Memory In Physical AI
Memory In Physical AI
In a Semiconductor Engineering interview, Cadence group director of product management Frank Ferro explains why memory is becoming a key element of physical AI. Humanoid robots, autonomous vehicles, and drones are all constrained by battery life, so they need efficient processing, denser and faster memory, and optimized data movement between processors and memory, with LPDDR6 becoming a core component of mobile edge intelligence systems.
LPDDR6 becomes an essential ingredient as intelligence moves into autonomous and mobile edge systems.
Humanoid robots, autonomous vehicles, and drones all share one thing in common — they can only function as long as the battery lasts. After that, they need to return to a charger. Today, in the case of humanoid robots, usefulness may only last a few hours per charge. So how can that time be extended? There’s no simple answer because no single component will solve all problems, but the underlying basics remain the same — efficient processing, denser and faster memory, and optimized data movement between processors and memories. Frank Ferro, group director for product management at Cadence, explains why data hierarchies are so critical for physical AI, what can be done on the memory side to speed up read/write times, and how to maximize performance while keeping costs down.
Ed Sperling
(all posts)
Ed Sperling is the editor in chief of Semiconductor Engineering.
Source: Semiconductor Engineering Chips and Packaging · semiengineering.com