Supercharging SoCs for AI
with High Performance and Efficient Power
With the introduction of AI to industrial automation,
factories are moving to autonomy.
Automation works in a static environment where robots perform that same activity over and over. It benefits manufacturing environments that are highly controlled and produce large volumes of identical products. But autonomy requires the robot to react and change without explicit human instruction, in effect, the robot must be able to reprogram itself and potentially interact with other robots. This has incredible potential for manufacturing in high mix, low volume environments, or made-to-order products.
In a typical industrial autonomous application, a system might need to process sensor inputs from cameras, computer vision, and GPS in real time so that it can anticipate and respond appropriately. Capturing, fusing, analyzing, and acting on vast amounts of data uses AI and a variety of other innovative technologies. Though the technology is available, mustering the compute capability to process AI workloads remains a challenge. Since there is no single compute platform that can meet the needs of autonomous workloads, the industry is looking to heterogeneous computing which uses a mix of specialized processing cores.
The processing workhorse that delivers the necessary compute capability is the SoC. Designing an SoC to meet the requirements for autonomous applications requires pushing performance to new levels while reducing power consumption. An intelligent clocking network can make the critical difference in achieving the optimal balance between performance and power.
Maestro provides system architects the tools to meet demanding SoC performance requirements for autonomous industrial applications. It enables system developers to create an all-digital application-optimized clocking solution that eliminates system inefficiencies due to on-chip variation, jitter, clock skews, setup and hold violations, peak current, and switching noise. Maestro lets users optimize clocking based on workloads and operating conditions. Chip designers can speed up or slow down regions of the silicon to optimize power, performance or both. Maestro offers an architectural solution that dramatically increases performance while lowering power consumption.