CUDA Toolkit 13.4 Adds Windows on Arm Support and Greater Control over Shared GPUs

Summary

CUDA Toolkit 13.4 extends CUDA application support to Windows on Arm and provides preview support for NVIDIA Rubin (compute capability 107), alongside updates to GPU management, CUDA Python, CCCL, Nsight tools, and core math libraries. Multi-Process Service V3 adds a scriptable CLI, named server instances, namespaces, TOML configuration, streaming multiprocessor partition controls, and cgroup-integrated GPU memory limits for shared, containerized GPU workloads. CUDA Compute Fabric Transport lets communication-library developers address named NVLink endpoints and issue asynchronous put, get, and reduction operations without mapping every remote allocation into virtual address space; it is exposed only through the Driver API. The release also decouples driver installation from the toolkit, exposes locality and unified-memory residency queries, and changes coherent platforms to CDMM by default while retaining NUMA support.

Takeaways

1. 1

   Multi-Process Service V3 supports named server instances, namespaces, TOML configuration, streaming multiprocessor partitioning, cgroup-integrated GPU memory limits, and a scriptable CLI for shared GPU workloads.

2. 2

   CUDA Compute Fabric Transport uses named logical endpoints, endpoint IDs, and offsets for asynchronous put, get, and reduction operations over NVLink; it is available only through the CUDA Driver API.

3. 3

   CUDA SDK installers no longer bundle the NVIDIA driver. On coherent platforms, CDMM is now the default memory-management mode, while NUMA remains available through a kernel module parameter.