NVIDIA announced the AI Cluster Runtime (AICR) to simplify Kubernetes readiness for NVIDIA Vera Rubin GPUs. AICR captures tuned configurations as reusable recipes, generates deployment bundles, and validates running clusters. This ensures Kubernetes environments are ready for Vera Rubin NVL72 compute trays on Day 1.
What happened
NVIDIA's AI Cluster Runtime (AICR) simplifies Kubernetes readiness for NVIDIA Vera Rubin GPUs, ensuring optimal performance and compatibility. AICR captures tuned configurations as reusable recipes, generates deployment bundles, and validates running clusters. The Vera Rubin NVL72 recipe family includes validated choices for the R615.23 open-kernel-module driver, DRA driver v0.5.0 with ComputeDomains, Ubuntu 26.04 with a 64k-page kernel, and platform-specific settings. End-to-end validation on Vera Rubin NVL72 reference clusters using GPU Operator v26.7.0 and DRA driver v0.5.0 achieved 744.71 GB/s bus bandwidth at 16 GiB over the NVLink fabric. AICR validation on early silicon identified an unmapped Rubin architecture value in go-nvlib, leading to a fix in the official DRA driver v0.5.0 release.
Source: NVIDIA