+) support via add-on
    • WiFi 6 and Bluetooth 5.4 via BeagleBoard.org BM3301 module based on TI CC3301
  • USB
    • 4x USB 3.0 (5Gbps) Type-A host ports
    • 1x USB 2.0 (480Mbps) Type-C device port and power input
  •  Expansion
    • 16-pin PCIe Gen3 x1 FPC connector
    • 40-pin expansion header
  • Debugging
    • 3-pin JST-SH console UART
    • 10-pin TAG-CONNECT JTAG header
  • Misc – Fan power and control connector
  • Power Supply – 5V via USB Type-C port
  • Dimensions –
  • OSHWA certification – US002616

BeagleY-AI SBC

BeagleY-AI block diagram
Block diagram

The board runs a Debian image with the XFCE desktop environment. Since the BeagleY-AI is open-source hardware you’ll find the (Altium) hardware design files on OpenBeagle and the documentation page is up, but currently empty, although a quick start guide (PDF) is already available. There’s currently no public datasheet or even a product page for the AM67A/J722S SoC. This can be easily explained as the Beagleboard.org Foundation was given early access to the upcoming vision processor and all I could find on the Texas Instruments website at the time of writing was an evaluation board for the AM67A and pin-compatible chips.

But there’s still time since the BeagleY-AI will only become available in June 2024 at a fairly good price tag of $70. You can already pre-order your board today on Element14 or Seeed Studio. It’s not the first SBC from BeagleBoard with an AI accelerator, as the company introduced the BeagleBone AI-64 in 2022 with an 8 TOPS accelerator and the BeagleV-Ahead powered by a TH1520 quad-core RISC-V processor that embeds a 4 TOPS NPU. More details may be found on the product page.

BeagleY-AI bottom side

Share this:
FacebookTwitterHacker NewsSlashdotRedditLinkedInPinterestFlipboardMeWeLineEmailShare

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress

ROCK Pi 4C Plus

4 Replies to “BeagleY-AI SBC features TI AM67A vision processor with 4 TOPS AI accelerators”

  1. “an open-source hardware, credit card-sized SBC powered by a Texas Instruments AM67A quad-core Cortex-A53 vision processor”

    Really? Is that true?? An open-source
    AM67A quad-core Cortex-A53 vision processor?

Leave a Reply

Your email address will not be published. Required fields are marked *

Khadas VIM4 SBC
Khadas VIM4 SBC