Wandboard 8MPLUS SBC Gets NXP i.MX 8M Plus SoC with 2.3TOPS NPU

TechNexion Wandboard Dual was one of the first Freescale i.MX 6 single board computers when it launched in early 2013 and the company has introduced a few more models over the years including Wandboard QuadPlus in 2017.

TechNexion is now working on a new version called Wandboard 8MPLUS powered by the recently announced NXP i.MX 8M Plus SoC that features a built-in 2.3 TOPS NPU for AI acceleration.

Wandboard 8MPLUS preliminary specifications:

  • EDM System-on-Module
    • SoC – NXP i.MX8M Plus quad-core Arm Cortex-A53 processor @ up to 1.8 GHz, Arm Cortex-M7 real-time core @ up to 800 MHz, Vivante GC7000UL 3D GPU, Vivante GC520L 2D GPU, HiFi 4 audio DSP, and 2.3 TOPS Neural Processing Unit (NPU)
    • System Memory – 2GB or 4GB LPDDR4
    • Storage – 32GB eMMC flash
    • Video Decode – HD (1080p +720p)
    • Video Encode – 1080p60 H.265, H.264
    • 260-pin EDM Type G board-to-board connector
    • Debugging – JTAG Interface by through holes
    • PMIC – NXP PCA9450
  • Carrier Board
    • Storage – MicroSD slot
    • Display Interfaces
      • HDMI 2.0a Tx with eARC support
      • Dual-channel LVDS expansion header, I2C touch connector
    • Audio – CTIA compliant 3.5mm audio jack, speaker connectors
    • Camera – Dual 4-lane MIPI CSI
    • Connectivity
      • Gigabit Ethernet (RJ45) via Atheros AR8035 Gigabit transceiver
      • IMX8M-Plus-4G model only
        • Dual-band 802.11 a/b/g/n/ac Wi-Fi 5 and Bluetooth via Qualcomm Atheros QCA9377
        • MHF4 connector for antenna
      • Optional 4G LTE via M.2 card and Nano-SIM card slot
    • USB – 3x USB 3.0 Host ports, 1x USB Type-C port
    • Serial – 2x UART
    • Expansion
      • M.2 slot
      • Expansion header with I²S, SDIO, CAN, UART, SPI, I²C, PWM, GPIO, etc..
    • Misc – Boot Select & Reset buttons, power LED
  • Power Supply – Via DC jack or USB-C port
  • Temperature Range – Commercial: 0° to +50° C
  • Certification and Compliance
    • USA – FCC ID: 2AKZA-QCA9377
    • Canada – IC: 22364-QCA9377
    • Japan – TELEC: 201-180629
    • European Union – EN 55032 / EN 55024; EN 300 328 v2.1.1; EN 301 893 v2.1.1
    • Australia/New Zealand – RCM
    • Certification – Compliant with RoHS / REACH directives

The company will officially support Android and Yocto Linux for the board, but if history is any guide, many other operating systems may eventually be supported as was the case for Wandboard.

The highlight of the board is really the NPU inside NXP i.MX 8M Plus processor, so just after providing an overview of the board, Marcel vandenHeuvel, TechNexion CEO, showcases an AI demo with Tensorflow Lite where object recognition is done in around 8ms when the NPU is enabled or about twenty times faster than with the CPU alone.


The demo runs on a full-featured i.MX 8M Plus development board as Wandboard 8MPLUS SBC is quite not quite available yet.

But TechNexion is now taking pre-orders for two variants of the board:

  • Wandboard IMX8M-Plus-2G with 2GB LPDDR4 for $134.00
  • Wandboard IMX8M-Plus-4G with 4GB LPDDR4, Wi-Fi and Bluetooth connectivity for $159

There’s no clear ETA, and the company promises to ship the board as soon as NXP i.MX 8M Plus processor is mass-produced. The latter often happens roughly one year after the announcement, so around January 2021. Bear in mind that processor launch timing is highly variable, so it could be a few months earlier or later.

Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.

Support CNX Software - Donate via PayPal, become a Patron on Patreon, or buy review samples
Subscribe
Notify of
guest
11 Comments
oldest
newest most voted
Willy
Willy
1 month ago

It’s a bit sad that 3 years after the much more power efficient A55 was introduced, some major vendors like NXP insist on selling the 8-years old A53, while lower end vendors like Amlogic and Rockchip have already upgraded their products.

Willy
Willy
1 month ago

For sure, but given that NXP is still creating new devices, it’s surprising they hadn’t started on these ones yet. Or maybe they consider that armv8.2+ will be in the imx9 family and as such will appear much later. But focusing on power-efficient AI by integrating NPUs and keeping the less efficient older CPU is a bit contradictory.

itchy n scratchy
itchy n scratchy
1 month ago

AML and RK won’t be held responsible if their products are not reliable, as they are targeting consumer not industrial, it was already a heavy birth for 64bit with the transition from i.mx6 to i.mx8…

crashoverride
crashoverride
1 month ago

I was actually “holding my breath” for this SoC. However, as with much that is pre-announced, by the time it hits market, I have lost interest. It needed to be 6 months sooner (its still not official) and 1/3 the cost.

Tim
Tim
1 month ago

What was your target?

theguyuk
theguyuk
1 month ago

Bigger, but Nvidia Jetson nano is around those prices.

dear_leader
dear_leader
1 month ago

Looks pretty fake to me, there are no bounding boxes and you have not showed any tf lite initiation on the board. the inferencing is not clearly real, because i cannot see any bounding boxes on the objects

Justice18660312
1 month ago

Seriously? It’s standard tensorflow lite test app:
https://www.tensorflow.org/lite/performance/gpu

Why inference must need bounding box? Welcome to join tensorflow community, you can learn more 🙂

engineer
engineer
1 month ago

Inference doesn’t require bounding box. the video shows image classification, but he says object detection (05:06), maybe that’s why other guy posted about bounding box. And also as per IMX 8mplus, it needs to run on NPU. and no information given on inference tools or inference engine
refer this link for more information
https://analyticsindiamag.com/what-is-the-difference-between-image-classification-object-detection-techniques/

SUSMITHA
SUSMITHA
1 month ago

The video doesn’t highlight much importance on AI Inferencing, its more like describing the hardware. I believe that 8m plus board is specialized for eIQ machine learning inferencing. You could have showed some demo and compare results with and without inferencing, show us the demo video of how you did it. That could have made more sense. the results appeared to be more fake for me.

Advertisements