Week In Review: Design, Low Power

AnsysGPT assistant; Stampede3 lands in Texas; Japan’s generative AI supercomputer; quantum entanglement; Semiconductor Education Alliance forms.


Arm is helping to address the ongoing talent shortage through its newly announced Semiconductor Education Alliance, with a long list of partners, including ArduinoCadenceCornell UniversitySemiconductor Research Corp.STMicroelectronics,SynopsysTaiwan Semiconductor Research Institute, the All-India Council for Technical Education, and the University of Southampton. The Alliance invites anyone who can contribute to the semiconductor skills pipeline to get involved.

Data centers

CoreWeave is opening a new 450,000-square-foot data center in Plano, Texas that will feature high-performance GPUs for for artificial intelligence, machine learning, pixel streaming, and other emerging tech.

Experts from AMD, Cadence, Siemens, Synopsys, and Lightmatter talk about how the shift toward chip design on cloud has sped up, whether the benefits of cloud are realized in chip design, and some of the most pressing challenges to chip design on cloud today.

Broadcom is now shipping its Trident 4-X7 Ethernet switch ASIC— a 4.0 Tb/s fully programmable switch designed for enterprise data center ToR (Top of Rack) boxes. The Trident 4-X7 offers native support for 400G connectivity to the next-generation spine/fabric technologies, which are making their way from the cloud into enterprise data centers.

Micron is now sampling its  8-high 24GB HBM3 Gen2 memory for HBM3 with bandwidth greater than 1.2TB/s and pin speed over 9.2Gb/s.

Design Tools, IP

Ansys uncorked a beta launch of AnsysGPT, a virtual assistant that offers engineering expertise across physics domains and provides 24/7 comprehensive technical support. Ansys created the assistant using ChatGPT technology from the Microsoft Azure OpenAI Service. The assistant, which is trained with Ansys’ data from Ansys resources and no proprietary information used during the training process, is expected to be more widely available in early 2024 after a subset of customers and channel partners test it.

Ansys announced its latest update, Ansys 2023 R2, to its engineering tools platforms, adding multiphysics solvers for 3D integrated circuits (3D-IC) and EVs.

Verification of complex, heterogeneous chips is becoming much more difficult and time-consuming. There are more corner cases, and devices have to last longer and behave according to spec throughout their lifetimes. This is where AI fits in. It can help identify redundancy and provide information about why a particular device or block may not be able to be fully covered, and it can do it in less time than relying on traditional methods, which are time-consuming. This Semiconductor Engineering tech talk looks at what’s behind these changes and where AI works best.

Development flows are evolving as an increasing number of optimization factors become interlinked. Shift left is just one piece of it. Dev teams are now asked to shift left, extend right, and stretch sideways.


The Texas Advanced Computing Center (TACC) at The University of Texas at Austin will be installing a Stampede3 supercomputer that has a 4 petaflop capability for high-end simulation. The computer will have 560 new Intel Xeon CPU Max Series processors with nodes that are enabled for high-bandwidth memory, adding nearly 63,000 cores for the largest, most performance-intensive compute jobs. National Science Foundation’s (NSF) is paying USD $10 million toward the project. The Stampede3 will be part of NSF’s supercomputing ecosystem.

The U.S. Navy DoD Supercomputing Resource Center will install 17.7 petaflops ‘Blueback’ supercomputer, at the Stennis Space Center, Mississippi. The computer will be a HPE Cray EX4000 system with both AMD and Nvidia hardware.

Researchers from Georgia Tech and Hanoi University of Science and Technology are using the Expanse supercomputer at the San Diego Supercomputer Center (SDSC) at UC San Diego to find superconductors. The researchers created AI/ML tools to help identify potential superconductors more quickly.

Japan‘s Ministry of Economy, Trade and Industry (METI) has commissioned the National Institute of Advanced Industrial Science and Technology (AIST) to create a supercomputer that can compute 2.5 times faster than its existing machine specifically to develop generative artificial intelligence (AI) in Japan. The supercomputer will be online as early as 2024.

Market research

The Semiconductor Industry Association published its 2023 State of the Industry Report about the U.S. chip industry, finding that since the CHIPS Act was introduced, USD $200 billion of private money has gone into new semiconductor ecosystem projects in the U.S.

ABIResearch warns of long-term fragmentation that could result in the generative AI products designed for the enterprise market if a “robust corporate strategy” is not in place. Deployments of generative AI systems that are isolated from business processes will drive fragmentation within an enterprise. The research firm estimates that generative AI applications such as ChatGPT and Stable Diffusion will add an additional U.S. $450 Billion to 12 verticals in the enterprise market over the next seven years.

Research firm Technavio estimates that the hyperscale data center market will grow at a CAGR of 26.44% between 2022 and 2027, increasing by USD $169.47 billion.


Clemson University scientists and their collaborators found a way to detect the quantum entanglement in quantum materials out of equilibrium and can control an artificially induced entanglement by using laser pulses. “We not only want to be able to detect entanglement, we want to be able to control it and keep track of it while we control it. One efficient way you can control it is with an intense laser, which sends a system out of equilibrium,” said Jordyn Hales, a a graduate student in the Clemson Department of Physics and Astronomy who is first author of the paper. “Laser controlled material means you can synthesize one material and shine the laser. The laser has different power, frequency, duration and polarization, so you can control it in different ways. That means for a single material, you have access to many different states in a large parameter space.”

MIT researchers created flexible quantum magnets for computers, robotics, and sensors. MIT Plasma Science and Fusion Center (PSFC) research scientists noticed, while modulated temperature and strain of a high-quality quasi-2D Cr2Te3 MBE-grown thin film, a novel Berry-curvature-induced magnetism occurs featuring an extraordinary sign reversal of the anomalous Hall effect. They concluded in the article in Nature, that the material “pristine ferromagnetic Cr2Te3 thin films [are] a fascinating platform for further engineering topological effects, given their nontrivial Berry curvature physics.”

Amid soaring temperatures across the world, wild fires burning, a report from the University of Copenhagen’s Niels Bohr Institute and Department of Mathematical Sciences that the ocean currents are slowing rapidly. The current slowing has long been theorized as an effect of the Earth’s ice caps melting, but this report estimates the currents will collapse with 95 percent certainty – between 2025 and 2095.

The chip industry is stepping up efforts to be seen as environmentally friendly, driven by growing pressure from customers and government regulations. Some manufacturers have been addressing sustainability challenges for more than a decade, but they are becoming more aggressive in their efforts, while others are joining them.


Flash Memory Summit awarded Google Cloud’s Amber Huffman this year’s lifetime achievement award for bringing flash storage into the mainstream by founding and driving standards for virtually all computing apps.

Upcoming events

  • 2023 Flash Memory Conference & Expo – August 8-10 (Santa Clara, CA)
  • DARPA: Electronics Resurgence Initiative (ERI) – August 22-24 (Seattle, WA)
  • Hot Chips 2023 – August 27-29 (Hybrid online & Stanford, CA)
  • NVMTS 2023: Non-Volatile Memory Technology Symposium – August 30-September 1 (Leuven, Belgium)
  • IEEE International System-on-Chip Conference (SOCC): SoCs/ SiPs for Edge Intelligence & Accelerated Computing – September 5-8 (Santa Clara, CA)
  • AI Hardware Summit 2023 – September 12-14 (Santa Clara, CA)
  • DVCON India: Design & Verification Conference & Exhibition – September 13-14 (Bangalore, India)
  • More events and webinars

Check out the latest Low Power-High Performance and Systems & Design newsletters for these highlights and more:

  • Improving Performance And Lowering Power In Automotive
  • Getting Rid Of Heat In Chips
  • HBM’s Future: Necessary But Expensive
  • Shift Left, Extend Right, Stretch Sideways
  • The Good And Bad Of Chip Design On Cloud
  • Large-Scale Integration’s Future Depends On Modeling
  • Using AI To Close Coverage Gaps


If you’d like to receive Semiconductor Engineering newsletters and alerts via email, subscribe here.

Leave a Reply

(Note: This name will be displayed publicly)