Peggy Aycinena is a freelance journalist and Editor of EDA Confidential at www.aycinena.com. She can be reached at peggy at aycinena dot com.
C-Sky Microsystems: Big Dreams and a 100-year Vision
October 5th, 2017 by Peggy Aycinena
Founded in 2001, C-Sky has “developed 7 types of embedded CPUs covering a wide range of embedded applications including smart devices in IoT, digital audio and video, information security, network and communications, industrial control and automotive electronics. It is the only embedded CPU volume provider in China with its own instruction set architecture, the Yun-on-Chip architecture developed in conjunction with Alibaba.”
C-Sky is a growing IP company serving an enormous market. I spoke recently by phone with Dr. Xiaoning Qi, CEO at C-Sky, who was in California attending meetings. No stranger to Silicon Valley, he previously served at Intel, Rambus, Synopsys, and Sun, after completing his Ph.D. under Prof. Robert Dutton at Stanford.
WWJD: How has your trip gone so far?
Xiaoning Qi: Very well. We presented our technology for IoT security at the IoT Device Security Summit [co-located with 2017 IoT DevCon]. I also attended the GSA Executive Summit on Tuesday in Menlo Park.
WWJD: What have you learned from these meetings?
Xiaoning Qi: From comments at the GSA meeting, where there were talks about AI, I learned that 5G is very exciting. And the keynote at the summit talked about opportunities in the future in China. It’s a particularly interesting world right now.
WWJD: How is C-Sky doing?
Xiaoning Qi: Very well. We received investment from Alibaba early last year, although we’ve been working on our R&D for a long time, our fundamental technology.
Right now, we continue to work on R&D, but at the same time we are going out and letting the company be known to the world. We are also ramping up sales and marketing, helping our customers realize their goals in the IoT market.
Additionally, yesterday in Beijing, we released the first secure NB-IoT [Narrow Band Internet of Things] chip, designed by ZTE and manufactured in China by SMIC.
WWJD: Has there been good response to that news?
Xiaoning Qi: Yes, and there was a release ceremony in Beijing. My CTO was there. If I hadn’t had a meetings previously scheduled here in California, I would have been there as well.
WWJD: Any announcements at DAC in Austin?
Xiaoning Qi: At DAC, I was on a panel where we talked about future of the IoT and semiconductor markets in the US, China, and in the world. It was very informative.
WWJD: What is the future of the semiconductor market in China?
Xiaoning Qi: China consumes currently at least one-half of all semiconductor products in the world. In the future, that percentage will go up and markets there will continue to grow. The IoT market in China is booming, because the population is there to support it.
WWJD: Do you think IoT will continue to depend on the older nodes?
Xiaoning Qi: Yes, most IoT will stay in the older technologies because of the cost. If you look at sensors, for instance, they are selling at a very cheap price. The [people who use them] can’t afford to use advanced technology which is much more expensive.
[Having said that], a customer just developed a chip for a surveillance camera using our CPU with a vision processor and manufactured at 28 nanometers. But for most IoT devices, the older nodes are [sufficient].
WWJD: Would edge devices be better at smaller technology nodes?
Xiaoning Qi: When those nodes become cheaper, it will be more cost-effective and that’s when people will move into those areas. If Moore’s Law goes forward, the density of transistor doubles and then why not?
WWJD: Is the IoT market just one market, or is it harder to describe than that?
Xiaoning Qi: [laughing] Yes, it is hard to describe. It actually [consists of] lots of future markets, although we are not quite sure about the profit part.
For our company, there are two things of interest. We’ve been involved in the traditional embedded market: smart cars, surveillance, printers. We are [addressing] surveillance with our CPUs, vision and video, and how to enhance edge devices for AR.
We are also working with Alibab on IoT for the future: How to connect the Cloud and edge devices, how to get analytics safely and efficiently from a powerful Cloud to the edge devices. All of these things are in the early stages for us, and we are working hard.
WWJD: Why is there any doubt about the profitability of the IoT market?
Xiaoning Qi: I believe there are two things that may [cause concern].
The IoT market spans everything from industrial to medical and consumers – all sorts of areas – but these areas are very much fragmented. They ‘inherited’ that situation from the embedded market.
The profitability of each market is actually very low, and the lifetime of the products is very short, so the chip designers are not quite sure they want to get into this market.
Typically for chips, you need high volumes [to succeed], but IoT chips are only [being produced] in the tens of thousands, which is quite small.
So the IoT market is fragmented, small volume, and short life cycle. Chip designers are not happy there.
WWJD: Will the IoT market be slow to develop as a result?
Xiaoning Qi: Yes, it will be slow. But from the technology perspective, it will move quickly.
First, we are trying to build an ecosystem, thinking about hardware and software together. As hardware providers, CPU providers, we cannot work on hardware alone. We need to work on the two together, we would like to build an ecosystem with our CPUs and our OS.
In the Wintel world, it is Intel and Microsoft. In the embedded world, it is ARM plus iOs, or ARM plus Android.
In our world, we [are working towards] Alibaba and our embedded OS, coupled with our CPUs. We would like to build such an ecosystem, at least for the Chinese market. From our perspective, we would then be able to build a good IoT market.
Second, from a technical perspective, we are building a unified architecture for the IoT market. We have the basic hardware and have built an abstract level for this hardware. Above that, there’s an operating system. We then build another layer of abstraction above the OS, and then the applications for the [various] domains.
The engineers or developers for a specific domain don’t need to understand the basics of the embedded OS. They just see the unified hardware and software platform. This is how we are trying to solve this ‘fragmented market’ problem.
WWJD: Hardware/software co-design has been a challenge for a long time.
Xiaoning Qi: Yes, but in EDA when we talk about hardware/software co-design, it’s different from what I’m talking about in the context of embedded IoT. Here it’s the more fundamental basics of operating system and computer architecture.
WWJD: I would argue they are the same, an attempt to develop the hardware in conjunction with the software to create a seamless end product.
Xiaoning Qi: Yes, that’s a good point.
And it’s from that perspective that we have worked closely with the Alibaba team. We own the architecture and know how to fit this into the operating system. They understand how to change the OS part to fit into a particular architecture. People working together like this can make a very seamless integration.
For us as a CPU IP company, half of our engineers are on the software side, developing design tools and a design environment – all of these to build the expertise for our hardware development.
WWJD: It seems very exciting.
Xiaoning Qi: Yes, it is.
And on the business side, we also would like to think like a software company, something we are learning how to do.
WWJD: How does a software company think?
Xiaoning Qi: The software guys can make a change very fast, and then try different solutions. If one doesn’t work, they fix it and go forward until it is released.
For hardware, [the process] is different. We have to keep up a fast pace as well, but the silicon takes a longer design cycle. We would like to [be able] to adjust the hardware and then tune the software at the same time.
When we do silicon design for the IoT market, for example, we need to provide value beyond the silicon. We need to provide the software stack, and ultimately for the IoT market, provide a solution and a service.
This is how it works for us when we collaborate with Alibaba, an online company with huge amounts of data. When we have the good IoT devices [we are developing], we can work on software integration through the Cloud. The Cloud will then be much richer, and will enable us to bring more customers to our products.
WWJD: What is the goal of a Cloud-based system connected to a billion edge devices?
Xiaoning Qi: To be able to send all that data into the Cloud, and to have access to huge analytics, to have the analysis done in the Cloud.
For example, for the Smart City, we need lots of sensors, lots of traffic information, sent to the Cloud. Then we need to process that data and make a proposal as to which road is best, which traffic lights to adjust, and eventually which roads to build or make wider.
WWJD: These billion edge devices will allow us to optimize the Smart City?
Xiaoning Qi: If the edge device is strong enough, the response will be that much faster. Strong edge devices process data without much latency, and react much faster that way.
For drones, for cars, for all these applications, using strong edge devices is the solution.
WWJD: That brings up the age-old debate: Strong edge devices represent decentralized intelligence and control, versus centralized intelligence and control with ‘weak’ edge devices.
Xiaoning Qi: Yes, but embedded CPUs in printers, for example – versus those in Smart Cars – do not need to be so intelligent.
They don’t need vision processors or facial recognition. They [represent] traditional embedded processors, which is also true for industrial controllers. Manufacturing machines can be controlled using traditional embedded processors.
WWJD: Which edge devices need to be very intelligent, to be very strong?
Xiaoning Qi: The vision processors in autonomous driving systems.
WWJD: So strong that the car can make its own decision without consulting with the Cloud?
Xiaoning Qi: That’s right.
Last year at a GSA meeting, there was a talk from a company that can collect data from the car to the Cloud, and back again. They claimed they had the latency under control. I thought it was good technology, but sometimes you have to make a decision in a tenth of a second, whether to hit another car or hit a pedestrian that is suddenly in front of you.
WWJD: Historically, we have relied on intelligent human reflexes to make that decision.
Xiaoning Qi: Yes, but the machine will soon be smarter than that.
WWJD: On the business side, where are the bulk of your customers?
Xiaoning Qi: They are mostly in China.
WWJD: Why belong to the ESD Alliance?
Xiaoning Qi: It is very helpful for us to know the world.
And we enjoy it because the organization used to be for EDA only. I worked with Synopsys for a number of years, and my adviser was Bob Dutton, so I was there when it was EDAC. I was always amazed at an organization where I could meet visionaries and many interesting people.
Recently, they added IP into the vision of [what EDAC was] and ARM joined. So for us as an IP company, it was also important to learn. We would like to learn from the big guys in the group.
WWJD: Are there topics you would like to see the ESD Alliance address?
Xiaoning Qi: Yes, we would like sessions with more information about IP.
I did attend an IP industry event, which was good, with good speakers. The exhibition was good and the vendors, we all had a chance to talk with each other.
Also, I have been able to meet some people in China working in IP. Bob Smith [Executive Director] has asked me to broadcast to these people the value of belonging to the ESD Alliance, and I said I would be happy to help.
WWJD: How many IP companies are there currently in China?
Xiaoning Qi: There are four.
WWJD: Isn’t it interesting to see Synopsys and TSMC emerge as major players in the IP market, companies that are not just IP companies.
Xiaoning Qi: In EDA, the profit margin in high. But when you compare EDA companies to the semiconductor industry, EDA is only $5 billion in total.
That is so small, companies like Synopsys, and TSMC, made a good move to go into the IP world, because there is more value there than just in tools. Chi-Foon Chan and Aart de Geus made the right decision there, because currently IP is half of Synopsys’ revenue.
For the foundries to go into the IP business also makes sense. In manufacturing, they need to have standard cells and libraries, so they can provide IP to their customers.
This puts the foundry in a better position to help their customers, because the IP would already be proven, for instance, in a TSMC process technology. I think now SMIC is doing the same thing.
WWJD: Also a business issue, how do you find new employees?
Xiaoning Qi: That’s actually an interesting and touchy question. We are a small company and have to compete with large companies for employees, competing in salary and life-work balance.
Typically in small companies that don’t have a lot of people, employees have to wear a lot of hats, but we have big dreams. We have to compete with Huawei and Alibaba for employees. But people come to work for us not just for the salaries, but to pursue the big dream. Yes, it’s a challenge to recruit.
WWJD: Do you ever recruit out of Stanford or Berkeley?
Xiaoning Qi: Right now, no. But some people studying here would like to work in China, would like to move back to work there.
WWJD: After your long years working here, have you moved back to China?
Xiaoning Qi: Yeah, I spend most of my time there now, although I’m also traveling a lot.
WWJD: Do you recall something funny that has happened to you in your career?
Xiaoning Qi: Not so much funny as interesting, perhaps. When I was an EE student at Stanford, electrical engineering jobs were very high profile and interesting.
But the semiconductor industry peaked in 2000, and although it was still very exciting to work in the hardware area, people found out that the industry peaked and they graduated started talking about it as a sunset business, one that was [in decline].
But I was always kind of optimistic, so I was thinking something must come along to rescue us. So then there was the IoT market, and now here we are.
WWJD: Your optimism seems to be an important part of your business model.
Xiaoning Qi: [laughing]. It think that’s right.
Since I was Chairman of the Chinese American Semiconductor Professional Association, Chenming Hu of Berkeley has been our honorable adviser. He told me that there are at least 100 years left for the semiconductor industry, and when I asked him if he was serious, he said yes.
In a similar way, when I was at Stanford some people thought the semiconductor industry would become the next steel industry. There is no more need for the technology itself to develop further, because we know that as a lot of industries becomes mature, we can still use the technology to build things. Applications are very important.
Right now, semiconductors are everywhere, in your home, in your pocket. The semiconductor industry is mature, but with a long future.
Subhasish Mitra is a professor at Stanford. We graduated the same year and are very good friends. He recently published a paper in Nature that looked at the whole [ecosystem] in semiconductors. He said there, we don’t need to have 7 nanometers, or 5 or 4, because memory is the problem.
He said CMOS [for logic] and nanotubes will solve the memory problem, so he is working on the more fundamental technologies. It’s very exciting. When I saw him recently, I said I would like to be one of his graduate students and be doing this work.
WWJD: Given your background, why do an IP company, and not an EDA company?
Xiaoning Qi: I know EDA, because I worked for some time on that side, and I know why people want to develop a tool that is more interesting. If you do EDA and make a tool, design is easier.
But the fact is, we really need a visionary in the EDA industry to come out with a new business model that will change the situation in EDA, moneywise.
I once asked Wally Rhines why EDA revenues are so small compared to the semiconductor industry. He said it’s because anyone with a computer and a brain can start an EDA company, but you need much more to do a semiconductor company.
So again, while I was looking at EDA and the people who design the tools, I was thinking to create something in terms of design itself. My thought was that working on the design would be more exciting.
Also, to do an EDA tool you can be a computer science guy, but I studied electrical engineering. An EE understands which devices you want to optimize, and although people with CS backgrounds may be very good at algorithms – speeding up calculations and simulations – to solve the right problems in design, you need an EE background.
Designers get my utmost respect, which is why I decided to work in IP.
Xiaoning Qi is CEO of C-Sky Microsystem Corp. Previously he held senior management and technical positions in Intel, Rambus, Synopsys and Sun Microsystems in California, working on IC devices, micro-processor design, and platform electrical design for semiconductor systems.
Xiaoning Qi is a member of Asia-Pacific Leadership Council, Global Semiconductor Alliance (GSA), and was chairman and president of Chinese American Semiconductor Professional Association. He has published more than 50 technical papers, one book, one book chapter and eighteen invited talks, and holds two US patents. He is also a reviewer for numerous IEEE journals and international conferences, and a senior member of IEEE.
Xiaoning Qi received his Ph.D. degree in Electrical Engineering from Stanford University in 2001.
Tags: Aart de Geus, Alibaba, ARM, Asia-Pacific Leadership Council, Bob Smith, C-Sky Microsystem Corp., Chenming Hu, Chi-Foon Chan, Chinese American Semiconductor Professional Association, ESD Alliance, GSA, IEEE, Intel, Mentor Graphics, Microsoft, NB-IoT, Rambus, Robert Dutton, SMIC, Stanford, Subhasish Mitra, Sun Microsystems, Synopsys, TSMC, Wally Rhines, Xiaoning Qi, ZTE