Open side-bar Menu
 EDACafe Editorial
Roberto Frazzoli
Roberto Frazzoli
Roberto Frazzoli is a contributing editor to EDACafe. His interests as a technology journalist focus on the semiconductor ecosystem in all its aspects. Roberto started covering electronics in 1987. His weekly contribution to EDACafe started in early 2019.

IP quality assurance; PyTorch-to-RTL; Risc-V growth; new semi subsidies in China; Google’s human brain mapping

May 28th, 2024 by Roberto Frazzoli

Governments around the world keep subsidizing their respective domestic semiconductor industries, with the most recent announcements coming from China and South Korea. Meanwhile, artificial intelligence is accelerating human brain research, with Google spearheading this effort. But first, a few EDA-related updates.

Siemens’ IP quality assurance solution

Siemens Digital Industries Software has introduced Solido IP Validation Suite software, an automated signoff solution for quality assurance across all design intellectual property types, including standard cells, memories and IP blocks. The suite – which includes Siemens’ Solido Crosscheck software and Solido IPdelta software – aims to shorten the time-consuming task of validating IP across all its design views such as logical, physical, electrical, timing, and power analysis contexts. It also provides version-to-version IP qualification for more predictable full-chip IP integration cycles and faster time-to-market.

Siemens’ PyTorch-to-RTL solution for AI accelerator design

Siemens Digital Industries Software has announced Catapult AI NN software for High-Level Synthesis of neural network accelerators on ASICs and SoCs. Catapult AI NN starts with a neural network description from an AI framework such as TensorFlow, PyTorch or Keras, converts it into C++ and synthesizes it into an RTL accelerator in Verilog or VHDL for implementation in silicon. Catapult AI NN brings together hls4ml, an open-source package for machine learning hardware acceleration, and Siemens’ Catapult HLS software for High-Level Synthesis. Developed in close collaboration with Fermilab, a U.S. Department of Energy Laboratory, and other leading contributors to hls4ml, Catapult AI NN enables AI experts to develop PPA-optimized accelerators for different applications without requiring them to become ASIC designers.

Risc-V updates

Andes, HiRain, and HPMicro will cooperate to build a Risc-V Autosar software ecosystem. The collaboration will combine the AndesCore Risc-V processor series, the HPMicro HPM6200 full line of products, and the HiRain Vehicle OS software platform solutions to jointly build the Risc-V ecosystem in the field of automotive electronics.

Esperanto Technologies, a developer of AI and HPC solutions based on the Risc-V instruction set, has signed a Memorandum of Cooperation with Japan’s Rapidus. The initial focus of the partnership is to enable future semiconductor designers to develop more energy efficient solutions for AI inference & HPC workloads for data center and enterprise edge applications.

Under a contract with the European Space Agency (ESA), Sweden-based Frontgrade Gaisler is designing a new Risc-V processor tailored to meet the requirements of microcontrollers for the space industry. For this project, Frontgrade Gaisler is concentrating on deterministic operation and minimal latency, then the processor will be integrated into radiation-hardened microcontrollers and FPGAs that support space missions.

Risc-V processors will account for almost a quarter of the global processor market by 2030, according to new research by Omdia. Between 2024 and 2030, Omdia forecasts Risc-V-based processor shipments to increase by nearly 50% per year, culminating in 17 billion processors shipped in 2030. The market research firm expects that 46% of those processors will be found in industrial applications, although the biggest growth over the forecast period – 66% annually – will come in the automotive segment. According to Omdia, one of the aspects that make Risc-V attractive for automakers is the possibility to “own” the design in a way that is impossible with a licensed ISA. Additionally, the rise of Artificial Intelligence is also expected to contribute to the continued growth of Risc-V.

“Chip war” updates

China has reportedly set up its third planned state-backed investment fund to boost its semiconductor industry, with a registered capital of $47.5 billion. The third phase will be the largest of the three funds launched by the China Integrated Circuit Industry Investment Fund, known as the “Big Fund.” The Big Fund has provided financing to China’s two biggest foundries, SMIC and Hua Hong Semiconductor, as well as to Yangtze Memory Technologies, a maker of flash memory.

Reuters has recently reported that two Chinese chipmakers are in the early stages of producing high bandwidth memory (HBM) chips. CXMT, China’s top manufacturer of DRAM chips, has developed sample HBM chips in partnership with packaging and testing company Tongfu Microelectronics; and Wuhan Xinxin is building a factory that will be able to produce 3,000 12-inch HBM wafers a month with construction slated to have begun in February this year. Separately, Huawei is reportedly aiming to produce HBM2 chips in partnership with other domestic companies by 2026.

Nvidia has reportedly cut the price of its H20 chip on the Chinese market, to make it more competitive compared to Huawei’s Ascend 910B. Due to US export restrictions, the H20 is now the most powerful Nvidia product sold in China.

South Korea launches a $19 billion support package for its semiconductor industry

The South Korean government has announced a new $19 billion program to support the domestic semiconductor industry. Most of the funding will be used to support tax credit and low interest rates for investment in semiconductor facilities; in addition to that, the package will also support R&D, human resource development, and infrastructure development for the previously announced “semiconductor mega cluster” to be built in the cities of Pyeongtaek, Hwaseong and Yongin in the southern region of Gyeonggi-do Province. One of the goals set by the South Korean government is to strengthen the country’s position in non-memory chips.

Lithium tantalate a promising candidate for volume manufacturing of photonic ICs

A research team from Swiss institute EPFL (École Polytechnique Fédérale de Lausanne) is proposing lithium tantalate (LiTaO3) as a material for the volume manufacturing of photonic integrated circuits. According to the researchers, lithium tantalate offers equal, and in some cases superior, properties to lithium niobate (LiNbO3), a material which has a limited industrial use because of its high cost per wafer and limited wafer size. In contrast, lithium tantalate has already been adopted commercially for 5G radiofrequency filters and therefore enables scalable manufacturing at low cost. The study shows that LiTaO3 can be etched to create low-loss (5.6 dB m−1) photonic ICs using a deep ultraviolet stepper-based manufacturing process.

Further reading: exploring human brain circuitry

The Connectomics team at Google Research has recently published an incredibly interesting work based on an electron microscopy reconstruction of a cubic millimeter of human brain cortex. The authors produced 1.4 petabytes of electron microscopy data and developed several AI-based tools for analyzing these data. The small brain sample contains about 16,000 neurons, 32,000 glial cells, 8,000 blood vessel cells (for a total of ~57,000 cells) and 150 million synapses. This project revealed never-before-seen structures within the human brain, such as a pair of neurons connected by more than fifty individual synapses, mirrored neurons in the deepest layer of the cortex, and axon whorls. The work was first released in 2021 as a preprint. Connectomics studies the brains of living beings using what could be described as an “electronic”, approach, that is, mapping the circuitry that connects neurons and trying to make sense of it from a function point of view.

Petabyte connectomic reconstruction of a volume of human neocortex. Left: Small subvolume of the dataset. Right: A subgraph of 5000 neurons and excitatory (green) and inhibitory (red) connections in the dataset. The full graph (connectome) would be far too dense to visualize. Credit: Google Research.

Logged in as . Log out »

© 2024 Internet Business Systems, Inc.
670 Aberdeen Way, Milpitas, CA 95035
+1 (408) 882-6554 — Contact Us, or visit our other sites:
TechJobsCafe - Technical Jobs and Resumes EDACafe - Electronic Design Automation GISCafe - Geographical Information Services  MCADCafe - Mechanical Design and Engineering ShareCG - Share Computer Graphic (CG) Animation, 3D Art and 3D Models
  Privacy PolicyAdvertise