---
title: Movidius Neural Compute Stick 2
h1: Movidius Neural Compute Stick 2
description: Getting started with the Movidius Neural Compute
Stick 2 with the third generation Myriad X Vision Processing Unit.
author: JWRR
date: February 17, 2019
theme: a
...
[View Markdown](index.md)
The Intel Neural Compute Stick 2 plugs into a USB3 socket and is a development
platform for the Movidius Myriad X Vision Processing Unit (VPU). The Myriad
X's low power consumption and small package size make it a good fit for IoT,
cloud-edge, image processing applications. It is well suited for image
classification and inference. It is not intended for training which is best
performed in the cloud or on power-hunger, high-end graphics cards. Some
potential uses are in drones for obstacle avoidance, object recognition,
security surveillance systems, AR/VR headsets, smart homes and smart cameras.
The Myriad X is power efficient using just about 1.5 Watts, but it can still
process up to 4 trillion operations per second. It has 16 vector processors
optimized for image processing pipelines and computer vision workloads. It
has over 20 hardware accelerators to perform tasks such as H.264 decoding and
encoding, optical flow and stereo depth perception algorithms.
The Myriad X is available in two IC packages. The NCS2 uses the MA2485 which
has an embedded 4 Gbit LPDDR4 memory built into the package. The MA2085
doesn't have embedded memory but has an external LPDDR4 interface that supports
up to 16 Gbits. The memory interface is 32-bits wide and can operate at 1600
MHz for date rates up to 12.8 GBytes per second. Both packages are 8.1mm x
8.8mm.
*Static Strap?*
The Myriad X has 16
[MIPI](https://www.mipi.org/specifications/csi-2)
lanes and can be connected directly to eight HD
cameras. Note, the NCS2 does **NOT** bring out any MIPI lanes, so all video
and data is from the USB3 plug. I would love a Myriad X development board
with a MIPI connector.
{{ toc }}
## Order the NCS2 (order)
The Intel NCSM2485.DK Movidius Neural Compute Stick 2 is available for about
$100.00 from stores such as
[Newegg](https://www.newegg.com/Product/Product.aspx?Item=9SIA6ZP8K46736) and
[Amazon](https://www.amazon.com/Intel-NCSM2485-DK-Movidius-Compute-Processing/dp/B07KT6361R/).
Make sure to buy the latest, not the previous generation.
## Install the Software (install)
The NCS2 uses the open-source OpenVINO toolkit. The software is available for
Linux, embedded enviroments and Windows. Ubuntu 16.04 LTS appears to be the
most mature. I''ll be using the NCS2 on a Windows 10 laptop.
[OpenVINO Installation Instructions](https://software.intel.com/en-us/articles/OpenVINO-Install-Windows)
describe the multi-step process. Follow the instructions. The install wraps
up with two demos that can be run without the NCS2.
## Useful Links (links)
* [Intel - movidius-myriad-xvpu-product-brief.pdf](https://newsroom.intel.com/wp-content/uploads/sites/11/2017/08/movidius-myriad-xvpu-product-brief.pdf)
* [Berkeley Design Tech. - Next-generation Intel Movidius Vision Processor](https://www.bdti.com/InsideDSP/2017/10/31/Movidius)
* [Anandtech - Intel Announces Movidius Myriad X VPU](https://www.anandtech.com/show/11771/intel-announces-movidius-myriad-x-vpu)
* [Linux Gizmos - Second-gen Intel Neural Compute Stick shows off new Myriad X VPU](http://linuxgizmos.com/second-gen-intel-neural-compute-stick-shows-off-new-myriad-x-vpu/)
* [Robotics Review - Intel NCS 2 Designed to Accelerate AI Applications at the Edge](https://www.roboticsbusinessreview.com/ai/ncs-2-intel-accelerate-ai-apps-edge/)
## NCS 2 / Myriad X Videos (videos)
While waiting for the NCS2 to arrive, here are some Youtube videos.
### Myriad X: Technology and Chip Architecture
### Intel Neural Compute Stick 2: A new USB stick with Artificial Intelligence
Synthesized voice but interesting use cases
## NCS / Myriad 2 Previous Gen Videos (oldvideos)
### Myriad 2 before Intel acquired Movidius
### Intel Movidius Neural Compute Stick | IoT Developer Show
### Deep Learning with Intel
OpenVINO stands for Open Visual
Inference and Neural network Optimization. OVIANNO seems like a more accurate
acronym... and the URL isn't taken yet... but what do I know.