Big tech firms invest in the RISC-V standard as artificial intelligence drives greater interest in chip innovation
Chinese chip maker Sophgo’s SG2042 chip is pictured at the 2023 World Artificial Intelligence Conference in Shanghai in July. The chip is based on the RISC-V open design, which is gaining in popularity as tech giants seek specialized hardware. PHOTO: CFOTO/ZUMA PRESS
A decade-old standard for designing semiconductors called RISC-V is gaining traction as technology companies look at making their own high-performance and specialized chips for artificial intelligence and mobile devices.
RISC-V, pronounced “risk five,” was developed in 2010 as a simplified, open-source chip “instruction set. ” It can be used to create a variety of chip blueprints for devices like smartwatches, laptops and data center servers, and is easier to customize for specific uses like AI computing. It is also free—unlike proprietary chip designs from market leader Arm—so processors can be built at lower cost with possibly better efficiency.
A surge of activity among tech heavyweights and more niche players for custom chips designed for specific functions is helping fuel RISC-V’s rise. “It’s the first real alternative that we’ve had in the market for many, many years,” said Richard Wawrzyniak, an analyst at market research and consulting firm SHD Group. Building blocks for RISC-V, such as developer communities and tools, are making chip designers more comfortable using it, he said.
Open source technology like RISC-V is offered free to developers, though companies often pay vendors for add-ons like enterprise features or tech support. The startup SiFive, for instance, sells licenses for its chip designs based on RISC-V. A chip instruction set is essentially the base-level in which software communicates with hardware, used to define a processor’s functions.
Meta is using RISC-V to power some of its AI computing, the company wrote in a May blog post. Running generative AI and other large AI models requires more efficient hardware systems, Meta wrote, so it designed its own rather than rely only on traditional graphics processing units. Meta declined to comment on RISC-V.
Google said it began supporting RISC-V for its Android mobile operating system last year. In October, Google announced a partnership with
Qualcomm to launch RISC-V based wearable devices. In November, Google’s research division released a set of open-source tools for AI and machine-learning based on RISC-V.
Hardware designed for specific, specialized applications will be the way to improve chip performance as transistors reach their physical limits, some experts say. That is the benefit of an open standard like RISC-V, which Google said allows for greater customization. Meta said its RISC-V processor cores are “heavily customized” for AI computing efficiency.
Other companies, including Qualcomm, Samsung, Nvidia, and even Intel—which produces a widely used chip instruction set called x86—are also exploring the technology. Qualcomm in August started a company with chip makers including NXP and Infineon to build and sell mobile and internet-connected RISC-V-based devices.
“With these older instruction sets, you would have to license something or get an architecture license to start to create a product,” said Ziad Asghar, Qualcomm’s senior vice president of product management. “But with this being an open instruction set, you can start without having to pay upfront licensing fees.”
That is a primary reason RISC-V appeals to companies compared with pay-to-use designs like Arm’s, said
Forrester research director Glenn O’Donnell. He noted that, “The licensing fees are not so deadly, and you can therefore make really cheap processors.”
Last year, Arm took the unusual step of suing Qualcomm, which accounts for more than 10% of its sales, over the company’s attempt to pay lower royalties on chips developed by a firm it acquired.
RISC-V-based unit shipments are projected to grow at a compound annual rate of 40%, according to estimates by SHD Group. Generative AI presents a major opportunity for developing new silicon based on RISC-V, the firm said, and RISC-V chips could generate revenues of nearly $100 billion and capture 25% of the SoC market, or “system-on-chip” products, by 2030.
The recent industry moves potentially threaten British chip designer Arm, which made its market debut in September. Arm has 90% of the market for smartphone, tablet and laptop processors. The company declined to comment on RISC-V.
RISC-V could present a bigger challenge if it continues gaining traction in China, whose tech companies have embraced the open standard as they seek to end dependence on Western chip technology.
Still, the “middleware” software supporting RISC-V, which makes it easier for enterprises to use without additional development work, has yet to be fully developed for high-performance applications like data centers, said Dell Chief Technology Officer John Roese.
“Until you have enough of a software and developer ecosystem, these things stay very niche,” Roese said, referring to RISC-V’s visibility among corporate buyers.
RISC-V International CEO Calista Redmond said the standard’s governing body plans to roll out certifications and more developer resources next year. “It’s very much a formidable third architecture,” she said. “Just a few years ago, there were only two formidable games in town: the x86 or Arm.”
Comments