With its open features, the RISC-V architecture and instruction set are gaining more and more attention. Whether it is China or Intel, Samsung, Qualcomm and other giants, RISC-V is also regarded as a potential alternative for ARM and even x86.”
In fact, NVIDIA is also very concerned about the RISC-V architecture. It has joined the RISC-V Foundation early and has done a lot of research. Recently, it has also revealed the possibility of applying the RISC-V instruction set in Deep Neural Network (DNN).
NVIDIA said that DNN needs high performance, high precision, and is sensitive to power consumption. It is not easy and costly to build a DNN accelerator. Therefore, it is planned to use low-power, high-bandwidth chip interconnect technology to accelerate multiple inferences. The chips form a mesh mesh network.
To read more, please visit: https://news.mydrivers.com/1/631/631962.htm. Please note that the original article is in Chinese.