SiFive RISC-V Core IP

SiFive Performance

High Performance RISC-V Processors

The Challenge

Since its inception, the RISC-V community has primarily served the deeply embedded processor market, where power, footprint, and cost constraints generally override the need for high performance. Add to these constraints the complexity of delivering new RISC-V technology that stands shoulder to shoulder with advanced proprietary architectures, and it is clear that a large category of processor consumers has, before now, gone underserved without a variety of high-performance choices from the market.

The SiFive Solution

The SiFive® Performance™ family of processors is designed for maximum throughput, while preserving efficiency for workloads as varied as branch-intensive operating system, and multimedia processing.

Ranging from 8-stage, dual-issue, in-order architectures equipped with 256-bit vector engines, to advanced out-of-order processors with best-in-class industry benchmark performance, the SiFive Performance Family represents the next wave and future of RISC-V compute.

The SiFive Performance family of processors is designed for maximum throughput, while preserving efficiency for workloads as varied as branch-intensive operating system, and multimedia processing. Ranging from 8-stage, dual-issue, in-order architectures equipped with 256-bit vector engines, to advanced out-of-order processors with best-in-class industry benchmark performance, the SiFive Performance Family represents the next wave and future of RISC-V compute.

Efficient performance vector application core

Highest performance application core

Performance P200 Series

  • 8-Stage, Dual-Issue, In-Order processor
  • 256b Vector Unit enabling versatile compute capabilities
  • Supports multicore coherence with up to 4 cores in a core complex
  • New SiFive Recode utility translates legacy SIMD code to SiFive vector assembly, preserving software investment while allowing for easy migration

Performance P500 Series

  • 13 stage, 3-issue high-performance out-of-order pipeline
  • SPECint 2006 – 8.65/GHz
  • 2.4 GHz in 7nm with a footprint of less than 0.38 mm**
  • Supports multicore coherence with up to 4 cores in a core complex

*Cortex-A75 performance data source:
https://www.anandtech.com/show/13503/the-mate-20-mate-20-pro-review/3

**Based on P550 with 32KB L1 I$ and D$ and 256KB pL2 vs estimated 7nm Cortex-A75 area from:
https://www.linleygroup.com/mpr/article.php?id=12151

SiFive Performance Target Markets

  • Enterprise Switching/Routing/Storage, Smart NICs
  • Edge Analytics, Big-Data Analytics
  • Autonomous Machines
  • Edge Compute
  • 5G Infrastructure/Base Stations
  • AR/VR/MR/XR