40 years ago, the founder of Intel, Gordon Moore, established Moore's Law according to which the number of transistors in a microchip doubles every two years. Until now, that is more or less how the industry operated. The problem today is that the chips have reached a stage where it's extremely difficult to reduce their size any further – the current size of a transistor is close to several atoms, thereby limiting the scope for any future descaling. As a result, progress in microchip technology has slowed dramatically, almost reaching a standstill.
One of those foreseeing this reality already five years ago was the HiPer consortium (High Performance System on Chip). The consortium was founded in 2014 by the Innovation Authority and includes commercial companies and academic research institutes operating together in order to develop new technologies, methods and tools for the development of complex SoC (System on Chip) platforms.
"The consortium concentrates on development in the field called "System on Chip", i.e., not a system with several chips but rather, a single chip containing an entire processing system. A whole computer consisting of a single micro-chip. This system is gaining prominence in today's electronics world and almost every current microchip contains such a system.
The consortium has developed important methodologies and tools for contending with the need for improved performance and shorter developmental cycles of microchips. If it currently takes a company between 1-2 years to develop a chip – subject to the ever-increasing complexity involved – development time can be markedly reduced with the appropriate tools. Such tools have been developed by the consortium – some of which don't exist anywhere else in the world. This saves companies at the consortium a lot of money by enabling them to avoid racing after the newest manufacturing technologies. For some companies, when such saving reaches tens of millions of dollars, it can be the difference between make or break.
All the companies in the consortium develop microchips and therefore almost every tool and methodology developed is relevant for them – even if the companies don't introduce them directly into their product, but rather, into the development processes. Two good examples of the importance of the tools developed in the consortium are that CEVA has already introduced technology developed in the consortium in conjunction with Tel Aviv University into its products and Mellanox (that since its acquisition by NVIDIA is called Mellanox NVIDIA) has applied certain technologies developed at the consortium in its products.
Aner Shoham, software engineer and veteran high-tech professional, is the consortium's technical director and one of the two entrepreneurs behind its foundation. The second, Shai Adar from CEVA Corporation, is the consortium's chairman.
"The fact that we foresaw the technological problems is one of the consortiums successes", says Shoham. "We also understood that solutions existed for some of these problems. We thought that if progress wasn't going to be as rapid as previously, we could forge forward in other ways and not necessarily with the manufacturing technology itself. In other words, the solution will not be to manufacture more compact and faster chips.
"The consortium was established in order to give Israeli companies the possibility of contending with the problems that emerge in the transition to more expensive and more complex manufacturing. This would be done by finding ways to do things differently i.e., to develop more efficient algorithms, as well as faster and more convenient tools specifically with existing manufacturing technologies.
"Let's talk numbers: manufacturing technology is advancing and the further it progresses, the more the numbers representing the size of the transistors decrease. We started out in the consortium five years ago with the development of a 28 nanometers microchip demonstrator. We adjusted ourselves to the developing technology worldwide and are now striving to produce smaller chips. The world is already talking about 7-8 nanometers, and maybe even less. You need to understand that manufacturing becomes more expensive as the numbers decrease. We enable companies in the consortium to delay somewhat the crazy rush to descale and to save a lot of money by doing so.
"How? If for example, you need a computer to operate faster in order to perform a calculation faster than before, one of the options is to move from 28 to 16 nanometers to achieve double the number of transistors and faster processing technology. In other words, I can now put 200 transistors in a space where previously I could only put 100. The problem is that the costs involved in such a transition skyrocket and not every company can afford it.
"What we said was that instead of moving towards a more advanced technology, maybe we can simply change the manner in which we perform the calculation so that it's more efficient and quicker even though we are using the same manufacturing technology. We are, in essence, enhancing the calculation process – thereby attaining significantly improved performance for the required space and output, and without changing existing manufacturing technology.
"That is exactly what the consortium is doing: enhancing performance - not by moving to a newer manufacturing technology but rather, by using more efficient algorithms in many cases and by using groundbreaking technologies. One good example is a technology called Dual Mode Logic (DML) that was invented by Prof. Alex Fish from Bar-Ilan and developed at the consortium. This technology enables better performance while preserving silicon space and saving energy.
Against all Odds
One of the consortium's achievements is the foundation of Alex Fish's special laboratory at Bar-Ilan University. This is how Prof. Fish describes the events: "The semi-conductor industry was the flagship industry of Israeli high-tech for many years, to such an extent that the presence here of all the important global corporations and development centers of companies such as Apple and Intel is taken for granted.
"In recent years, the challenge of this illustrious and knowledge-intensive industry was to establish Israeli-owned startups so as to avoid dependence solely on foreign corporations.
"As technology continues to advance - we are now talking about mass production at technology of 7 nanometers and less – the complexity of microchip design is constantly increasing. This requires enormous knowledge, effort and manpower and therefore production costs now total tens of millions. The upshot is that there are less startups and existing companies need to compete head-on with global corporations possessing almost endless resources. The key therefore is to be super-creative and innovative.
"This is where the Innovation Authority comes in. The Authority views the field of microchip development as being of great importance and is attempting to identify the challenges involved. The Authority has established three consortiums in this field within a short period of time. The demand from the HiPer consortium was clear: extremely high innovation. Without this, it will be difficult for the industry to compete internationally.
"The situation I described in the semi-conductor world poses a tremendous challenge for academia, one that requires an engineering effort it cannot afford. There are only a few academic groups in the world that can design micro-chips on the scale of the leading commercial corporations.
"At HiPer, we decided from the outset to construct a SOC lab that would enable us to contend with complexities and complications. A SOC lab was established as part of the EnICS laboratories at the faculty of Engineering at Bar-Ilan University. EnICS labs operate according to a special model whereby there is extremely close cooperation between four academic groups: the groups share common infrastructures, the equipment belongs to everyone, and all the students sit in the same laboratory together. The groups' financial decisions regarding the purchase of equipment and budget management are made jointly, in full agreement, and efficiently. The new SOC lab was established as a joint infrastructure of EnICS researchers with the immediate aim of advancing the HiPer consortium.
"The goal was to bring to the SOC lab an experienced team of engineers who weren't an ever-changing group of students. In other words, experienced engineers who made it possible for us to bridge the gap between idea and practical implementation. This is a professional team that knows how to design chips using the most advanced technologies available.
"The HiPer consortium taught us how much the Authority's expansion into additional fields can leverage Israeli industry", explains Moshe Avrahami, Director of the Industrial Sector in the Innovation Authority's Technological Infrastructure Division. "The lack of high-quality manpower in chip design is an example of the shortfall we were exposed to thanks to those academic researchers. These skilled workers were trained during the development process itself and were pounced on by the industry. We are currently completing a training program for high-quality workers as part of the MAGNET consortiums in order to meet this need."
"Because this team of experts is not constantly changing, the knowledge stays put. We have succeeded in recruiting people with tremendous experience and formidable reputations and "raised" others by ourselves. Everyone works full-time for us enabling us to achieve exceptional things that no other academic group can achieve. All this involves very high costs so we couldn't have initiated the project without the Innovation Authority. Against all odds and in face of all the risks, they believed in the idea and in our dreams and such faith is not trivial.
"The laboratory concentrates the activity of the entire consortium. We take the most interesting technologies developed within the consortium's academic and industrial teams and put them into a technological demonstrator. In other words, a chip that is a kind of platform operated by the central laboratory.
"The objective is to develop generic technologies that are directed towards high performance and towards attaining performance improvement, design efficiency etc. These generic technologies we have developed do not have a single specific designation – they can be used for a range of needs by all the consortium's companies", emphasizes Prof. Fish. "The presence of a single entity enabled us to overcome the need for each company to work separately on something that served only their specific needs. The central entity aims to take 4-5 very interesting technologies and demonstrate them in a complex generic chip. In my opinion, this could not have happened without the Innovation Authority - a central, neutral entity devoid of commercial interests that focuses on advancing knowledge and science.
"At the outset of the consortium we recruited young engineers, trained them, and with the help of Ceragon Corporation, a participant in the consortium, we produced the first demonstrator – an experimental chip that proves the capabilities of the innovative technologies we developed.
"Thanks to the cooperation of all the consortium members, we are currently planning the second demonstrator with extremely advanced 16-nanometer technology upon which there are some very interesting new technologies which we have developed. The lab is responsible for all stages of the design and implementation from A to Z. This demonstrator places us in a completely different position as a professional entity compared to where we were at the outset of the consortium – from the perspectives of knowledge, manpower and capability.
"What's next? The Authority has recently established a new consortium called GenPro with the aim of developing an Israeli processor based on RISC-V technology. Our lab is also responsible for the architecture of one of the labs in this consortium, for coordinating collaborations and for designing the demonstrator. When the consortium closes, the lab will be a knowledge center capable of providing assistance to other labs throughout the country."
Merging the Logo
One of the partners in the HiPer consortium was Mellanox that creates microchips for a unique communications protocol. Elad Mentovich, one of the company's senior developmental engineers, lists four major advantages in operating within the framework of the consortium:
"The first advantage is the option of bringing together different companies that contribute knowledge and skills and that don't compete between themselves because they operate in different markets."
"The second advantage is in the establishment of the national laboratory at Bar-Ilan University as the focal meeting point of all the different companies. In this way, the companies work together around one laboratory that receives knowledge from each of them and which builds something better than each individual company could have constructed separately, thereby realizing the true potential of each company.
"The third advantage is that this is a superb platform for checking new technologies. It is very expensive to build new technologies, and when it's done with state cooperation, we can create technologies that might serve us three or five years from now.
"The fourth advantage is the ability to add functionality to the company's current microchip, something that saves the need to invest large sums in purchasing expensive technology.
"The consortium has joined the global trend of developing an entire "on chip" system, explains Mentovich. "The objective was to make a real, functional and useable chip – the first advanced microchip produced by academia. There are several challenges in the new technologies that we defined together at the consortium: new forms of logic, different types of memory, new methods for accelerating materials within the chip, and a simpler, more generic connection between hardware and software.
"The chip can be viewed as a piece of Lego with each of the partners making an effort to place his piece so that it connects with the others. Each partner develops a block within the large chip and the lab at Bar-Ilan integrates between them all.
"The perception is that if we succeed in creating cooperation between industry and academia, then if you are a company that doesn't want to construct all the Lego from scratch, but rather just add your valued input – you have an easier starting point.
"The consortium is very successful. No one believed that we would even succeed in producing the first chip, but we did, and it worked. We achieved new technologies and now the aspiration is to integrate everything together.
"If in the past the perception was that such a consortium stood no chance due to the risk and difficulty involved, we have proven that it's a great success and that a microchip does not necessarily have to be the result of one company's exerted efforts, but rather, a result of cooperation with academia. And now the lab exists on its own right.”