In April 2019, RK announced its flagship AI chip – RK3588, which was originally planned for mass production in 2020. However, due to various reasons such as the pandemic and IPO impacts, the chip was delayed. It is now reported that the chip will begin sampling in Q1 2021.
Image from the internet
Six major features highlight the flagship product positioning:
1. Advanced 8nm process technology
8-core CPU, 4 cores A76 + 4 cores A55, maximum frequency 2.0G
GPU adopts the latest four-core Odin architecture
2. 6T NPU computing power, using RK’s self-developed NN architecture
3. Self-developed dual 8K video codec: 8K60fps video decoding capability, 8K30fps video encoding capability
4. 4800W ISP 3.0 architecture, supports multi-camera input
5. Supports 8K60fps display output; HDMI2.1, MIPI, EDP quad-screen easy display output
6. Rich high-speed peripheral interfaces: USB3.1, PCIe3.0, SATA3.0, Gigabit Ethernet, etc.
From the specifications, it can be compared with HiSilicon’s previous flagship chip 3559AV100, which has exceeded the specifications.
This chip is designed for the AIoT market, and currently, the main application scenarios are:
Consumer electronics (C-end): PC and tablets, 8K TVs
Smart security (B-end): Smart front-end IPC, Smart NVR (32-channel 1080p decoding)
Display devices (B-end): Video conferencing, electronic whiteboards
Main specifications of Hi3559AV100
HiSilicon’s 3559A is currently in extreme shortage in the market, with prices soaring to 4500 RMB, making it difficult for most companies to proceed. This delayed chip from Rockchip may take a share in the 3559A market next year. Based on the pricing speculation for its high-end flagship model, it may be similar to the previous market price of 3559A, possibly ranging between 45-70 USD depending on different customer shipment volumes.
Related links from previous issues:
HiSilicon 3559A is out of stock; what should be used as a replacement for edge intelligence?
Why is there no unified industry standard for AI chip performance testing?
From the table, SigmaStar’s multiple AIoT chips released at the 2020 conference
Scan the code to follow, sharing interesting, valuable, and objective observations of the AI industry
