Anyscale Ray optimizes NVIDIA GB300 NVL72 with NVLink Domain placement

Anyscale has introduced NVLink Domain-Aware Placement Groups for its Ray framework, designed to optimize performance on NVIDIA's GB300 NVL72 systems. These new placement groups ensure that tightly coupled actors are scheduled within the same NVLink Domain, which is a rack-scale fabric connecting 72 NVIDIA Blackwell GPUs and 36 NVIDIA Grace CPUs. This topology-aware scheduling primitive aims to maximize GPU communication and preserve system intent, building upon previous node-level placement strategies. AI

IMPACT Enhances performance for large-scale AI training on specialized NVIDIA hardware.

RANK_REASON This is a software feature update for optimizing hardware, not a new model release or significant industry event.

Read on Anyscale blog →

• Anyscale
• Blackwell GPUs
• DGX H200
• GB200
• GB300 NVL72
• Grace CPUs
• Hopper
• NVIDIA
• NVIDIA DGX H100
• NVLink
• NVLink Domain
• NVLink Domain-Aware Placement Groups
• Ray

• infra
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →