AI Chips

References:

General-purpose Chips (CPU) vs. Specialized Chips (GPU, FPGA, ASIC)

Different types of AI chips are useful for different tasks. GPUs are most often used for initially developing and refining AI algorithms; This process is known as “training”. FPGAs are mostly used to apply trained AI algorithms to real-world data inputs; This is often called “inference”. ASICs can be designed for either training or inference.

“An AI chip a thousand times as effcient as a CPU provides an improvement equivalent to 26 years of Moore’s Law-driven CPU improvements.”

GPU Vendors:

  • Nvidia
  • AMD
  • 景嘉微电子

FPGA Vendors:

  • Xilinx
  • Intel
  • 上海复旦微电子(1998)
  • 安路科技(2011)
  • 紫光同创(2013)
  • 高云半导体(2014)

ASIC Vendors:

  • Google, TPU
  • Tesla,
    • 2021-06, 72 TOPS.
    • 2021-08-19, D1 chip (362 TOPS), used in Dojo supercomputer (25 chips per tile, 120 tiles)
  • Amazon,
  • Alibaba,
  • Tencent,
  • Baidu, 昆仑
    • 昆仑1,2019.12,256TOPS-Int8, 150W,Samsung代工。
    • 昆仑2,2021.8.18,7nm量产,2~3倍性能提升。

More

Created Aug 31, 2021 // Last Updated Sep 5, 2021

If you could revise
the fundmental principles of
computer system design
to improve security...

... what would you change?