SNUG announcements; meshless multiphysics simulation; flaws in AI-generated RTL; open-source alternative to CUDA; skyrmion-based memory

Is pain a positive thing for character-building? Or is it just that humans instinctively need to find a reason to justify pain? Quite an off-topic question here – but not that much, after all, if it stems from a speech given by Nvidia CEO Jensen Huang. For his take on character-building, see the “Further reading” paragraph at the end of this week’s news roundup. But first, some on-topic technology updates.

New Synopsys announcements from SNUG Silicon Valley

Here’s a quick overview of some of the announcements Synopsys made on occasion of its recently held annual Synopsys User Group (SNUG) conference in Silicon Valley. In the area of multi-die designs, 3DSO.ai is a new AI-driven capability built natively into Synopsys 3DIC Compiler, a unified exploration-to-signoff platform. 3DSO.ai offers optimization for signal integrity, thermal integrity, and power-network design. Also targeted at multi-die designs, Synopsys Platform Architect – Multi-Die accelerates design timelines, delivering – according to the company – a six to twelve month “shift left” from RTL for the analysis of performance and power, while accounting for the interdependencies between multiple dies and allowing early partitioning decisions. Synopsys also unveiled two new hardware-assisted verification solutions: ZeBu EP2, the latest version in the ZeBu EP family of unified emulation and prototyping systems; and HAPS-100 12, Synopsys’ highest capacity and density FPGA-based prototyping system. Additionally, the company introduced Synopsys Cloud Hybrid solution, which enables users to burst from on-prem data centers to the cloud during peak needs – automatically splitting the job based on available capacity and eliminating manual data transfers. Lastly, Synopsys announced that it has completed the acquisition of Netherland-headquartered Intrinsic ID, a provider of Physical Unclonable Function (PUF) IP.

Altair to extend its meshless technology to electronics

Altair has announced the upcoming release of Altair SimSolid for electronics – promising fast, easy, and precise multi-physics scenario exploration for electronics, from chips to PCBs and full system design. SimSolid is an already existing Altair product, which so far has gained adoption in industries such as aerospace and automotive. According to the company, SimSolid’s main benefit is its ability to eliminate geometry simplification and meshing, the two most time-consuming and expertise-intensive tasks done in traditional finite element analysis. As a result, it is up to 25x faster than traditional finite element solvers, and effortlessly handles complex assemblies. Extending Altair SimSolid’s meshless technology to electronics will enable the tool to tackle intricate challenges like signal integrity, power integrity, and electromagnetic compatibility/interference, all while making simulations more accessible and efficient.

60% of LLM-generated RTL code is prone to security weaknesses

DVCon US’ best paper award went to a group of Infineon researchers for their work titled “All Artificial, Less Intelligence: GenAI through the Lens of Formal Verification.” The paper outlines a method to verify and address hardware CWEs (Common Weakness Enumerations) in RTL designs generated by generative AI from different LLMs (Large Language Models). This has resulted in the creation of the ReFormAI dataset, which contains 60,000 SystemVerilog RTL designs that can be utilized to train LLMs and ML algorithms to avoid generating CWE-prone hardware designs. The study reveals that approximately 60% of the hardware designs generated by LLMs are prone to CWEs, posing potential safety and security risks.

UXL Foundations to reportedly offer an open-source alternative to Nvidia’s CUDA

Announced in September 2023, the Unified Acceleration (UXL) Foundation is a cross-industry group committed to delivering an open standard accelerator programming model that simplifies development of performant, cross-platform applications. Hosted by the Linux Foundation’s Joint Development Foundation (JDF), the Unified Acceleration Foundation is an evolution of the Intel’s oneAPI initiative, which has witnessed significant growth as an open programming model that spans different architectures such as CPU, GPU, FPGA, and accelerators. Participating organizations and partners include Arm, Fujitsu, Google Cloud, Imagination Technologies, Intel, Qualcomm and Samsung. According to a recent Reuters report, the main goal of the UXL Foundation is to offer an open-source alternative to CUDA, the Nvidia software which is one the strongest assets behind Jensen Huang’s primacy in AI acceleration.

Export restrictions reportedly prompting China to explore quadruple patterning and an open-source ecosystem

Huawei and a Chinese foundry – presumably SMIC – have reportedly submitted patents for a “self-aligned quadruple patterning” (SAQP) solution. This would enable the companies to produce 5-nanometer chips using their existing lithography equipment – as they do not have access to the latest EUV equipment due to U.S. export restrictions. The use of quadruple patterning is a less-than-ideal solution in terms of manufacturing yield and chip cost; these drawbacks, however, can be overlooked in applications that are not price-sensitive. Export restrictions are also prompting China to leverage open-source technologies, both in hardware and in software. At the recent “Xuantie Risc-V Ecological Conference” held in Shenzhen, China, Alibaba’s research department reportedly promised to deliver a server-grade Risc-V processor later this year, and showed off a Risc-V-powered laptop running openEuler – the open version of the EulerOS that Huawei created by adapting Red Hat Linux. The notebook also runs DingTalk – Alibaba’s Zoom-and-Teams-like collaboration suite – and Libre Office.

Skyrmion-based memory promises 1,000x power reduction

A research team from National University of Singapore has come up with a way to electrically control skyrmions, tiny magnetic whirls – 10,000 times smaller than the width of a human hair – that form within specific magnetic layers when they are made extremely thin. The researchers presented the wafer-scale realization of a nanoscale chiral magnetic tunnel junction (MTJ) hosting a single, ambient skyrmion. The MTJ realizes three non-volatile electrical states and can electrically write and delete skyrmions with switching energies 1,000 times lower than the state of the art. According to the team, this work provides the much-anticipated backbone for all-electrical skyrmionic device architectures.

Improving MEMS with additional scandium content through a new laser-based equipment

Increasing the percentage of scandium, a rare-hearth element, in aluminum scandium nitride (AlScN) films allows to achieve low dielectric loss and twice the piezo coefficient of current sputtered films, optimizing electrical conversion to drive enhanced sensitivity in MEMS RF filters and better performance in MEMS microphones. Further, improved piezoelectric qualities make it feasible to replace lead zirconate titanate (PZT) with lead-free AlScN. Now “Pulsus” – a new pulsed laser deposition (PLD) tool from Lam Research, developed in collaboration with French technology hub CEA-Leti – delivers AlScN films with the highest scandium content available, at least 40%.

Further reading

Here’s part of the transcript from a speech recently given by Nvidia CEO Jensen Huang at the Stanford Institute for Economic Policy Research (SIEPR), addressing an audience of Stanford students: “One of my great advantages is that I have very low expectations (…). People with very high expectations have very low resilience, and unfortunately resilience matters in success (…) Greatness is not intelligence. Greatness comes from character, and character isn’t formed out of smart people. It’s formed out of people who suffered and so (…) I wish upon you ample doses of pain and suffering.” The video of this part of the speech is available here.

This entry was posted on Thursday, March 28th, 2024 at 8:08 am. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.