Page 45 - EE Times Europe March 2022

EE|Times EUROPE 45

Inference Sensing and In-Memory Computing Chip Startup Claims 20 TOPS/W

Figure 3: Application reference for ADA20X (Source: Reexen)

application scenarios, if high-power–efficiency computing chips are The objective of the second project, IMOCO4.E, is to provide verti-
used to perform such functions as gesture recognition and eye track- cally distributed edge-to-cloud intelligence for machines, robots, and
ing in video processing, the operational duration of AR devices can be other systems with control components. The project’s total budget
increased and the user experience enhanced, according to Reexen. exceeds €30 million.
In summary, ADA20X has the following features: For both projects, Reexen has been participating as a technology
• The computing-in-memory architecture works like the human solution provider. Other EU partners in the projects include research
brain and breaks the system limitation of the memory wall institutions such as Imec (Belgium), Fraunhofer Institute (Germany),
caused by the separation of storage and computing with tradi- and Grenoble University (France); semiconductor manufacturers
tional von Neumann architecture. This is the key breakthrough of including STMicroelectronics and X-Fab; and other European technol-
neuromorphic computing. ogy companies.
• The architecture offers a high power-efficiency ratio because
its power consumption will not increase sharply in tandem with INTERNATIONAL R&D TEAM
the complexity of the application’s computing tasks. It thereby Reexen’s competitive strengths as a startup are its chip technology
addresses the dual problems of high computing power consump- intellectual property and its experienced international R&D team,
tion and heat dissipation. which includes scientists from Imec, ETH Zurich, Peking University,
• It supports both convolutional and transformer neural networks Tsinghua University, and other research institutions, as well as senior
(CNNs and TNNs), as well as neuromorphic computing engineers from top semiconductor manufacturers such as Qualcomm,
architectures such as spiking neural networks (SNNs). The HiSilicon, Samsung, and Intel.
architecture is flexible, and the efficiency of array computing is Reexen’s vision is to Thus far, Reexen has obtained
not attenuated. 30+ patents, published more
• It can support both the event input of dynamic vision sensors become the ADI of the than 20 ISSCC/JSSC papers,
(such as event cameras) and the image input of traditional cameras. AI age. Its inference and accumulated extensive
• Configured with an imaging signal processor (ISP) based on experience in chip design and
in-memory computing, the ADA20X can support event and image sensing and in-memory mass production.
preprocessing simultaneously. The nearly €15 million in
The third series, ADA30X, is positioned as a high-performance SoC computing are a natural Series A funding that closed in
designed to perform multisensor (LiDAR plus vision) fusion signal pro- extension of traditional July will be used primarily for
cessing. It targets AR/VR control, autonomous-driving master control product development and to
below the Level 3 grade, and autonomous-driving secondary control analog and mixed- expand the R&D team. Reexen
above L3. The ADA30X is still in the planning stage and won’t be in signal processing. currently has 60 employees
mass production until the end of 2023. and has five new projects on
the agenda for this year. The
EU COLLABORATIVE RESEARCH PROJECTS planned expansion of its Swiss R&D center will increase the number of
Reexen has a research and development center in Switzerland and experienced European engineers and scientists.
actively participates in cooperative EU research projects. It is currently According to Hongjie Liu, Reexen’s vision is to become the ADI of
involved in two such programs: StorAIge and IMOCO4.E. the AI age. ADI is synonymous with analog and mixed-signal chips,
The StorAIge initiative aims to develop high-performance, ultra- and Reexen inference sensing and in-memory computing are a natural
low–power, and safe SoC solutions to improve the EU’s competitive extension of traditional analog and mixed-signal processing. ■
advantage in AI commercialization for edge applications. The total
budget for the project is estimated at nearly €100 million. Steve Gu is a senior industry analyst of AspenCore China.

www.eetimes.eu | MARCH 2022