5 Head-Turning Features Of Intel's Movidius Neural Compute Stick

Compute Stick For Deep Learning

Intel subsidiary Movidius has launched a new small form factor device, the Movidius Neural Compute Stick, which is designed to make artificial intelligence a plug-and-play application.

The compute stick will help partners double down on deep learning and artificial intelligence as these applications gain traction in the market.

"The Myriad 2 VPU, housed inside the Movidius Neural Compute Stick provides powerful, yet efficient performance – more than 100 gigaflops of performance within a 1W power envelope – to run real-time deep neural networks directly from the device," said Remi El-Ouazzane, vice president and general manager of Movidius, in a statement. "This enables a wide range of AI applications to be deployed offline."

Here are five features of the Intel Movidius Neural Compute Stick.

Applications

Intel's newest compute stick acts as an artificial intelligence accelerator that doesn't need to be hooked into a cloud-based system, making it easier to add deep learning capabilities to existing platforms.

The compute stick has several functions. For instance, it can automatically convert trained Caffe-based convolutional neural networks into embedded neural networks that are optimized to run on Myriad 2 VPU. The compute stick also enables tuning for optimal real-world performance at ultra-low power, while validation scripts allow developers to compare the accuracy of the optimized model on the device to the original PC-based model, according to Intel.

Finally, Intel said its device can behave as a discrete neural network accelerator by adding dedicated deep-learning inference capabilities to existing computing platforms for improved performance and power efficiency.

Myriad 2 VPU

The Movidius Neural Compute Stick is powered by Myriad 2 vision processing unit technology, a low-power processor designed to handle specific tasks like eye and gesture tracking or environment mapping. The VPU, which uses 12 parallel cores to run vision algorithms, helps the Neural Compute Stick understand controllers' hand and arm gestures from a distance or recognize faces and analyze behavior.

Other companies are using Movidius' Myriad 2 VPU – including drone manufacturer DJI and video surveillance product manufacturer Dahua Technology.

Specs

The Movidius Compute Stick supports the Caffe deep-learning framework and connects to PCs through a USB 3.0. Consumers will need a 64-bit PC that runs Ubuntu – and has 1GB of RAM and 4GB of onboard memory – to work.

The compute stick consumes 1 watt of power and features up to 100 gigaflops of performance.

Pricing And Availability

Movidius Neural Compute Stick is available for purchase through select distributors for a market-suggested retail price of $79. Intel is working with an array of partners to distribute its newest compute stick, including RS Components and Mouser.

Intel AI Efforts

Intel has made strides over the past year to tighten its focus around artificial intelligence through building out its stack of hardware and software offerings. Months after acquiring Nervana Systems in August, the company in November launched the Nervana platform, which includes comprehensive "deep learning" solutions.

At the same time, the company revealed a product roadmap, including the first silicon optimized for neural networks to deliver high performance for deep learning, as well as compute density at a high-bandwidth interconnect. This silicon, code-named "Lake Crest," has been tested in the first half of 2017 and will be available to customers later in the year. Intel also introduced a new product, code-named "Knights Crest," which will tightly integrate its Intel Xeon processors with technology from Nervana Systems.