After seven years as a professor in the electrical engineering department, President John Hennessy took a sabbatical from Stanford in 1984 to found MIPS Computer Systems Inc. and commercialize his work with RISC processing. The company went public shortly thereafter and enjoyed great success as the pioneer of MIPS architecture. Sir Hossein Yassaie is the CEO of Imagination Technologies Group, which acquired MIPS Technologies last year to use in future CPU development.
The Daily sat down with both Hennessy and Yassaie to talk about the acquisition as well as the history of MIPS and the realities of the semiconductor IP (intellectual property) business and startup companies.
The Stanford Daily (TSD): Can you just explain a little bit about how things happened with the early days of your company and what led up to this point with the MIPS technology?
John Hennessy (JH): The MIPS [Microprocessor without Interlocked Pipeline Stages] architecture had its roots in a research project here at Stanford started in the early 1980s with the simple question: What would the age of single chip CPUs mean for how we should design processors? Stanford, together with Berkeley and IBM, really…stumbled upon an incredible discovery about a better way to design instruction sets for modern computers than had been previously thought about.
And I say “stumbled upon” because really what happened was the constraints of having to fit a processor with the limited number of transistors at that time meant that we had to rethink the whole process of how to design the interface between the software system and the hardware system. That’s what led to the fundamental discovery, but I think it wasn’t until some years later that people really realized how fundamental it was.
In fact, in the beginning there was a lot of skepticism. The IBM project was kept secret for many years, and people didn’t know about it. And it became very easy for many people in the industry to say, “This is a bunch of crazy academics here — this isn’t really a great idea.” It had to prove itself over time.
In more recent times, the emergence of the embedded computer space, with everything from smartphones and tablets to all kinds of smart desktops to really high-end computer video game systems, has meant that an architecture that could deliver a combination of high performance low cost and be able to operate in a regime with low power had real advantages.
What emerged were the second-generation advantages of the RISC [Reduced instruction set computing] approach that we piloted here and at the other institutions. That’s in fact been the driver for the last few years as the computing world has shifted increasingly away from desktops and big machines to things that people carry around or they use in many ways outside of the context of a normal computing task. So that’s sort of a capsule history of 25 or so years with.
We started the company not because we wanted to be entrepreneurs. We were, in fact, reluctant entrepreneurs — we started the company originally because a famous computer scientist convinced us that, if we didn’t start the company, the technology would just languish on the shelves. [The technology] was very threatening to the existing players out there and their machines, and if they adopted this technology, it was going to obsolete their products.
Gordon Bell was then the founder of the second-largest computer company [Digital Equipment Corporation], and he said, “You’ve got to go do this yourself, otherwise you’re going to have a bunch of papers sitting on a shelf that nobody’s ever going to use.” And, in retrospect, it turned out to be very insightful.
Digital Equipment Corporation actually had a person working with us on sabbatical and when they wanted to take the research and transfer it to their headquarters back East, they couldn’t convince the people to take it. IBM cancelled their first attempts with the technology.
In fact, one of the founders of MIPS helped me found the company, because when IBM cancelled the project, he so believed in this technology that he said, “I’m not going back to IBM. If you guys want to do something here, I’m in.”
And, at the time, people were making big computers that cost a lot of money. Room-sized. And we were talking an improvement by a factor of five in cost, conservatively, and maybe a factor of 10. So it was a big change. People were thinking, “Well, I’ll just sell fewer computers, and I’ll make less money,” instead of saying, “Suppose you delivered that much power — what could it do in terms of what could it do and how many people could afford them?” They didn’t see that, and it was early on.
And that’s another reason I say it often takes an entrepreneur to really take it out there, somebody’s who is really been involved in the creation of something. If you’re involved in it, you see the glass as half full. Yeah, there are some problems to be solved, but the key insight is there. Everybody else sees the glass as half empty and says, “This isn’t going to work. This is a research project.”
I still remember this funny story — I was on a panel with Dave Patterson, who was doing a similar project at Berkeley with IBM. And somebody says to him, just after we had started the company, “You know Hennessy here on the panel? He just got a million dollars to invest from the venture capitalists, and he’s going to start this company. If this technology’s not going to work, what should he do? Take the money and go to South America.” [Laughing]
TSD: How does acquiring MIPS accelerate your activities at Imagination, and what was the thinking behind the move?
Hossein Yassaie (HY): We fundamentally believe that the processor market is a big market and is an area that is growing. As I think John touched on, this market started with areas such as game consoles and embedded applications and then mobile phones happened and tablets, which is a very big area.
On the horizon, everyone talks about connected devices and Internet of Things. We went from tens of millions to hundreds of millions to a few billion, and now I think we’re heading toward tens of billions of things which are clever and smart so the processor business is very important for us.
We’re a key IP company with significant customers, so we’ve had our own processor development internally since 1995, and it’s very much a philosophy along the same lines as MIPS that we were working on in terms of multithreading and RISC architectures. But it’s just in the last year in discussion with the MIPS executives and team [that] it became clear that there is a mechanism for us acquiring MIPS in a structure that means we double the team size.
We [had] about 200 people on our processor team, and now we’re double that, and that combination means [that], in terms of capability and know-how, and the additional ecosystems MIPS brings to us — particularly with the fact that MIPS is in the Android tree — our process of getting where we want to be with the processor technology scaled up, both in times of resources and where we are in the marketplace.
TSD: There was some speculation in the early 1990s that MIPS and other powerful RISC processors would overtake the Intel IA32 architecture for desktop and mobile computers, but this hasn’t really been the case. Were you surprised that this didn’t catch on and why do you think that the architecture never really caught on and has stayed mostly out of the desktop?
JH: I think lots of things evolved as the technology shifted. There are a whole lot of complicated reasons for why things happened on the desktop the way they happened. One of the very hard things to do is to get companies to collaborate, even if by collaboration they would be stronger than if they fractured themselves.
Intel was the big player outside the RISC space, and in the RISC space things just fractured. A couple of times we had a shot at bringing it all together and getting enough players on one platform that you would have a shot at Intel, that never quite happened for a number of complicated political reasons.
Then what happened was the rise of this shift to mobile and embedded technologies — it happened on Internet switches, it happened on desktops, on desktop machines supporting television sets and set-top boxes. And that market and the video game, the high-end video game market just took off. Boom. Exploded.
And all of a sudden that was growing much faster than the main desktop business. And things have shifted dramatically. Now if you look beyond ARM and MIPS, they dominate the rest of that marketplace out there and Intel still has this standardized marketplace. But the advantages that the RISC architectures have in terms of efficiency, both in use of silicon which means they are cheaper and, equally importantly, power. Power has become a big, big issue in these things. Unlike a big computer somewhere [when] you have a fan to cool it [or] you have it plugged into the wall, you don’t have that.
TSD: Are those factors you talked about with efficiency things that make MIPS attractive in low-cost markets maybe for the future?
HY: From Imagination’s point of view, we completely agree with the comment John just made about power. One of the major constraints in mobile devices is that there is not a cable that is plugged into power and there is not a fan in there. So you have a set power envelope that you have to live with. And not only that, the demands that consumers have: they want it be to be as good as their PCs.
And actually we’re pretty much there. That’s really, if you look at the history of Imagination, we built our business in the mid-90s saying, “OK, graphics is a big deal for mobile. Against all the odds let’s put our efforts into creating a very powerful solution for mobile and get graphics into the devices and handsets.”
Now going forward, everything needs to be smart, everything needs a CPU and power is still a constraint. We certainly believe MIPS is one of the most efficient architectures that exists on this planet and certainly have expectations and aspirations that it will grow from this base with the help and energy we can put behind and become a big player in the market space.
JH: Efficiency has become the key issue — that’s the key phrase. It was less important early on when people were not as cost sensitized. It’s one thing when you’re building computers that cost several thousand dollars. It’s another when you’re building a smartphone that only cost $200 or you’re building an embedded device that may only cost $50 or $25. Then it’s a completely different game.
HY: There’s also another interesting thing that has happened recently. If you go back in the early days of say the mobile phone market, a lot of these systems were closed. A particular operating system, a particular manufacturer — it wasn’t really an open market.
With what is out there with Android and the Linux environment that exists, the market is there and as long as there is a good architecture that can deliver the performance and power and cost, amazing things could happen.
The other aspects of this for Imagination is that MIPS is a fantastic architecture, and there are a lot of great people in that company. In the licensing business, stability is very important. Once you have a long-term plan and commitment behind an architecture and there’s no changes and concerns and issues, then things will change.
Existing markets that MIPS has been in, such as network or set top boxes, wherever MIPS was a key driver, those customers will feel once again much more confident about the future and the security of MIPS.
There’s a lot of people out there, their lives and engineering work in some way have been touched by MIPS. They have true love for the architecture. We’re counting on this to allow us to take MIPS to the highest level of success and glory that is possible.
TSD: Do you think attempts at developing the MIPS Architecture are going to happen organically as you try to become more efficient, or are you stepping back to take a direct approach to make bigger leaps and bounds?
HY: From my point of view, both of those things are certainly on the cards. There’s a lot of good stuff about MIPS architecture, one of the amazing things about it is that it’s the first 64-bit microprocessor on the planet. And the world is heading that way fast. So there’s a lot of things you can build on.
But when it comes to low power and efficiency there’s always new things you can add and apply. We certainly plan to build around what is there in terms of the history, knowledge and experience and combine that with what we’ve learned in our work in the mobile space from the graphics or what have you and press on and take this architecture forward.
TSD: Now that the technology is out there, how do you balance turning a profit with the need to keep improving and innovating with the architecture?
HY: We are a public company, and I can tell you that, when you are a public company, you have to keep many people happy, internal and external, and it’s a tough challenge. But my approach to these things is that, at the end of the day, I am a mathematician and a scientist and an engineer.
I’m doing what I’m doing because I think there’s a good chance of success. You have to do the right thing. Of course profit and revenue is important, but not at the expense of doing the wrong thing. You have to have a balance — you cannot jeopardize the long-term future for short-term decisions, particularly in the IP business.
We generally have five-year plans, and one of the big things I’ve started is getting my team together and telling them, “Ok guys, it’s time to fix and finalize the roadmap for the next few years. You have to do that and put the investment in, and that’s the nature of the beast. It’s a long-term business with long-term focus but you have to find ways to keep the money side also delivering.
JH: I think this is true for not just tech companies, but all companies right now. The fundamentally difficult problem is keeping long-term innovation alive and doing that while keeping your current business running, which you have to do. If you’re a public company, the shareholders will desert you quite quickly.
But all companies have that fundamental problem of how do you keep that innovation engine alive because that’s the secret of growth and taking advantage of new markets and new opportunities.
TSD: If you were to go back to when you first started this work, could you have anticipated this level of success with the MIPS architecture? And when you’re looking back 50 years from now, what do you hope people are saying about it?
JH: We knew that we had discovered something fairly fundamental — we understood that. We didn’t understand how all the pieces fit together. The leap generated by the early MIPS architecture the very first RISC architecture, and it was significant enough that you could say this really was a breakthrough.
You never know what the successes will be like. Especially when you’re at a small company, and you’re fighting for your life, and you’re up and down. At one point, the company was down to 10 days of operating capital — this was many years ago in the early ’80s — but with 10 days of operating capital, you have to do something or you’re going to be out of business shortly.
And that’s what its like for startup companies. The fact that the architecture bloomed, it really bloomed with this movement out into the embedded space. And I’m excited to see Imagination pick up and build on this incredible history and take this to the next level as we go forward. I think the world is going to get better, and we’re going to see lots of interesting breakthroughs and uses of this technology, and I’m excited to see Imagination do it.
HY: I think in particularly having been to a couple of exhibitions recently and talking to customers, there’s definitely strong support and interest for the growth and success of MIPS. This is important for Imagination as it is for the industry.
The industry needs alternative CPUs giving choice to people, and we certainly recognize the history of MIPS, and our aim is to develop and safeguard and be a good custodian taking this architecture forward and creating a kind of solution in IP that many customers can use and commit to for the long range.
This interview has been condensed and edited.