← Back to the feed
AI & softwareResearch findings

Performance Optimization of Autonomous Driving Perception Using Multi-Chiplet NPUs

Original published 2024.11.25 07:59 KSTAdded to Blackdeck 2026.11.12 07:34 KST

What has changed?

A novel scheduling strategy for efficiently deploying perception workloads on multi-chiplet AI accelerators

Key points

  1. Multi-chiplet NPUs show strong potential for autonomous driving perception workloads
  2. Performance profiling of Tesla Autopilot pipeline on various chiplet accelerators conducted
  3. Proposed scheduling strategy achieves 82% throughput increase and 2.8x engine utilization improvement

Connect it to your work

Performance analysis of Tesla Autopilot perception pipeline shows improved throughput and engine utilization on chiplet-based accelerators

What remains unproven

The abstract reports simulation results; real-world performance is not established by this input.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivPerformance Implications of Multi-Chiplet Neural Processing Units on Autonomous Driving Perception