SiFive and CEVA work together on machine learning processors

January 08, 2020 //By Ally Winning
SiFive and CEVA will partner to design ultra-low-power domain-specific Edge AI processors for a variety of high-volume use cases.
SiFive and CEVA will partner to design ultra-low-power domain-specific Edge AI processors for a variety of high-volume use cases.

The partnership is part of SiFive’s DesignShare program and will be centred on RISC-V CPUs, CEVA’s DSP cores, AI processors and software. These features will be integrated into SoCs that are intended for a wide variety of end markets. The first use cases considered will include smart home, automotive, robotics, security and surveillance, augmented reality, industrial and IoT.

The two companies will develop the industry-specific, scalable SoCs to meet the challenges of maximizing performance, extending battery life and adding intelligent features to devices not suitable for cloud-based AI inference due to security, privacy and latency concerns.

The Edge AI SoCs will be supported by CEVA’s CDNN Deep Neural Network machine learning software compiler to create optimized runtime software. CDNN has a wide range of network optimizations, advanced quantization algorithms, data flow management and fully-optimized compute CNN and RNN libraries that allow cloud-trained AI models to be deployed on edge devices for inference processing. CEVA will also supply a full development platform that is based on the CEVA-XM and NeuPro architectures for the development of deep learning applications using the CDNN, as well as DSP tools and libraries for audio and voice pre- and post-processing workloads.

More information

www.sifive.com/designshare

Related news

Rutronik UK adds Redpine multiprotocol wireless SoCs and modules

Smart Energy 1.4 chosen for UK smart meter roll-out

Qorvo: More Power to Zigbee 3.0: Wi-Fi, Zigbee and Thread ‘unmeshed’​

nRF52840 multiprotocol SoC now also supports Zigbee


Vous êtes certain ?

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

Vous allez être rediriger vers Google.