Hardware
CUDA Toolkit 13.4 Adds Windows on Arm Support and Greater Control over Shared GPUs
CUDA Toolkit 13.4 introduces Windows‑on‑Arm support and preview functional compatibility for the NVIDIA Rubin architecture (compute capability 107), enabling cross‑platform CUDA development on ARM sys
CUDA Toolkit 13.4 introduces Windows‑on‑Arm support and preview functional compatibility for the NVIDIA Rubin architecture (compute capability 107), enabling cross‑platform CUDA development on ARM systems. The release modernizes GPU resource management with Multi‑Process Service V3, offering a scriptable CLI, named server instances, TOML configuration, and cgroup‑integrated GPU memory limits; it also adds the Compute Fabric Transport API for NVLink data movement, expands CUDA Python (texture/surface programming, NUMA‑aware memory), and improves CCCL 3.4 with higher‑bandwidth kernel variants and standardized C++ parallel algorithms. Developers can download Toolkit 13.4 to access these features and consult the MPS V3 quickstart and CUDA Python release notes for detailed usage.
Related
- Inside NVIDIA Rubin GPU Architecture: Powering the Era of Agentic AI
- Introducing CUDA Rust: Two Tracks for Writing GPU Kernels
- How to Run Isolated Tenant Kubernetes Clusters on Shared GPU Infrastructure
Source: NVIDIA Developer | 2026-09-09