SiFive - April 23, 2021

What’s new in AI & ML from SiFive

Introducing the SiFive Intelligence X280

This week has been an exciting one for our team at SiFive, with a number of key announcements that are beginning to allow us to publicly share our strategy for the exponentially growing AI/ML market. While we are only scratching the surface of the impact AI/ML will have on our industry and our lives, it’s clear that the workloads, algorithms, and requirements are going to be constantly evolving and improving for the foreseeable future.

These changing requirements make it exceedingly difficult to align workloads with purpose-built silicon. Hardwired accelerators may work well for matrix multiplication and convolution layers today but simply aren’t efficient or adaptable enough for more modern models, such as transformer-based models like BERT or recommendation models. What’s needed are adaptable AI and ML solutions with a flexible, future-proof approach enabled by efficient, programmable building blocks of hardware IP designed with a software-first ideology.

This leads us to our biggest announcement this week, SiFive Intelligence. Today at the Linley Spring Processor Conference, Dr. Chris Lattner, President of Engineering and Product at SiFive, introduced the SiFive Intelligence X280. The SiFive Intelligence X280 extends RISC-V with SiFive Intelligence Extensions, integrating dedicated AI acceleration technology and extended data type support. SiFive Intelligence includes turnkey software solutions that utilize the X280's features to provide great AI inference performance and support for the full capability of TensorFlow Lite. The result is a programmable, scalable, and configurable platform to meet modern AI/ML processing requirements from the edge to the cloud, providing out-of-the-box compatibility with a wide range of popular machine learning models.

We’re pleased to debut SiFive Intelligence and to be able to tell you that one of the first customers of the SiFive Intelligence X280 is AI developer Tenstorrent, an innovator in AI training and inference processors led by industry leaders Ljubisa Bajic and Jim Keller. The heterogeneous combination of Tenstorrent’s Tensix processing cores alongside the SiFive Intelligence X280 will bring efficiency and programmability to stay ahead of the requirements of next-generation AI workloads. Read more about SiFive's collaboration with Tenstorrent.

This week also saw the announcement of a new partnership for SiFive, with Renesas licensing the SiFive RISC-V Core IP portfolio and collaborating with us on next-generation RISC-V solutions for the automotive markets. Our product portfolio includes 32 and 64-bit embedded real-time cores; advanced, high-performance RISC-V processor cores; and SiFive Intelligence. We’re excited to see what the future will bring as we collaborate to meet the automotive application requirements and the new AI workloads that will come from it. Read more details on the Renesas blog post, Renesas and SiFive Partner to Jointly-Develop Next-Generation High-End RISC-V Solutions for Automotive Applications.

The future of AI processing requires a software-first mindset built on a robust and open ecosystem like RISC-V. We combine RISC-V vectors and SiFive Intelligence Extensions to offer highly power- and memory-efficient compute without sacrificing programmability. RISC-V was invented by the founders of SiFive to build an instruction set architecture for the new golden era of computing. The new SiFive Intelligence line, which is available now, combines all these elements to deliver the flexibility and performance needed by modern AI platforms.