Data movement is the new performance battleground in semiconductor design
For much of the semiconductor industry’s history, performance debates have centered on compute throughput and memory capacity. Faster processors, wider vectors, and larger caches have been the primary levers for system architects, often treating the movement of data between them as a secondary concern.
Today, a different constraint is asserting itself as AI data center workloads proliferate, architectures diversify, and systems extend beyond traditional computing into the physical world. Data movement, rather than processing or storage, is increasingly defining the limits of performance, power efficiency, predictability, determinism, and scalability.
VP of Product Management and Marketing at Arteris.
This shift around data movement is already visible. Across applications like advanced SoCs, AI accelerators, chiplet-based systems, and now in physical AI applications such as robotics, industrial automation, and intelligent vehicles, it has become clear that transporting the vast quantities of data required by such workloads is more demanding than processing...
Copyright of this story solely belongs to techradar.com. To see the full text click HERE
