U.S. Cyber Trust Mark; Cerebras/G42 AI supercomputers; Cadence-Rambus deal; Wi-Fi 7; 32Gbps GDDR7 DRAM; Ultra Ethernet Consortium; Llama 2; Stellantis’s $11B deals; battery developments; CWE vulnerabilities; organic bioelectronics.
The Biden-Harris Administration announced the U.S. Cyber Trust Mark, a cybersecurity certification and labeling program to help consumers choose smart devices less vulnerable to cyberattacks. The Federal Communications Commission (FCC) is applying to register the Cyber Trust Mark with the U.S. Patent and Trademark Office and it would appear on qualifying smart products, including refrigerators, microwaves, televisions, climate control systems, and fitness trackers. Amazon, Best Buy, Google, LG, Logitech, and Samsung said they will support the program, with others likely to follow.
Cerebras and UAE-based G42 unveiled a network of nine interconnected supercomputers aimed at reducing AI model training time. The Condor Galaxy network has a total planned capacity of 36 exaFLOPs. The first AI supercomputer, Condor Galaxy 1 (CG-1), has 4 exaFLOPs and 54 million cores. CG-2 and CG-3 will deploy in the U.S. in early 2024.
Cadence inked a deal to acquire Rambus’ PHY IP assets for SerDes and memory interfaces, which Cadence says will be important for large data centers and AI applications. Rambus will still retain its digital IP business, including memory and interface controllers and security IP. Terms of the deal were not disclosed.
Samsung released a next-generation 32Gbps GDDR7 DRAM to boost capabilities in automotive, artificial intelligence (AI), and high performance computing (HPC) applications, featuring 1.4x better performance and 20% better power efficiency compared to its 24Gbps GDDR6 DRAM.
The Ultra Ethernet Consortium (UEC) launched a complete Ethernet-based communication stack architecture that will be scalable, cost-effective, and “capitalize on Ethernet’s ubiquity and flexibility for handling a wide variety of workloads” including cloud, networking, high performance computing (HPC) at scale, and artificial intelligence (AI). Founding members include AMD, Arista, Broadcom, Cisco, Eviden, HPE, Intel, Meta, and Microsoft.
Wi-Fi 7 brings better speeds and data handling, but it does little to bridge various communications technologies, making it more difficult and more expensive to design chips that must integrate and support multiple wireless technologies, including different versions.
Meta and Microsoft introduced the next generation of their open-source large language model (LLM), Llama 2, which is free for research and commercial use. Meta is also working with Qualcomm to optimize the execution of Llama 2 large directly on-device without relying on the cloud services, saving costs for developers and offering users a private, more reliable, and personalized experience.
Cadence collaborated with GlobalFoundries (GF), Hoerzentrum Oldenburg gGmbH, and Leibniz University Hannover to develop a binaural hearing aid system-on-chip (SoC) prototype. Cadence’s GF 22FDX platform provided up to 50% lower power at the same high-performance frequencies (vs. 28nm), enabled by the adaptive body bias (ABB) feature.
MLCommons announced MedPerf, an open benchmarking platform that evaluates artificial intelligence (AI) models on real-world medical data, and invited healthcare professionals, patient advocacy groups, AI researchers, data owners, and regulators to join the effort.
Expedera uncorked a family of artificial intelligence (AI) inference edge neural processing units (NPUs) for use in always-sensing camera applications on smartphones, security cameras, home appliances, and other power-constrained devices.
Synopsys released a portfolio of interface IP for the TSMC N3E process, offering a fast path integration and enabling chip designers to accelerate development of their AI, high-performance computing (HPC), and mobile designs.
Intel agreed to let ASUS manufacture, sell, and support the Next Unit of Compute (NUC) 10th to 13th generations systems product line, and develop future NUC systems designs.
MIT researchers developed a new privacy metric, Probably Approximately Correct (PAC) Privacy, which can help engineers create machine-learning models while protecting sensitive training data by adding the smallest amount of noise possible.
Argonne National Laboratory released a report highlighting the crucial importance of responsible development of artificial intelligence (AI), including challenges related to technology, implementation, and application.
The U.S. Department of Commerce launched a website for companies to participate in US-EU data transfers as part of the recently signed Data Privacy Framework.
Stellantis implemented a broad semiconductor strategy featuring supply agreements for automotive chips with a purchasing value of over €10 billion (~US$11.2B) through 2030. Partners, including Infineon, NXP, onsemi, Qualcomm, aiMotive, and SiliconAuto will supply chips such as silicon carbide (SiC) MOSFETS, microcontroller units (MCUs), and system-on-a-chip (SoCs).
onsemi and mobility company BorgWarner expanded their silicon carbide (SiC) collaboration, worth over $1 billion total. By using onsemi’s EliteSiC technology, BorgWarner aims to increase the range and performance of its electric vehicles (EVs) through increased power density and higher efficiency.
Automotive OEMs are boosting their investments across the semiconductor ecosystem as stepping stones toward electrification and autonomy, and they are starting to encounter some of the same issues chipmakers have been wrestling with at advanced nodes — massive compute performance, thermal and power issues, reliability over extended lifetimes, and a highly diverse and geographically distributed supply chain.
Infineon introduced four products with automotive applications:
Renesas collaborated with AMD on a reference design for next-generation space avionics systems that integrates radiation-hardened components for power management and includes four new ultra-compact Intersil-brand ICs. Meanwhile, Renesas’s automotive Cyber Security Management System (CSMS) was defined and implemented according to International Standard ISO/SAE 21434:2021.
Siemens announced SK Hynix achieved ASPICE automotive semiconductor software quality certification in Korea, enabled by Siemens’ Polarion ALM solution for Application Lifecycle Management.
Nissan became the first Japanese automaker to adopt Tesla’s North American Charging Standard (NACS), beginning in 2025.
The National Highway Traffic Safety Administration (NHTSA) opened a special crash investigation into a fatal accident on July 5 in California involving a 2018 Tesla Model 3 that may have been using advanced driver assistance systems (ADAS), according to Reuters.
Researchers at Ohio State University proposed a vehicle-in-virtual environment (VVE) way to develop and evaluate connected and autonomous driving functions. They claim VVE is safe, efficient, and low-cost, and reduces the risk of deploying test vehicles on public roads.
A dry battery manufacturing process eliminates solvent and shows promise for a cleaner “battery that is durable, less weighed down by inactive elements and able to maintain high energy storage capacity after use,” according to scientists at the Department of Energy’s Oak Ridge National Laboratory.
Fig. 1: Making battery anode film. Source: ORNL
A new and safer fluorine-containing electrolyte for lithium-ion batteries works in sub-zero conditions, making it easier for drivers of electric vehicles (EVs) to maintain charge in winter, according to scientists from the U.S. Department of Energy’s (DOE) Argonne and Lawrence Berkeley national laboratories.
Argonne National Laboratory will use J.D. Power’s EV Index to access new electric vehicle (EV) data sets to “better understand drivers’ interest in EVs, patterns of EV adoption, experience with charging stations, preferences, and challenges.
Touch and display driver integration (TDDI) architecture is predicted to increase 54% year over year in 2023, with shipment volume rising from 24 million units in 2022 to 38 million units in 2023, according to Trendforce.
RowHammer is a well-known read-disturb phenomenon. Now, researchers at ETH Zürich demonstrated and analyzed another widespread issue, RowPress, which affects DDR4 chips.
Even when process isolation is implemented correctly at the software level, it may be compromised by the hardware. Researchers at University of Kaiserslautern-Landau considered a threat model where the attacker is able to inject faults and execute user-level programs.
Hardwear.io called for researchers and hackers around the world to submit interesting hardware offense/defense research to share with the security community at the ninth Hardware.io Netherlands conference on Nov. 2 and 3, 2023. Call for papers (CFP) closes on Aug 20.
The MITRE-managed Common Weakness Enumeration (CWE) community published a list of 15 weaknesses on the cusp of being included in the Top 25, which continue to be prevalent and severe enough to cause concern.
CISA issued numerous alerts, including seven industrial control systems (ICS) advisories and one addition to the known exploited vulnerability catalog concerning Microsoft Office and Windows.
DARPA is working with universities and industry partners to create artificial intelligence (AI) agents capable of collaborating with humans to analyze “image, video, and multimedia documents during time-sensitive analytical tasks for national security, where reliability and robustness are essential.”
Netcraft raised more than $100M from Spectrum Equity to help “accelerate progress toward our mission of detecting and disrupting cybercrime at scale.”
UK-based Optalysys raised £21M (about US$27M) to advance its photonic computing technology and “unlock a new form of secure processing known as Fully Homomorphic Encryption (FHE), aiming to close vulnerabilities in cloud systems in a more efficient manner than current solutions.”
A malicious artificial intelligence (AI) language tool, WormGPT, is capable of producing a persuasive, cunning email that could be used for sophisticated phishing and business email compromise (BEC) attacks.
The U.S. National Institute of Standards and Technology (NIST) invited people to host an event, distribute career information, or engage with others through social media, during Cybersecurity Career Week, October 16 to 21, 2023.
The Biden-Harris administration published the National Cybersecurity Strategy Implementation Plan (NCSIP) aiming to defend critical infrastructure, disrupt and dismantle threat actors, shape market forces, invest in a resilient future, and forge international partnerships.
The United States must allocate more mid-band spectrum for licensed commercial 5G use to enhance its economic and national security priorities. If it is not addressed, the present spectrum shortage in the United States could hinder technological innovation and give China a path to global leadership in the increasingly connected twenty-first century.
The Center for Strategic and International Studies (CSIS) will host an in-person and online event on July 25 on the strategic national security imperative for allocating more mid-band spectrum for licensed commercial 5G use.
Columbia University and Columbia University Medical Center researchers created an organic bioelectronic device capable of advanced processing, data transmission, and device powering. The vertical internal ion-gated organic electrochemical transistor (vIGT) incorporates a “miniaturized hydration access conduit to enable megahertz-signal-range operation within densely packed integrated arrays in the absence of crosstalk.”
National University of Singapore researchers captured 2-D light patterns into DNA by utilizing optogenetic circuits to record light exposure into DNA, establishing a living digital camera and paving the way towards integrating biological systems with digital devices.
Researchers at MIT, New York University, and the University of California at Berkeley created a framework to teach a robot how to succeed after it fails. The system uses an algorithm to generate counterfactual explanations for the failure, then asks a human for feedback, then uses the feedback to generate new data to fine-tune the robot’s machine-learning model.
Researchers at several German universities found transition metal dichalcogenide (TMDC) transistors have potential for flexible electronics applications.
Find upcoming chip industry events here, including:
Upcoming webinars are here.
Read the latest automotive, security, and pervasive computing articles, or check out the latest newsletter.
 |
 |
 |
 |
 |
 |
 |
Leave a Reply