In an in-depth interview, Henry Samueli, Broadcom's co-founder and CTO, predicted the future of our multi-gigabit Wi-Fi, LTE (4G) and home broadband. Moore's Law is a bigger challenge, but Broadcom is also planning a high-end CPU.
It's time to focus on Bontone CTO Henry Samueli. Over the years, CPU manufacturers like Intel, IBM and Apple have won many honors in the processor industry. But with the development of mobile technology and cloud computing, communication chips are becoming more and more important. Broadcom is the leader in the communications chip market.
On Wednesday (12.4), Broadcom announced a new processor, symbolizing a new trend in the computer industry. The small new chip BCM20736 allows different devices to talk via wireless Bluetooth technology and support wireless charging. The chip's route is the two hot new trends: wearable computing and the Internet of Things.
The first category of wearable devices are fitness equipment, Google Glass and a wide range of smart watches. The second type is to get the network out of the PC and mobile phone, connected to other items, such as cars, thermostats, indoor plants, smart power meters, right, and coffee pots.
Samueli founded Broadcom in 1991. At that time, wireless communication was still rare in the industry, and consumers were unheard of. Now, electromagnetic waves carry more than just data, but they can carry electricity. But wired communication is still important throughout the Internet.
Stephen Shankland of CNET, a technology information network, interviewed Samueli, who talked a lot about his own vision of the future of communications technology. The following is an interview record (with deletions) covering wireless charging, LTE mobile networks, and Gigabit home broadband.
Q: You have a Bluetooth chip that supports wireless charging. In what direction do you think wearable computing is moving? What is the development direction of the Internet of Things? It is often difficult to tell the hype from reality.
Henry Samueli: There are a lot of advertising, but there is no reason. The wearable market is indeed a very exciting market because it really covers hundreds of markets. When you talk about the Internet of Things, you cover hundreds of industries. The core of the Internet of Things is communication. That's what we offer, a communication chip that consumes very little power and is cheap. You can embed it in whatever you want to create. The beauty of the IoT market is that the threshold for entry is quite low. You don't have to be a big company. Thousands of startups will be formed to deploy such products in the future.
Buying this chip from Broadcom is equivalent to buying a simple solution. Our entire product line, WICED, wireless Internet connecTIvity for embedded devices, is a very easy to use platform that allows you to connect software to any sensor you might find. Whether it's an environmental sensor, a medical sensor that measures body functions, or a home security sensor, you need to connect it to a communication device. Then it will be networked.
It takes a lot of thought to lay a layer of sensors on the creation. If you want to connect all the roads, traffic lights, street lights, houses and power meters, it will be a big investment. Where can this product be popular? At the beginning, what market is financially justified?
Samueli: The infrastructure still takes time, and this product takes time. Each power meter will be a smart power meter. Sensors are everywhere in this environment. The short-term goal is consumer products that people will buy from Amazon. We started to focus on health sensors. Not just fitness. Now, you are wearing a smart bracelet to detect your steps. In the end, you will have more and more complex sensors to detect your heart rate or other body functions.
Smart watches are another interesting opportunity. We have seen many big companies and startups spend their time on smart watches. Smart watches can replicate many of the features that can be found on a smartphone. Will the market for smart watches be as big as a smartphone? Probably not, but the smart watch market may still be a huge market.
At home, we are beginning to see more and more home appliances that support wireless connectivity. Smart thermostats, smart smoke detectors and all flat-panel displays talk to your tablet and smartphone and connect to a complete multimedia device network at home.
All these devices are charging? You are using the A4WP (Alliance for Wireless Power) standard. Can you use this to charge all devices? Today's wireless charging technology is a very short-distance charging technology: you have to put your device on a charging pad.
Samueli: Now, yes. But people are eagerly awaiting wireless power over longer distances. You may be able to wirelessly charge your device longer. This is still in the experimental stage, but in a few years, you may really be able to do so. I think this is how you end up charging many devices. Or if you can reduce the power consumption enough so that a button battery can be used for a year, you probably think this is fine.
You registered the A4WP standard. Wouldn't it be better if we only had one wireless charging standard?
Samueli: We have to adopt a standard. There are currently three major wireless charging alliances, and they are actively negotiating with each other to seek the possibility of convergence. If you don't have a common standard, you can't create a mass market. If you put your device on a charging pad and you can't charge it because it doesn't match, you might be upset. Integration is a must, and I think there will be a unified standard in the future.
Why did you choose A4WP?
Samueli: We are very concerned about consumers, and consumers are pushing us to choose this standard. Adopting other standards, we can do it very easily. In terms of core technology, the three standards are similar, and you can even make a multimode chip to fit them at the same time.
The idea of ​​energy harvesting technology is to use human energy, heat, solar energy or radio energy around the device. Can this charging method be sufficient to power a useful device? Or is the collected current not too small?
Samueli: The energy collected can only support a very low level sensor. I am sure there will be some appliances using this energy harvesting technology, but I don't think that will become the mainstream market segment.
I am curious about what you think about LTE mobile wireless networks. You acquired LTE assets from Renesas Electronics. In the LTE market, you have not been in the front, this time is to catch up?
Samueli: The acquisition ended on October 1. The move is to accelerate our development in the LTE market. The Renesas team was originally a Nokia modem team and began researching LTE in 2002, 2003 and 2004. Their technology has been developed for more than 10 years, so their technology is very mature and qualified and proven. We are finally excited to enter the LTE market.
Who can catch up with Qualcomm? Now that you have a second supplier, what will your marketing strategy be like? Or is it more active?
Samueli: It's important to have a lot of vendors on the market that offer this technology. When you have such a large market, consumers will seek multiple resources for their equipment. Entering this market is a big opportunity. This means we will have a complete solution, not just LTE. Broadcom is a leader in connectivity technology, Wi-Fi, Bluetooth, GPS and NFC (Near Field Communication). Combining LTE baseband will allow you to build a complete solution platform for your customers. This will put you in a very important position. We believe that we have the ability to make great achievements in this market.
Why do you want to buy to get you to this step? LTE has been around for quite some time, and Broadcom is at the heart of wireless communications. Why not develop independently in the company?
Samueli: LTE is growing faster than most people expect. From the moment it was announced, it quickly got on track. We did have an internal development roadmap, but we just need to speed up its development.
As far as I understand, with LTE, if you want to truly achieve high-speed wireless data transmission, you have to use a higher frequency band. In order to achieve this, you have already placed your base stations close to each other because the high frequency signals are attenuated faster and the transmission distance is shorter. In order to realize the potential of LTE, will wireless network operators need to build more new towers?
Samueli: Our plan is extended to the next decade. Over the next 10 years, the existing infrastructure will provide us with adequate service, as the upgraded version of LTE or LTE can transfer data to hundreds of megabits per second. But if you look at it for more than a decade, then you need a transmission speed of several gigabits per second, then you need a smaller base station, so you need to expand infrastructure on a large scale. .
You also do network equipment for fixed-line communications, not just wireless. What do you think about fiber optics connected to the family? You see Google is pushing Google Fiber. Will this be bigger and stronger? Or is it cost-effective?
Samueli: Yes. The driving force comes from the demand for services from home consumers. You should look at the Ultra HD (high resolution 4K video) you just deployed. You'll get faster digital TV data transmission and the high-speed Internet connection consumers want. You will need a gigabit data modem - a cable modem, a DSL modem. Fiber optic transmission and passive fiber optic networks are part of it. Broadcom is a market share leader in those markets. Bringing higher-speed data transmission to the home is because consumers must have this demand.
What do you think of the next 10 years? How many American families will be hooked up with fiber? I think the ratio is still relatively low. I am surprised how much people can squeeze from the copper wire.
Samueli: The existing infrastructure is mainly a hybrid fiber-optic coaxial cable network. The cable company first converts the program signal into an optical signal and then transmits the optical signal to a node with an optical fiber. The node then converts the optical signal into an electrical signal for transmission through your home coaxial cable. Consumers only see coaxial cables. This hybrid structure will last for a long time, certainly for at least 10 years. The penetration of the fiber will still be low for some time.
How fast can you transfer coaxial cable infrastructure data?
Samueli: Now we are working on the next generation DOCSIS standard - Data Over Cable Service Interface SpecificaTIon. DOCSIS 3.1 is being defined. That would allow data to be transferred to the home at 5 to 10 gigabits per second. Such high-speed data transmission will easily meet the needs of consumers for the next 10 years.
What restrictions will be encountered on that?
Samueli: Very few. We are designed to be compatible with existing coaxial cable infrastructure. All you need to do is upgrade the cable head (the cable company's network equipment) and use a new box that supports DOCSIS 3.1. Then you will have a DOCSIS 3.1 cable modem and you will have 5 gigabits per second of data transfer speed. This will happen soon, and within three years, you will start to see that it will be deployed in a fairly meaningful way.
What will benefit from 60GHz Wi-Fi technology like 802.11ad? Or is it not so practical? After all, ultra-high frequencies mean that the signal distance is very short.
Samueli: This has to be weighed, whether it is signal distance or data transmission speed. If you really want to download a file at 5Gbps at an ultra-fast speed and only need a signal distance of 1 meter, then you might use 60GHz technology, so I won't give it up. There may be some interesting applications that require very fast data transfer speeds and short-range wireless technology. I think 60GHz technology has potential for the next three years.
Do you have any projects that support 60GHz technology?
Samueli: We have internal research and development work on 60GHz technology. We are developing some related products.
Are you planning to bring those products to the market in the next few years?
Samueli: Yes, of course.
You have been doing communication chips and radio chips all the time, but there are also many application processors and CPUs outside. Now that you've recruited customers who buy a large package of integrated chips, what are you doing on the mobile phone application processor?
Samueli: Today we are one of the world's largest application processor shippers. The difference is whether you see the span of the application processor, you have very high-end thin chips, and then you have many mainstream 1.5GHz processors, which are the mainstay of shipments. Then you have some low-end CPUs below 1GHz. The market we choose is the mainstream large capacity segment. In fact, all of the mobile SOCs we ship today (system-on-a-chip processors that integrate multiple functions into one package) are equipped with dual-core or quad-core ARM CPUs from 1.2GHz to 1.5GHz. Very high-end 2GHz...
Like Qualcomm Snapdragon 800s.
Samueli: Right . That is a lower capacity segment. As time goes by, we will further resolve those high-end parts. But now we have carefully considered and decided on the capacity positioning.
What is the best way to predict Moore's Law and keep your computer performance better? When you want smaller transistors, you run into obstacles. What would you do to solve this problem, photonics, spintronics, III-V semiconductor nanomaterials? Following the traditional silicon CMOS (complementary metal oxide semiconductor), what is the next step?
Samueli: I am as accurate as you are. Nobody knows. This is where the problem lies. CMOS has been surprisingly used for 50 years, hitting the ring, and then expecting another technology to replace it and continue to run another 50 years, which is an optimistic idea. My personal opinion is that this will be a difficult process. We will do our best to promote Moore's Law. We will extend it to another 15 years or so, but no one has come up with a viable technology to further expand the transistor. We really might walk into a dead end. In the next 50 years, you may not see such scaling, as we have seen in the past 50 years. We will be faced with the basic physical limits to slow down the performance of the computer.
We have seen some limits - not Moore's Law. Strictly speaking, that only controls how many transistors can be placed on a silicon chip. What is to say is the limit you are trying to make the performance of the computer exceed. Power constraints have pushed chip makers to very multicore designs. Do you think software will keep up with the multi-core world? Or, fundamentally, that is the tricky problem most programmers have to face?
Samueli: I don't think this will be an insurmountable challenge. Every time you put a challenge in front, people will come up with creative solutions in software. I agree to use the multicore approach to handle the power limitation issues encountered on the CPU. Therefore, you will see more mainstream multi-core chips, and even mobile phones will have quad-core processors. People are figuring out how to use very efficient compilers and load balancers to distribute processing across multiple cores.
You announced the ARMv8 server processor (the technology is licensed by ARM Holdings, UK). Will ARM make a difference in the server market? You are paying attention to network functions. What is your prediction? Is there any room for the server market to be included in Intel?
Samueli: If you look at our portfolio, we will find that we are more concerned with the segment of the communications processor than the pure server market. It is still an infrastructure market, but this infrastructure refers to the infrastructure that runs communication processing, such as deep packet inspection and security processing. Today we ship MIPS-based processors and we have a portfolio of MIPS-64 custom processors we designed. Recently we announced the ARMv8 version of the custom architecture we will be developing. When we introduced the product to the outside world, we had the highest performing ARM CPU in the world. We are talking about a 3GHz class CPU with 16nm CMOS manufacturing technology. This will point to the communications processing market, where we have the best foothold and ability to penetrate the market, and since we acquired NetLogic, we have been doing well in that area.
Can some of these products be applied to the server market? can. Maybe we will expand this market over time. It depends on whether people will successfully fight ARM in the server market.
Optical Socket Panel, also known as optical fiber panel or information panel.The main function of the panel is to fix modules and protect cables at the information outlet.Optical Socket Panel is widely used as a termination point for the feeder cable to connect with drop cable. The fiber splicing, splitting, distribution can be done in this box, and meanwhile it provides solid protection and management for the FTTx network building. It can be used indoor or outdoor.
Optical Socket Panel
Optical Patch Panel,Fiber Optic Patch Panel,Fiber Optic Panel,Optical Socket Panel
Chengdu Xinruixin Optical Communication Technology Co.,Ltd , https://www.xrxoptic.com