Calista Redmond, chief executive of the microprocessor consortium RISC-V International, is a fan of the wild days of chip competition back in the 1980s.
"This is the biggest opportunity to change the trajectory of computing and hardware that history has seen since the 80s, and that gets me excited every day," said Redmond in a recent interview with ZDNet via Zoom.
She was referring to the flowering in the '80s of numerous different computer chip architectures. They included not only Intel's x86 processors but IBM's POWER architecture; MIPS-based processors made by companies such as NEC and Toshiba; Digital Equipment Corp.'s Alpha series of processors; Sun's Sparc processors; Motorola's PowerPC series; and Hewlett-Packard's PA-RISC series, to name just some of the more obvious chips.
Many of these processor families faded over the decades, leaving two main processor camps, x86, and the ARM processors made by the U.K. firm of that name, which is owned by Japan's Softbank Group, and which is being sold to Nvidia.
RISC-V, which was born a decade ago in a laboratory at the University of California at Berkeley, an academic effort of professors David Patterson and Krste Asanović, is basically the Linux of the chip world, a set of instructions that can be used by all chipmakers and owned by none so that it can be freely modified -- "sort of the kernel for open hardware, just as Linux is the kernel for open software," as Redmond puts it.
RISC-V has not only steadily gained industry support, but it is also starting to lead to technological breakthroughs.
One of the early backers of RISC-V, intellectual property startup SiFive, is partnering with Intel to make possible the fabrication of RISC-V-based chips in Intel's new foundry effort.
And Nvidia's deal to buy ARM may push companies to consider RISC-V, according to some chip executives. "It's a great thing for RISC-V; it's a great thing for any alternative," Xilinx CEO Victor Peng told ZDNet in an interview in May.
To Redmond, the present moment in time offers another shot at diversity in computing that was cut short in the '80s.
"There were a lot of processors that were vying to be the heart and soul of computing," observes Redmond.
However, "Whether it was the early PCs or handsets that came after that, everything had been on a proprietary approach; the fledgling open efforts didn't have all the right ingredients for success at the time.
"This is a monumental shift and change in computer history as we go forward from here," said Redmond. "We are seeing huge, massive investment."
Redmond comes to the role of running the RISC-V consortium with plenty of hardware experience as well as experience in bridge-building between parties.
She came to the organization nearly three years ago after almost thirteen years at IBM. She ran the "ecosystem" for Big Blue's Z Series mainframe business and was president of the OpenPOWER foundation, the non-profit founded in 2013 to develop an ecosystem for POWER chips. Redmond also spent over two years on the board of directors of the Open Mainframe Project, the organization formed in 2015 to bring Linux to mainframes.
She is, in other words, highly experienced in how coalitions take shape and how they can be nurtured.
"I run a tight ship, and part of that is membership, and part of that is growing our members," said Redmond. "We have a tremendous amount of folks who have signed on, ranging from students to entrepreneurs to startups all the way up to multinationals." The membership has doubled in the past year, to more than 2000 members. "That's traction we are continuing to cultivate across the community."
Redmond's enthusiasm for changing computing is shared by RISC-V's CTO, Mark Himelstein. "We are at an inflection point," Himelstein told ZDNet in the same interview with Redmond. "There are going to be hundreds of millions of cores going out this year because of the renaissance in integrated design for IoT and SoC, and so forth."
"Even if somebody has one chip on a board, they may have ten RISC-V chips for a specialized purpose," says Himelstein. Being able to source intellectual property from an expanding ecosystem makes the flexibility of RISC-V chips about more than just a license-free arrangement, he said.
"Our bumper sticker is, Don't duplicate, innovate," said Himelstein, referring to extensions that can be a common set of capabilities for all RISC-V users.
"We are making tracks and working on what the community thinks is important," said Himelstein. "We have expanded especially our software effort significantly," with fifteen working groups, he said.
That includes, said Redmond, an expanding set of contributions on software over and above the RISC-V instruction set specification itself. "It's also a marker of success as we grow beyond base hardware elements and tools and design resources into the software and other aspects of an ecosystem," including "the operating systems, the specific applications and workloads that are spanning across industries."
The rise of open source is aiding the development of that software ecosystem.
"We already have operating systems which are well versed in the notion of running on multiple architectures," she pointed out. "Canonical and Ubuntu and SusE are already investing in multiple architectures, and RISC-V is a no-brainer. They want to be there too."
Redmond cites as progress not just the increasing volume of development but the increasing complexity of parts. "It's the dimensions of the cores," she observed. "RISC-V began in academia, but then moved quickly to embedded and other small, simple, low-power designs."
"And the interesting events now are that we're seeing RISC-V proliferating across all kinds of computing, not just relegated to that one corner, but in fact growing into multi-core, the largest of systems, the largest of chips, the scale-out variety, everything from embedded to enterprise, including workloads that are difficult for even proprietary architectures to transcend."
"Everywhere from soldering irons to super-computers," Himelstein added.
That path from soldering iron to super-computer is happening as Redmond builds a consortium that protects and nurtures the instruction set without itself having any commercial ambition.
In the case of Nvidia and its purchase of ARM, Redmond notes Nvidia is "a long-time supporter" of RISC-V and has expressed its "strategic intent to remain on RISC-V."
"An interesting angle is that sometimes RISC-V is a both/and situation," she reflected. "There are instances where there can be both RISC-V and other architectures on the same chip." Himelstein concurred, observing that "there are a lot of people that are multi-denominational."
Some of the progress in RISC-V is hard to see because no matter how successful RISC-V proves, the world may never know the full extent of its usage. That is because while ARM and other commercial technology providers make their licensees sign documents, no one using RISC-V has to disclose usage.
RISC-V International asks vendors to disclose usage voluntarily but does not compel such disclosure.
Asked if measuring the advancement of RISC-V can be mysterious, Redmond replied, "that's somewhat fair to say: unfortunately, we can't show everyone's roadmaps" of their planned chip design using the instruction set.
She notes a number of already public facts, such as the European Processor Initiative, which is trying to "adopt an open approach to computing," with which RISC-V is "very involved," she said. "In Asia-Pacific, you see a lot of activity happening, from handsets to automotive," she observed, straight through the automotive supply chain, especially in Japan.
Pakistan has "declared RISC-V their national [chip] architecture," she noted, and India has a Shakti chip project built on RISC-V. In North America, "many multinationals are infusing RISC-V as part of their overall chip strategy," she said, including Nvidia and Google.
Clearly, an open instruction set could greatly benefit cloud computing companies such as Google and Amazon. Patterson, the co-inventor of RISC-V, was a consultant inside Google for several years in the development of the TPU processor for machine learning. Alibaba is the only cloud company that has publicly disclosed the use of RISC-V.
Asked about whether other cloud companies are working on RISC-V, Redmond demurred. "My apologies," she said with a smile.
In Redmond's view, the steady march of RISC-V is assured by the fact that the consortium she is helping to build means an ecosystem can evolve much faster than it did with either Intel or ARM.
"Back in the '80s, there were a lot of processor wars, and the shakedown kind of landed on Intel and later ARM, and they all took decades to cultivate this ecosystem," reflected Redmond.
"That's something that RISC-V faces as well, but I can assure you this is not going to take decades to enable that ecosystem or to solve compatibility and portability.
"Those are things we have a line of sight to if not solutions to already."