Neural AI processor in a USB stick offers 100 Gflops performance
Do you need to develop or add AI computing tasks such as vision processing, picture recognition, voice processing or any such streaming content processing which requires artificial intelligence processors to accelerate computing at the edge device, which does not have required processing engine to process complex neural computing algorithms? Now you have a neural computing hardware on a USB stick from Intel offering developers such artificial intelligence (AI) accelerator in handling computing loads which require vector processing. This AI silicon stick can be used to develop, tune and deploy AI applications. Its like adding a set of AI specific processor cores to your host computer processor .
Though you have cloud to process heavy loads of AI computing, but is not always the best solution. A computing platform at the edge in automotive, robot, or in a smart home environment is better option than relying on cloud computer, so that it ensures less dependency on networks with no hinderance in operation.
Intel Movidius neural computing stick can also enable design of some software running exclusively utilising this stick on some of the edge computers which are already in the field.
Though Intel has offered this product for developers, what is missing is availability of any popular operating system or software which is readily programmed to accept such hardware accelerators. In that sense this product is more for software developers rather than end-users who can readily used this product to get the AI at edge through plug and play. But there is a chance for some innovative integrator to bring out a software product which can exclusively work with this AI Computer dongle, provided Intel support this product in future.
Intel Movidius Neural Compute Stick can automatically convert a trained Caffe-based convolutional neural network (CNN) into an embedded neural network optimized to run on the onboard Movidius Myriad 2 VPU. Intel also provides other tools for AI.
The key features of this hardware accelerator includes:
1. The Myriad 2 VPU inside the Movidius Neural Compute Stick provides 100 gigaflops of performance within a 1W power envelope.
2. Support runing real-time deep neural networks directly from the device.
3. Perform inference at the edge .
4. Automatically convert a trained Caffe-based convolutional neural network (CNN) into an embedded neural network optimized to run on the onboard Movidius Myriad 2 VPU
5. Layer-by-layer performance metrics for both industry-standard and custom-designed neural networks enable effective tuning for optimal real-world performance at ultra-low power.
6. 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.
Price :US$79 sold by Mouser and RS
Picture: Mutiple Movidius Neural Compute Sticks conected to a notebook computer