SiFive launches Performance P670 and P470 RISC-V energy efficient processors

SiFive’s expansion in the rapidly evolving RISC-V market continues with the announcement of the P670 and P470 processors - two new models from the Performance lineup that aim to address the need for high performance and efficiency in high volume small-size devices like wearables, smart home appliances, industrial IoT, AR/VR headsets etc. Thanks to the addition of vector instructions, the new solutions are recommended by SiFive as powerful alternatives for the “legacy ISAs,” plus they can offer greater flexibility for the Android Open Source Project ecosystem that already welcomed Qualcomm and Samsung among its prominent supporters.

Technology highlights for the P670 and P470 processors include extended virtualization support, full out-of-order vector implementation with first to market RISC-V vector cryptography, improved security via SiFive’s WorldGuard system, and Advanced Interrupt Architecture compliance.

The P670 processor is designed for premium wearables, networking infrastructure, robotics and mobile devices. It is built on a 5 nm node and can be clocked at more than 3.4 GHz. SiFive estimates that the P670 can score more than 12 SpecINT2k6 per GHz, which is excellent for a processor with 50% reduced area compared to ARM’s Cortex-A78. Additionally, this model offers higher single thread performance and twice the compute density compared to “legacy solutions,” and integrates 2x 128-bit vector ALUs.

As an efficiency-focused and area optimized solution , the P470 processor is suited for entry-level wearables, as well as smart devices. Compared to ARM’s Cortex-A55, the P470 offers 2.75x the performance and 4x the compute density. It is built on the same 5 nm node and can also run at up to 3.4 GHz, with an estimated score of more than 8 SpecINT2k6 per GHz. It integrates one 128-bit vector ALU, but SiFive plans to release a P450 model that excludes the ALU for an even smaller die.

Buy the VisionFive 2 RISC-V AI Single Board Computer on Amazon