NVIDIA Releases CUDA 13.5 with General Availability for Vera Rubin Architecture
In short: NVIDIA has launched CUDA 13.5, reaching general availability for its new Vera Rubin GPU architecture. This update introduces significant enhancements to multi-process service (MPS) resource scheduling, locality domain partitioning, and deeper integration with SIMT libraries via CUDA Tile.
This summary was generated automatically by AI from NVIDIA Developer's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1General availability for NVIDIA Vera Rubin GPU architecture.
- 2New locality domain flags for supplemental SM allocation.
- 3MPS v3 introduces weighted timeslicing, load balancing, and cgroup-based routing.
- 4CUDA Tile now supports cross-GPU communication on NVLink/NVSwitch and SIMT library interop.
- 5CUDA Python 1.2 adds batched 1D memory copies and flexible graph updates.
| Parameter | Before | Now |
|---|---|---|
| Vera Rubin Architecture | Early developer access | General availability (GA) |
Why it matters
These updates allow for better performance density in data centers by providing tools to squeeze more utility out of each GPU partition, especially when running multiple concurrent AI models on the same hardware.
What it means for AI agents and contact centers
If your infrastructure runs high-throughput inference, test the new MPS v3 weighted timeslicing to prioritize latency-sensitive voice agent workloads. Also, evaluate the CUDA Tile SIMT interop to see if custom kernels can achieve better throughput when processing audio streams.
🧪 Worth Testing
Testing is recommended for optimizing GPU utilization and resource isolation in multi-tenant inference environments.
Sources
- NVIDIA DeveloperOfficialPrimary sourceOriginal article →„NVIDIA CUDA 13.5 Brings NVIDIA Vera Rubin GA Support“4 Nov 2026, 20:46
- Published by source
- 4 Nov 2026, 20:46
- Found by our system
- 4 Nov 2026, 21:06
- Summary generated
- 4 Nov 2026, 21:06
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy