Interviews with chip company CEOs are always thought-provoking. In addition to the topics we usually talk about, such as success and failure in the market, revenue and competition, plans and traps, there are also some technical details desired by the technology house. Recently, the register, a foreign magazine, conducted an interview with Calista Redmond, CEO of risc-v international,
The existence of risc-v deserves arm's vigilance
In the interview, Calista Redmond talked about Intel's not too worried about its x86 business, so it became a platinum member of ISA alliance, and what risc-v would look like in 2022 - it began to enter the field of notebook computers, AI, Google super large chips and edge fields, and planned to complete the progress made by arm in 20 years within five years.
Redmond gives a bold view: "risc-v may not pose a threat to Intel, but arm needs to be vigilant about the existence of risc-v."
Redmond pointed out that risc-v has made significant progress, which is indeed based on evidence. After all, obtaining the support of the industry is a great achievement. Risc-v has become a strong competitor in the CPU industry and attracted the attention of the industry. At present, risc-v is gradually establishing its own ecosystem.
These are the only way for risc-v, but there is still a way to go before the real ecological prosperity.
The processor market is a very strange market, and there is no real sign that risc-v will become a killer processor.
The real challenge is not technical capability, nor can it do things that competitors cannot do. For example, risc-v ships in large quantities in the form of embedded and SOC chips, and is the only isa where designers can freely add or modify architecture features at the ISA level. Compared with the market advantages of its two competitors, these are still insignificant.
Better than arm, you can't make "big money"
Look at arm. In 2021, arm sold 30 billion cores, that is, an average of four cores per person in the world, earning $2.7 billion.
Arm's achievements have proved three points. First, the global demand for advanced mobile and embedded cores is huge; Second, arm occupies an absolute leading position in the mobile market; Third, even so, arm didn't make much money.
Maybe $2.7 billion will be considered a lot of money, and the profit margin can not be underestimated, but it only accounts for 0.6% of the global mobile phone revenue of $450 billion. As the core device of mobile phone, the profit earned by ARM processor is not much. If the dominant position of arm changes dramatically, it is definitely not because there are enough profits in the market.
Intel has an annual revenue of $80 billion, which is because Intel has a huge chip factory. For risc-v, Intel isa IP is very similar to arm, and the division of IAS IP is very unique.
Intel and arm are the rulers of their respective fields. Although they have tried to occupy cities and territories in the market dominated by each other for decades, their fields have hardly overlapped. Last year Apple The Mac Mini M1 launched by the company is the first mainstream desktop computer based on ARM processor after acorn computer in the early 1990s (raspberry Pi is not a mainstream desktop computer).
As for smart phones based on Intel x86, there is no record in history. There is no second place in the PC field and mobile field, not even the other party's Isa. Over the past few decades, the development momentum of industry tools and theoretical knowledge, coupled with a large amount of investment in innovation, the mainstream CPU market is divided up by two competitive but dynamic monopolies. If you want to compete with any one of them, you can only catch up with it.
Cost performance should become the competitive advantage of risc-v
To make risc-v truly prosperous, risc-v also needs to have its own territory.
AI / ml can be one of them. There are many accelerators in this field. The design is very diverse, and there is no winner in the industry.
The Internet of things is also a great opportunity, but this is not a new opportunity. The development of the Internet of things has experienced an early experimental stage and is locked in the high-performance standard architecture. In fact, large companies are ready to divide new industries.
Even if a new field does appear, risc-v must have a core competitiveness. Open source does not provide sufficient capital and technical advantages, because most original equipment manufacturers will license the core like arm rather than design their own core. Now SOC can add a lot of architecture innovation without changing Isa.
Frankly speaking, risc-v needs high cost performance. A month ago, the first risc-v SBC visionfive appeared. It is like raspberry PI, but it is larger, slower and more expensive. PI started its own industry because it is based on the mobile phone technology of Broadcom SOC for many years and has the advantages of being smaller, faster and cheaper than its competitors, but risc-v can't bring anything.
Risc-v cannot achieve great success by surpassing PI. It needs to find some application fields in a very attractive way. What is specific is not important, but it must be ensured that risc-v must ensure the first position in this field and leave no room for the second.