talep üzerine web semineri

AheadComputing: A novel approach towards RTL metrics driven power methodology for high performance core

Tahminî İzleme Süresi: 20 dakika

Paylaş

AheadComputing logo image

Early RTL power analysis for high-performance AI cores often lacks actionable metrics and is applied too late, allowing inefficient clocking and logic activity to escape into costly downstream fixes. This session presents a scalable, automated, metrics-driven RTL power methodology enabling early, data-aware optimization during the architecture and RTL phases. Fast RTL power analysis is integrated into daily regressions using real workloads and synthetic tests, correlating power and performance on identical execution content while tracking instruction-level power cost. Continuous monitoring of clock, data, and register activity drives targeted optimizations such as clock-gate refinement and logic shaving. Applied to a production AI core, the methodology reduced clock-gate inefficiency from ~45K to ~3K flops within weeks and exposed major power bugs well before backend implementation.

Konuşmacı ile tanışın

AheadComputing

Yosi Hen

PnP and PM Architect

Yosi Hen is a Performance and Power (PnP) and Power Management Architect at Ahead Computing, where he leads the power architecture for the NUE Core, a high-performance RISC-V processor targeting advanced process nodes. His current work encompasses power management flows, dynamic voltage and frequency scaling (DVFS), thermal management, and power analysis and optimization leveraging industry-leading tools including Siemens PowerPro.

Prior to joining Ahead Computing, Mr. Hen spent nearly two decades at Intel Corporation, where he held progressively senior roles in power and performance architecture. As a PnP Architect Lead and Group Manager, he led power convergence efforts for next-generation client products and built Intel's first multi-die power model. He also pioneered power optimization methodologies utilizing machine learning techniques. In his most recent role at Intel, he served as a SoC/Power Management Architect responsible for GPU and AI accelerator integration.

Mr. Hen holds a Bachelor of Science in Electrical Engineering from the Technion – Israel Institute of Technology and an International Executive MBA from the University of Haifa.