4+ day, 7+ hour ago   (1695+ words) Kilowatt-class AI accelerators are forcing engineers to rethink what counts as adequate coverage across workload, thermal behavior, test hardware, and time. On a spec sheet, 1kW of maximum power looks like a single operating condition. In practice, two workloads can hold…...

6+ day, 7+ hour ago   (130+ words) Just because a chip is deemed good doesn't mean it will work as expected. The post Smart Outlier Detection appeared first on Semiconductor Engineering. Nearly all leading-edge designs today are multi-die assemblies, which means that one defective die can render…...

1+ week, 5+ day ago   (85+ words) Engineering methodologies are evolving toward more integrated and intelligent approaches across three interconnected domains: optimization, automation, and AI. The post The Evolution Of Intelligent Systems: From Optimization To Automation to AI appeared first on Semiconductor Engineering. Traditional approaches to signal…...

1+ week, 6+ day ago   (298+ words) Researchers at Huawei, ETH Zürich, and HUST published a technical paper titled “FLINT: Efficiently Leveraging High Bandwidth Flash for Capacity-Scalable LLM Inference Acceleration.” Abstract: “LLM inference is increasingly constrained by accelerator memory capacity rather than compute throughput. This constraint is…...

2+ week, 2+ day ago   (214+ words) Researchers from Stanford University and SLAC National Accelerator Laboratory published a technical paper titled “Deep Learning to Automate Parameter Extraction and Model Fitting of Two-Dimensional Transistors.” Abstract Excerpt: “We present a deep learning approach to extract physical parameters (e.g.,…...

2+ week, 6+ day ago   (152+ words) The race is on for localized intelligence. The post Redefining Roles For Edge And Cloud AI appeared first on Semiconductor Engineering. The rapid build-out of AI on the edge marks a fundamental shift in where intelligence is located, how much…...