Close Up Photos of UDOO Bolt Ryzen Embedded V1000 Development Board

The first AMD Ryzen Embedded V1000 development board – UDOO Bolt  – was  launched on Kickstarter last week, but for whatever reasons, the company did not provide close up photos of the board.

But luckily for us, TLS – regular reader & commenter, and occasional contributor for CNX Software – is now at Computex 2018 and spent some time at UDOO’s booth where he took time to take some photos and share with us.

UDOO Bolt Ryzen Embedded V1000 Development Board Top
Click to Enlarge

The top view of the board shows the two SO-DIMM slots and Ryzen Embedded processor, as well as an better looks at the ports and headers around the boards.

UDOO-Bolt-Sideview
Click to Enlarge

The board should support up to 4 displays via its two HDMI 2.0a ports, and two USB type C ports with DP alternate mode. While, AMD Ryzen Embedded V1000 SoCs support 10 GbE, UDOO selected TRXCOM TRJG19901DNL Gigabit Ethernet jack, probably to keep costs under control.

UDOO-Bolt-Board-Bottom
Click to Enlarge

The bottom side reveals the M.2 Key B 2260 socket for SATA SSD and M.2 Key E 2230 socket on the top left, and M.2 Key M 2280 slot for NVMe SSDs on the right side.  Sadly the component names are a little hard to read, but the 32GB eMMC flash should be found right under the two M.2 slots, and the Microchip ATMEGA32U4 for the Arduino part on the top right just under the unpopulated embedded controller I/O header.

Support CNX Software - Donate via PayPal or become a Patron on Patreon

7
Leave a Reply

avatar
4 Comment threads
3 Thread replies
0 Followers
 
Most reacted comment
Hottest comment thread
5 Comment authors
onebirMark L JacksonJohann McDoePaul MTLS Recent comment authors
  Subscribe  
newest oldest most voted
Notify of
TLS
Guest
TLS

Some additional bits. The next revision will be made in a couple of weeks. Udoo has to add a third party controller for the GPIO, as apparently the AMD Ryzen SoC doesn’t have any usable native GPIO’s.

Paul M
Guest
Paul M

Taking a look at wikipedia, this board has the same footprint as Nano-ITX
comment image

so it should be possible to find a case, but the ports will be in a different position.

TLS
Guest
TLS

It’s actually the same size as mini-STX as well, but the screw holes are different. It does seem like the screw holes line up with the Nano-ITX standard, so nice catch. Port layout should be a minor issue, but most Nano-ITX boards don’t seem to have ports on both sides.

Johann McDoe
Guest
Johann McDoe

At least two bad design decisions: soldered BIOS/UEFI flash instead of removable package and no status LED on the Ethernet port.

TLS
Guest
TLS

This is revision 0.1, so a lot of things might still change.

Mark L Jackson
Guest
Mark L Jackson

Is this the Ryzen/Vega CPU/GPU SoC ? This is what I need, lots of compute capability to run a real-time physics engine as part of a next gen hybrid powerplant ECU, controlling IC using physics models rather than silly maps, Maps were just a hack needed back in the early 80’s implementing fuel injection, shit we got an extra 10+ zeros on compute now. way overdue to be rid of them, calculating a in-cylinder trapped mass and analysing cylinder pressure to identify combustion phasing is easily done. Using artificial neural networks to match the efficiency maps of Li-Ion, graphene super-capacitor, electric motor and IC operating three different combustion modes to achieve best power or economy. Operating Li-Ion cells at high load to keep mass down on a motorcycle, just as hard if not harder than IC operation, degradation models, current/voltage/state of charge models, temperature/current models … Cost per watt/hour in $ / co2 reduction, maintaining IC above 40% thermal efficiency.

onebir
Guest
onebir

Yes, this can do 3.6 FP16 Tflops:
https://www.amd.com/en/products/embedded-ryzen-v1000-series

But
1) this is enough to train decent-sized NNs (assuming enough memory and reasonable memory bandwidth). Maybe you could train your NN offline & use a much less powerful GPU (eg on an ARM board) to do inference?
2) off the shelf options for AMD GPGPU usage are limited at present (just PlaidML/Keras afaik)