All News | Boards | Chips | Devices | Software | Archive | About | Contact | Subscribe
Follow LinuxGizmos:
Twitter Facebook Pinterest RSS feed
*   get email updates   *

Snapdragon 820E mini-PC supports AI on the retail edge

Oct 3, 2018 — by Eric Brown 1,603 views

VIA unveiled a “VIA ALTA DS 3 Edge AI” mini-PC that runs Android 8.0 on a Snapdragon 820E. VIA previously announced it is bringing FogHorn’s Lightning IoT edge intelligence platform to its Edge AI systems.

VIA Technologies, which earlier this year released a VIA Smart Recognition Platform board powered by a Qualcomm Snapdragon 820 has now released an Android-based mini-PC built around the long-life Snapdragon 820E model. The $399 VIA ALTA DS 3 Edge AI follows other Alta DS systems including last year’s ALTA DS 4K mini-PC, which runs Android on a quad -A17 Zhaoxin ZX-2000 SoC and an earlier Alta DS 2 that runs Android on VIA’s own dual -A9 VIA Elite E1000 SoC.

Last month, VIA announced a partnership in which it will bundle FogHorn Systems’ Android- and Linux-compatible Lightning product portfolio for edge intelligence on VIA Edge AI systems. Adlink has a similar deal with FogHorn (see farther below).

VIA ALTA DS 3 Edge AI, front and back
(click images to enlarge)

The VIA ALTA DS 3 Edge AI provides an Android 8.0 “Oreo” BSP for the Snapdragon 820E, which also powers the Arrow/Qualcomm DragonBoard 820c SBC. The Snapdragon 820E is an almost identical, embedded variant of the Snapdragon 820, joining other 10-year lifespan embedded Snapdragon variants such as the Snapdragon 600E. Like the 820, it offers four Kryo cores — two at 2.35GHz, and two at 1.6GHz — plus a 624MHz Adreno 530 GPU, a Hexagon 680 DSP, and a 14-bit Spectra ISP.

ALTA DS 3 (left) and Snapdragon 820 (and 820E) block diagram
(click images to enlarge)

Like the VIA Smart Recognition Platform, the Alta DS 3 taps Qualcomm’s Neural Processing SDK for running “one or more neural network models trained in Caffe/Caffe2, ONNX, or TensorFlow using the CPU, GPU or DSP,” says VIA. The system is designed primarily for “multimedia-rich New Retail applications that boost customer convenience and engagement,” including applications that combine AI algorithms with “immersive multimedia signage display content.” Other applications include “facial recognition check-in systems for hotels, train stations, and airports or payment authentication systems for cashierless stores, vending machines, and ticket kiosks.”

ALTA DS 3 on vertical stand

The 175 x 118 x 25mm mini-PC is loaded with 4GB LPDDR4 in a POP package, as well as 16GB eMMC and an SD slot. There’s also an M.2 slot for 2280 NVMe SSDs.


The ALTA DS 3 is further equipped with a GbE port, a mini-USB port with “fastboot mode only”, and 3x USB 3.0 ports. Dual independent or mirrored 4K displays are available via the dual HDMI ports, which are described as offering an “integrated HDMI 2.0 transmitter and dual DSI to HDMI 1.4 converters.”

ALTA DS 3 detail views
(click image to enlarge)

The system provides a USB-only miniPCIe slot, a SIM card slot, and an antenna mounting hole. The U.S. version offers an LGA module with 802.11a/b/g/n/ac and Bluetooth 4.1 supported by two additional antenna connectors. You can swap this out for a VIA EMIO-2573 4G LTE mini-PCIe module, which also appears to be an option on the international model.

You also get dual audio jacks (WCD9335 codec), a 12V DC jack, a power button with LED, a TPM 1.2 security chip, and a Kensington lock. The 0.61 kg system supports 0 to 40°C temperatures, and there’s a vertical stand holder and optional VESA. Optional services include hardware and software customization and “cloud-based device management,” which may refer to the new FogHorn Lightning platform.

FogHorn’s Lightning

On Sep. 4, VIA announced a partnership with FogHorn Systems to offer its Lightning edge intelligence software on VIA’s Edge AI systems. The software supports industrial IoT and edge computing applications, helping to “accelerate the deployment of edge-to-cloud solutions for a myriad of industrial, transportation, and smart city applications,” says VIA.

FogHorn Lightning Edge ML
(click image to enlarge)

VIA did not mention which of Lightning products it will offer. The portfolio includes a new Lightning EdgeML release that leverages existing machine learning algorithms to bring AI capabilities to edge devices. EdgeML “enables complex machine learning models to run on highly-constrained compute devices such as PLCs, Raspberry Pi systems and tiny ruggedized IIoT gateways,” says FogHorn. Other Lightning packages provide data visualization, pre-processing, analytics, and more.

In addition to supporting the ALTA DS series, Lightning will be available with the previously mentioned VIA Smart Facial Recognition, as well as VIA Mobile 360 ADAS and VIA Mobile360 Surround View Computer Vision. It will also run on VIA AMOS and Artigo embedded systems that lack SoCs such as the Snapdragon 820E with special AI acceleration features. We don’t mean to single out VIA here for broadening the AI hype beyond credulity — just as most embedded systems launched in 2016 were called IoT platforms, in 2018 they’re called AI edge intelligence systems and the like.

We first saw FogHorn mentioned in July when Adlink announced it was providing FogHorn Lightning as part of its new Adlink DXS IoT digital experiments as-a-service offering. This soup-to-nuts IoT cloud analytics service is further explained in this Digitimes story.

(Update: A few days after this story posted, Foghorn released Lightning 2.0. The cross-platform software has improved scalability, automated processes, sensor fusion, and “edgification” of machine learning models, as well as an updated UI.)

Further information

The VIA ALTA DS 3 Edge AI is “sold out due to unexpected demand” at the VIA store, but pre-orders are available for $399. More information may be found in VIA Technologies’ ALTA DS 3 announcement and product page.

(advertise here)

Print Friendly, PDF & Email

One response to “Snapdragon 820E mini-PC supports AI on the retail edge”

  1. Shane says:

    Okay why not buy a usbc hub, and then buy a Pixel smartphone. You can still find them around the place, plus it has a faster SD 821 SoC. If you don’t mind spending a little more, then you can pick up a Pixel 2 which has an even faster SD 835 SoC. Plug the USBC hub into the Pixel device, and you have a device better than this one.

Please comment here...