A recent blog review from Semiconductor Engineering rounds up hardware-related posts from Synopsys, Siemens, Cadence, and others. The entries span interconnect standards, memory technology, and validation strategies, offering a snapshot of where compute hardware is headed.
Siemens' Sriram Bhaskar Teja Mundru describes how UALink over UCIe 3.0 improves cross-vendor interoperability, standardizing interfaces, form factors, and flow control to enable scalable AI systems. Cadence's Shyam Sharma details the GDDR7 standard's key features, including channel operation, encoding and signaling, command and clock models, and protection mechanisms.
On the testing front, Teradyne's Ben Mitchel argues that every AI and data center device must be tested under real-world conditions. Keysight's Mike Wager similarly considers how to validate applications across the full range of devices and environments without maintaining individual test suites. The roundup also touches on on-device AI, with Arm's Ben Clark combining LiteRT, quantization, KleidiAI, and SME2 to reduce latency in mobile toxic speech detection.