[GTALUG] "RISC-V technology emerges as battleground in US-China tech war"

[D. Hugh Redelmeier]

| From: Evan Leibovitch via talk <talk at gtalug.org>

| I think there's something very different at play here. While I'm sure
| existing chipmakers are whispering in Congress' ear, they don't have any
| selling to do. Since the US has already put export controls on advanced
| chipmaking technology and equipment in China, RISC-V can be trivially
| advanced as a path to circumvent such controls. It's no secret that the HQ
| of the RISC-V consortium was moved from the US to Switzerland explicitly to
| inhibit any one country (or bloc of countries) from inhibiting its progress.

Right.  There are a bunch of hard things about making a competitive 
processor chip or SoC.  Designing an Instruction Set Architecture is not 
one of them (I've done it myself).

Launching a new architecture and getting enough infrastructure to be 
useful seems to be the hardest.  RISC-V is almost there.  Its design is 
complete enough to manufacture.  There are a bunch of SoCs that you can 
buy now.  Linux runs on them.  That genie is out of the bottle.

The remaining barriers have nothing to do with RISC-V International.

Designing fast implementation designs ("micorarchitecture") is hard.  
There are open implementations but they lack credibility.  That may 
change.

ARM licences a bunch of microarchitecture and most ARM chips use one or 
two of these.  (The major exception is Apple -- they use their own 
microarchitecture.)

The ISA also is missing ancillary functions:

- PMIC
- USB
- GPU
- PCIe / NVMe
- TPU
- SATA
- DRAM controller

Fabricating chips from these designs is hard too.  That's what the US has 
cracked down on: ASML in the Netherlands has a lock on making masks.  
This ban hurts a tiny company like the Netherlands a lot.

TSMC, headquartered in Taiwan is the leading fab in recent years. They are 
barred from selling some stuff to PRC.  This hurts them a lot.  It might 
be sufficient to cause PRC to invade.


A US ban on RISC-V International contributions from the USA will have a 
few effects:

- development will be slowed

- China will pull further ahead in control and contribution to RISC-V

- US looks like an unreliable partner.  (Look at NVidia's stock this 
  week.)

In effect China will be ahead from this move.

The only way to keep technical leadership is to advance more quickly than 
you competitors.