SiFive, a provider of commercial RISC-V processor IP, today announced the first open-source RISC-V-based SoC platform for edge inference applications based on NVIDIA’s Deep Learning Accelerator (NVDLA) technology.
The demo will be shown this week at the Hot Chips conference and consists of NVDLA running on an FPGA connected via ChipLink to SiFive’s HiFive Unleashed board powered by the Freedom U540, the world’s first Linux-capable RISC-V processor. The complete SiFive implementation is well suited for intelligence at the edge, where high-performance with improved power and area profiles are crucial. SiFive’s silicon design capabilities and innovative business model enables a simplified path to building custom silicon on the RISC-V architecture with NVDLA.
NVIDIA open-sourced its leading deep learning accelerator over a year ago to spark the creation of more AI silicon solutions. Open-source architectures such as NVDLA and RISC-V are essential building blocks of innovation for Big Data and AI solutions.
“It is great to see open-source collaborations, where leading technologies such as NVDLA can make the way for more custom silicon to enhance the applications that require inference engines and accelerators,” said Yunsup Lee, co-founder and CTO, SiFive. “This is exactly how companies can extend the reach of their platforms.”
“NVIDIA open sourced its NVDLA architecture to drive the adoption of AI,” said Deepu Talla, vice president and general manager of Autonomous Machines at NVIDIA. “Our collaboration with SiFive enables customized AI silicon solutions for emerging applications and markets where the combination of RISC-V and NVDLA will be very attractive.”