Tech Talk: Analog Simplified

As advanced-node chips are added into cars, and usage models shift inside of data centers, new questions surface about reliability.

Is the industry heading toward another hardware/software divide in machine learning? Both sides have different objectives.

Cost of porting tools and IP will limit choices at partial nodes and create confusion at others.

This more than Moore technology is still ramping up, and problems need to be solved, but it could lead to some fundamental changes.

Are optimizations really a zero sum gain? Where to gamble and when to play it safe.

Why new approaches are needed to tie together training and inferencing.

The race is on to simulate thermal and electromagnetic effects.

Heat is becoming a bigger problem for chips developed at new process nodes and for automotive and industrial applications.