مقالات التقنية

How RISC-V has become a viable third processor architecture | Calista Redmond


Check out all the on-demand sessions from the Intelligent Security Summit here.


The economy and the event business aren’t strong, but the RISC-V Summit drew about 1,000 people to San Jose, California, this week to hear the latest on the open-source processor.

RISC-V International CEO Calista Redmond said the numbers were down from last year, but she attributed that it part to conference attendance budgets hit by the downturn as well as lingering COVID concerns.

The nonprofit group’s membership has been steadily growing from a ragtag group of feisty academics to some of the biggest tech companies like Google and Nvidia. Over the past decade, the group has groomed RISC-V into a viable alternative to proprietary Arm and Intel-based processors. Last year, Deloitte predicted that RISC-V chip revenues would grow from $400 million in 2021 and reach $1 billion in revenue by 2024.

RISC-V International has seen more than 26% membership growth year-over-year, with over 3,180 members across 70 countries. The RISC-V community ratified six extensions and specifications this past year, with 10 additional specifications and extensions are expected to be ratified in the next 120 days.

Event

Intelligent Security Summit On-Demand

Learn the critical role of AI & ML in cybersecurity and industry specific case studies. Watch on-demand sessions today.

Watch Here

RISC-V International now has 81 technical groups. More than 30 of those groups were added in 2022 alone to focus on fast-growth markets for RISC-V, such as security, system-on-chip infrastructure, automotive and AI/ML.

One of the benefits of RISC-V is that it is sanction-free. As an open-source platform, RISC-V is not affected by export restrictions. This makes it appealing to companies, especially in China, that have been affected or fear being affected by those restrictions, Deloitte said.

Deloitte said that companies are planning on using it for different storage, graphics, and machine-learning applications. Even Intel is partnering with RISC-V player SiFive. Arm argues that it has more features and has more support options for developers. But Redmond notes rivals can be more expensive.

Redmond noted that since RISC-V is royalty-free, it isn’t as easy to do the accounting and figure out the exact number of RISC-V chips that have shipped. But estimates put it at around 10 billion cores in the market. That makes it small compared to x86 and Arm, but it’s growing. The RISC-V opportunity is also spreading out.

Redmond noted that Ventana, SiFive, and Andes all have the ambition of trying to penetrate the datacenters of the world with RISC-V processors that can handle special tasks that other processors aren’t designed to do.

Here’s an edited transcript of our interview.

Calista Redmond is CEO of RISC-V International.

VentureBeat: How many people are coming this week?

Calista Redmond: We’re over a thousand. We don’t have final numbers yet.

VentureBeat: Is that pretty similar to the past?

Redmond: Composition-wise, I think it’s somewhat similar. We see a bit lower than what we were hoping for, but I’d attribute that mostly to some of the budget constraints that companies have been having as they go into the fourth quarter, as well as probably some COVID concern and caution. We asked everyone to be vaccinated to join our event. We’re trying to maintain a lot of protocols that way.

VentureBeat: Was there a year you canceled the event?

Redmond: We went virtual one year. Last year we were together with DAC, and then the year prior to that we were virtual. 2019 was our last stand-alone event here. I think our numbers are pretty consistent with what the rest of the industry is seeing in terms of in-person events and distribution of in-person and virtual and things like that.

Some of the biggest companies in the RISC-V market.

VentureBeat: As far as big themes, is RISC-V still gaining ground on other [chip architectures]?

Redmond: Oh, absolutely. Absolutely. It’s a combination of a lot of new ground, uncharted territory, green field. A lot more consideration being given to some of those existing workloads. This year you saw a lot of announcements that span a deeper performance end of the spectrum of computing, not simply in one end. As well as further inroads in some industries. Automotive has been taking off. In HPC we’re seeing a lot of great traction.

A lot of folks are excited about datacenter. The datacenter is an area that–it’s a tough nut to crack. It’s also an area where there’s both existing workloads as well as some new capabilities that are being brought into data centers. Hyperscale, other large implementations, to bring acceleration. AI, ML, those types of capabilities, as well as the neverending expansion of IOT and other types of endpoint technology and coordination.

VentureBeat: Ventana, SiFive, they seem to have intentions to challenge the big guys, the biggest processors.

Redmond: There are a lot of our members who are focused on getting some of those high-performance aspects. You’ve named some of them. Andes is also on that list. They have some higher-performance things coming on now. I don’t know that all of those high-performance ones are challenging incumbents, but there’s such an explosion in opportunity. There’s room for everyone.

VentureBeat: I saw David Patterson’s post on the fallacies. That was pretty good. He’s very able to explain these things. The neutral, Switzerland argument seems to be a very important one still.

Redmond: We’re still there. I don’t think that it’s a top of mind thing these days. We’ve been there for a few years now.

VentureBeat: But the notion that the architecture is neutral–

Redmond: I don’t know that that has anything to do with Switzerland. When you think about global standards, it’s really something that’s in the public domain everywhere, regardless of any incorporated entity behind it. Some standards may not even have an incorporated entity behind it. At RISC-V we’re the stewards of this particular global standard, as well as the extensions and specifications around it. Once you bring something into the public domain, it’s no longer held by any one company or country. RISC-V in the public domain as a global standard is not held by Switzerland. Folks like Switzerland as far as the RISC-V entity, but it doesn’t have anything to do with the technical deliverables.

VentureBeat: But you draw attention because there are so many problems associated with others that are the closed architectures, that have some very geographic concerns.

Redmond: Proprietary architectures are held by single companies. RISC-V is not–the architecture is not held by a single company. RISC-V the architecture is contributed to the public domain. It’s similar to Creative Commons or other global standards like Ethernet. Anyone, anywhere can describe their designs as Ethernet. It’s not about Switzerland. It’s about how we’re in the public domain versus proprietary architectures that have their destiny tied to a particular company.

VentureBeat: That seems to explain one reason for momentum behind it. Are there a lot of other things, though, that also explain some of the momentum, why it’s growing?

Redmond: Massive opportunity. If you draw up a business case on where you want to strategically go, you want to have the design flexibility and freedom to get there. And that massive opportunity is being realized. More than 10 billion cores in the market already. Those implementations have been, a lot of them, in the green field opportunity. Is it the green field opportunity? Is it the opportunity to take an existing design and move into an adjacent market?

More than 1,000 people attended the RISC-V Summit in 2022.

When you build a business case, it’s about massive opportunity in a high growth space. We see that in semiconductors. It’s about time to market. Can you accelerate that path? Well, you’ve accelerated beyond a lot of the barriers to entry already. The initial investment cost to get into that space is much easier to overcome when you start with an open global standard like RISC-V. That’s one of the areas.

The next area that would be considered in the business case is, can you differentiate? Can you become extremely competitive versus your peers? Well, you can’t do that if you’re locked into a proprietary constraint. You can do that with the design flexibility and freedom that you get in RISC-V. And by the way, in addition to being able to do what you want, it’s designed for that. It’s designed so that you have software acceleration top of mind, that you have tight controls on power consumption, on code size, on the various design elements. Vector is a great example of something we bring in that is a very competitive differentiator, something you can do on RISC-V that you can’t do elsewhere. If you’re seeking to go into some of that exploding opportunity, you want to go with speed and with something that’s as competitive as possible.

VentureBeat: With open source standardized things, people often say that route is slower, that going with something that’s known and proprietary speeds you along into the market. It sounds like you’d consider that a fallacy. No one owning it, in one sense, could be a recipe for nobody taking responsibility.

Redmond: No, it’s everyone’s vested interest. The way it works, you’re not sitting back and waiting for a community to develop your product. You’re accelerating and engaging in the pieces of that puzzle that are critical to you. All of these building blocks are there. You take those, as an independent organization or entity or company or researcher or what have you, and say, “All right, I’ll take this nugget and go add on to it with my competitive differentiation and I’m going to market.” You don’t have to negotiate for years with millions of dollars to get that base building block. An architecture license can be quite an inhibitor to entering a new space. I’m not sure that it’s a slower path unless there’s a piece of that base nugget that you’re waiting for. At this stage, we’ve overcome many of those base building blocks so that you have what you need to go after a space.

RISC-V cores are being used in all sorts of devices.

Now, that’s not to say we’re 100 percent complete across all elements. We have a lot of work underway. For those elements that are strategically important to you, that’s an opportunity to either go off on your own and support it yourself into eternity, or engage in a work group to get to a ratified extension or specification to modularly build into your design. And you can do both. You can enact your own proprietary extension for now and then, in your next generation, come back and bring in that other piece.

But what you might also appreciate is that you want to design for long term strategic durability. That means you don’t want your product or your solution to become obsolete with the next generation. It’s that compatibility and that interoperability and all those other great features that you want to persist through time, because you’re also not just working in isolation with your product. You’re also trying to tie into a larger ecosystem.

Profiles are our answer to how you get to a combination of those common building blocks, composed such that you can ensure compatibility baked down to that level, to help the ecosystem rather than the ecosystem having to say, “Let me compose a design. How would that work?” Having some base designs that are readily available. That’s what our profiles are. Like a reference design, but not as an implementation. It’s a positive profile. It’s accelerating the ecosystem, making the onramp a lot easier and faster.

VentureBeat: What kind of feedback do the members give? What do you jump on to implement some of the things they suggest?

Redmond: We’re very member-driven. When those priorities come to us, step one is, is this important to multiple people? We try to pull those efforts together. There’s an entire structured governance process. That doesn’t mean it’s slow. There is discipline to our process that says, “All right, let’s pull this together. Let’s charter a work group. Let’s get some chairs and vice-chairs and let’s get the work done. Let’s bring it through a review and ratification process.” We ensure that this is not a singularly driven organization with deliverables that have been put together or ratified by just a single organization. It’s something for the benefit of the community.

VentureBeat: Looking at blockchain, f you’re going to bring something new into RISC-V, you have to assemble the experts to assess whether it affects the architecture?

Redmond: The members who are deep into that come together very quickly. It’s not as if somebody has an idea but has no one around to look for wisdom on it.

VentureBeat: The hot topics in tech, it’s AI, blockchain, and metaverse. I don’t know if it sinks down to the level of RISC-V, as to whether people are discussing these things, if they’re going to have to have their future processors execute against them.

Redmond: I haven’t heard a lot of engagement around metaverse topics yet. I’m not sure if there’s someone we could point you to on that. We have around different 85 groups in motion now, which is why we need lots of discipline across these groups. We need to work in harmony. That’s part of the benefit of bringing–if you have six companies all working on the same piece of the puzzle, bring it together in this group. Let’s invent it once, not multiple times, and you can all go work on your more differentiated stuff once we finalize that together.

RISC-V Summit expo floor.

VentureBeat: Those 85 groups, do they have a certain kind of makeup? Are they application-driven?

Redmond: Those groups are different technical deliverable work groups. We have committees that oversee combinations of these groups that are within different areas. There’s software, security, other things that are horizontal committees. We have verticals around priv and un-priv implementations on things. Then we also have special interest groups around specific industries, industry areas. Android or automotive or data center, things like that. They get together and say, “Let’s agree on the priorities for what our output is going to be.” Or strategic directions that they need to take, those strategic decisions that can be made as you cultivate these new variables. All of them have end utility or applications in mind as they set those priorities.

VentureBeat: Are you counting the chips that ship with the architecture? Where is that?

Redmond: Our licensing model doesn’t actually tie back to all the individual implementations. As an open license, you don’t have that kind of visibility. We don’t have a granularity on that type of information. But we’re looking at where are the trajectories, what are we seeing and understanding across the market, to understand where those investments in RISC-V are taking hold and beginning to grow. I’m sure you’ve seen our presentations on our growth so far. The Qualcomm figure from yesterday, just from them, was 650 million cores. But that’s a drop in a larger bucket of 10 billion cores that we’re estimating in market, just based on anecdotal information that we hear from a lot of our larger constituents.

VentureBeat: David Patterson’s suggestion that it’s the third major architecture right now, the third contender, is supportable.

Redmond: We like to think that there are an even three. Choose your favorite. That’s just it. Choice encourages and enables innovation and competition. It avoids the lock-in.

VentureBeat: It’s interesting that this part of the whole ecosystem is going in that direction. I don’t see other parts of the ecosystem unlocking. Everybody calls Big Tech that for a reason. They all seem to have monopolized their own sectors.

Redmond: Here’s a nice story. Once upon a time, datacenters were pretty much a single logo investment. You picked your servers and you’re going to get your matching networking protocols and everything. And then along came Ethernet. It disrupted that entire space. Who thought that would be the disruption there? Pretty soon you can plug and play whatever you want in that data center. Semiconductors were creating a base standard approach to that same equation, that same design point. You’ve seen that in operating systems. Once upon a time Microsoft ran everything. Along came Linux. Pretty soon planes would drop out of the sky without it. Linux underpins most of the major systems in the world now.

Whether it’s a global standard or other things put into the public domain, such as software, that has fueled a lot of creativity in business. If your focus is more on the hardware side, there are some other examples where global standards have taken hold. Different peripheral interconnects and things with different systems, or in systems themselves there are things like the Open Compute Project, which has been something that you see in hyperscale. But that creates that common recipe or common footprint that a broader set of stakeholders can engage in. You are no longer a single supply shop. You have supplier diversity, things like that. Open collaboration is not new in hardware. It just takes on a slightly different flavor, a slightly different direction. But I do think there continues to be more room for that.

SiFive is making licensable RISC-V CPUs.

VentureBeat: Have you seen much threat of forking, where people disagree on something?

Redmond: People surface that as a concern, but it has no merit. When you think about the sheer forces working against forking, no one in their right mind would actually do it. There is no nation or company wealthy enough to truly build their own ecosystem around another ISA at this point. If you were to do that, to take it and fork it off into the weeds, it’s on your shoulders then to support that into perpetuity. Even with RISC-V, no one would want to do that, because RISC-V has gained so much ground globally. That global importance of everyone being able to work from the same base model is incredibly important. The forces working against fragmentation or forking are pretty significant at this point.

That comes down to not just the cost, but that long term strategic importance and path that companies want to take. If you were to fork or fragment, you’ve just added incredible cost and development time and other resources to supporting the ecosystem that you need to build.

VentureBeat: Is the reuse of different pieces that get developed a good measure of progress in the ecosystem?

Redmond: For the ecosystem, there’s been a history of being able to work on multiple architectures. You look at an organization like Canonical. They work on all the architectures. They make it a point of pride that they work on all the architectures. Many software organizations have that. It only makes business sense. You don’t want to be a single business interest. You want to have multiple business interests. That’s where having your application be able to run effectively across multiple architectures is important. That’s what Android has said. The time is right. Let’s go. Let’s do this on RISC-V as well as other architectures. But you’ve seen that history repeat itself. Businesses have said that it’s a good business to participate in this space.

VentureBeat: Is there some hope that some of the biggest companies will go this direction? It’s interesting that some have done it quietly. Nvidia has a bunch of RISC-V right now.

Redmond: You can even look back at Jensen’s comments. In the early days of learning about the ARM acquisition, even at that time he said that regardless of what direction the ARM transaction went – and I’m not directly quoting him, because I don’t exactly remember his words – they were committed to RISC-V in their strategic path forward. RISC-V offers something that alternative architectures don’t. Nvidia has continued to be a great member of the organization. We have someone from Nvidia on our board of directors.

We have other, I would say, considerable stakeholders engaged in the community, whether that’s Qualcomm and their 650 million cores that they’ve shipped, or Intel. Intel is a great example. They have their own architecture, but they’ve decided that it’s in their business interest to also embrace RISC-V. It’s another part of their business, but that’s of strategic importance to them. They put a billion-dollar flag on it this year. You look at organizations like Google. Google leveled up their membership with RISC-V. Not this year, but a year ago. The reason they did was that they said, “You know what, there are many parts of our organization that are engaged in RISC-V. This is strategically important to our company.” It is growing. We are seeing this deeper investment in RISC-V as it grows in its strategic importance.

RISC-V rack cluster
RISC-V rack cluster

VentureBeat: I cover a lot of metaverse now. The metaverse seems like the most electronics-intensive application possibly imaginable. The only problem is that Moore’s Law comes to an end just in time for that.

Redmond: There’s a couple of schools on that. Number one, you get Gelsinger up there saying that there’s more left to the periodic table that we haven’t used yet. That’s kind of fun. But number two, you still have the danger of physics. It gets to be very problematic at a certain size.

VentureBeat: If you then try to solve that kind of huge problem at a time when you don’t have as much innovation happening on the chip level, I could see that consuming an awful lot of energy. There’s something about this whole big picture that doesn’t work. Maybe it’s nuclear fusion now.

Redmond: Exactly! That’s a physics problem we can’t get to. If you’re in the gaming and metaverse space fairly often, did you hear any talk of RISC-V over there?

VentureBeat: Not so much that–only really in the respect that geopolitics is the enemy of applications with a global ambition. Anything that balkanizes the world, whether it’s the gaming world or the software world or the hardware world, even the console space, is bad for what people ultimately want to get to.

Redmond: The commercial side, that’s a formidable challenge, no doubt. Over the course of the last few decades, open collaboration, global standards have risen above some of the tensions to build bridges. I remain pretty hopeful that we’ll get to a better space.

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.

مقالات ذات صلة

اترك تعليقاً

لن يتم نشر عنوان بريدك الإلكتروني. الحقول الإلزامية مشار إليها بـ *

زر الذهاب إلى الأعلى