Blog Review: Aug. 26

Cadence’s Paul McLellan shares some highlights from Hot Chips, including the massive growth in deep learning models, the basics of designing neural network models, and challenges involved in different approaches.

Mentor’s Colin Walls explores memory management units, its job of translating an address used by the CPU to an alternative address, and why this remapping is desirable and useful.

Synopsys’ Marc Serughetti and Johannes Stahl find that prototyping, both FPGA-based and virtual, are crucial pieces to faster hardware verification and pre-silicon software bring-up.

Ansys’ Craig Hillman points to why automotive electronics require consideration beyond conventional approaches to ensure reliability when auto electronics are exposed to much wider environmental effects and failure mechanisms are difficult to accelerate.

Arm’s Alex Shang, Yabin Zheng, Mary Bennion, and Alex Avramenko consider a new way to upscale images from lower resolutions to support 4K, and 8K screens that uses deep learning to generate pixels intelligently, improving the visual quality over traditional interpolation upscaling while still being deployable on edge devices.

In a blog for SEMI, John West of VLSI Research notes that COVID-19 has forced all the players in the semiconductor manufacturing supply chain to reassess how much inventory they need to hold, with a particular focus on critical subsystems.

Rambus’ Suresh Andani takes a look at PCIe 5 adoption for hyperscale data centers and where it will be deployed in the data center architecture to support an increasing number of low-latency and time-sensitive applications.

And don’t miss the blogs from last week’s Manufacturing, Packaging & Materials newsletter:

Editor in Chief Ed Sperling sees no shortage of demand for more compute power, but the underlying assumptions must change.

Executive Editor Mark LaPedus talks with Semico Research’s CEO about his outlook for the semiconductor biz.

Lam Research’s Russell Dover explains how to simplify analyzing large amounts of tool data to identify the root of problems.

SEMI blogger Walter Custer warns that while chips and capital equipment have outperformed so far this year, different dynamics are in play entering the second half.