Syntiant processor ups neural performance 25x

January 13, 2021 // By Peter Clarke
Syntiant processor ups neural performance 25x
Syntiant Corp. (Irvine Calif.), an analog AI startup focused on audio interfaces, has launched its second-generation Neural Decision Processor, the NDP120 with 25 times greater tensor performance than its previous generation devices.

Like the previous generation NDP100/101 the NDP120 is optimised for audio and sensor processing for always-on applications in battery-powered devices.

The NDP120 applies neural processing to run multiple applications simultaneously, including echo-cancellation, beamforming, noise suppression, speech enhancement, speaker identification, keyword spotting, multiple wake words, event detection, and local commands recognition.

The chip includes the Syntiant Core 2, neural network inference engine, with a configurable audio front-end interface. The NDP120 also supports multi-modal sensor fusion, including sensors for infrared detection, multi-axis acceleration, tilt, magnetic field, and pressure.

The Syntiant Core 2 is a tensor processing core designed to support tensor processing and ease of programming. This second-generation architecture delivers 25x the tensor throughput of the Syntiant Core 1 found in the NDP100 and NDP101 devices that are currently shipping in high volumes (see AI audio startup ships one million ICs, raises funds ).

However, the company did not discuss any migration between manufacturing process nodes that might have allowed for a scaling-up of the neural networking core. In the past Syntiant has described its technology as being an analog neural network and has been described as performing processing in a flash memory array.

The Syntiant Core 2 training development kit includes tools for running bit-exact simulations directly within high-level modeling languages, such as Tensorflow and Keras. All major frameworks can port to the Syntiant Core 2 runtime, including native support for multiple types of convolutional kernels, kernel striding, kernel dilation, downsampling layers, fully connected layers, pointwise operations, and a variety of activations.

Compression is supported allowing network architectures to mix 1, 2, 4, and 8-bit weights and higher precision bias terms. For the most challenging tasks, the Syntiant Core 2 supports high precision modes, including 16-bit inputs and outputs. With design tools supplied by Syntiant, it is possible to explore the complete power, latency, memory, and parameter requirements on the Syntiant Core 2 interactively.

The NDP120 is equipped with an Arm

Vous êtes certain ?

Si vous désactivez les cookies, vous ne pouvez plus naviguer sur le site.

Vous allez être rediriger vers Google.