Nsight Compute 2026.2 - New Features
…250). For more details, refer to ERR_NVGPU . NVIDIA Nsight Compute Updated the overall layout to pin several tool windows by default. Improved CUDA Tile support on the Source page. Added a…
Search
Tracked topic
NVIDIA CUDA
People also ask
What Is CUDA?
CUDA is NVIDIA's platform for accelerated computing, providing the software layer that enables applications to harness the power of GPUs. Developers can program in languages such as C++, Python, and Fortran or leverage GPU-accelerated libraries and frameworks like PyTorch. This flexibility lets developers integrate GPU computing into any layer of their software stack to achieve optimal functionality and performance.The CUDA Toolkit, an integral component of the CUDA platform, provides the compiler, libraries, and developer tools required to develop GPU applications.
NVIDIA CUDAWhat’s CUDA All About Anyway?
Learn about the CUDA ecosystem that helps developers solve real-world challenges.
NVIDIA CUDATop stories
NVIDIA Dynamo Snapshot: Fast Startup for Inference Workloads on Kubernetes | NVIDIA Technical Blog
5d agoExtract More Kernel Performance with NVIDIA CompileIQ Auto-Tuning | NVIDIA Technical Blog
6d agoDevelop High-Performance GPU Kernels in C++ with NVIDIA CUDA Tile | NVIDIA Technical Blog
6d agoNVIDIA CUDA 13.3 Enhances GPU Development with Tile Programming in C++, Compiler Autotuning, and Python Updates | NVIDIA Technical Blog
6d agoAccelerating AI-Powered Chemistry and Materials Science Simulations with NVIDIA ALCHEMI Toolkit-Ops | NVIDIA Technical Blog
…Linux (primary), Windows (WSL2), macOS NVIDIA GPU (A100 or newer recommended), CUDA compute capability ≥ 8.0 CUDA Toolkit 12+, NVIDIA driver 570.xx.xx+ Installation To install ALCHEMI Toolkit-Ops, use the…
Automating GPU Kernel Translation with AI Agents: cuTile Python to cuTile.jl | NVIDIA Technical Blog
…Julia 1.12+ and NVIDIA CUDA 13.1+ driver NVIDIA Ampere, NVIDIA Ada, or NVIDIA Blackwell GPU (compute capability 8.x, 10.x, 11.x, 12.x) An LLM agent with file…
Optimizing Communication for Mixture-of-Experts Training with Hybrid Expert Parallel | NVIDIA Technical Blog
…Tong Liu Tong Liu is a DevTech engineer at NVIDIA, specializing in optimizing Mixture-of-Experts (MoE) large language model training and CUDA kernel development. He has contributed to key features in…
AI Infrastructure for Autonomous Vehicles
…Training and Optimizations NVIDIA offers Docker containers for your preferred deep learning framework through the NGC catalog . GPU Acceleration Enable GPU acceleration using NVIDIA CUDA ® Toolkit and NVIDIA CUDA Deep Neural Network…
NVIDIA Aerial
…It includes a runtime for executing the pipelines on NVIDIA Aerial™ RAN computer platforms. NVIDIA Aerial CUDA-Accelerated RAN NVIDIA CUDA libraries for layer 1 (L1) and layer 2 (L2) RAN, to…
Nsight Compute 2026.1 - New Features
…Fixed issues in node-level profiling of CUDA device launchable graphs. For a complete overview of all NVIDIA® Nsight™ Compute features and access to resources, please visit the main Nsight™ Compute page…
NVIDIA JetPack SDK Downloads and Notes
…Supported Linux Distros Canonical Ubuntu 24.04 for Jetson AI Compute NVIDIA CUDA® 13.0.0 CuDNN 9.12.0 NVIDIA TensorRT™ 10.13.3.9 Graphics Vulkan 1.4 Vulkan SC…
NVIDIA JetPack SDK Downloads and Notes Archive
…Supported Linux Distros Canonical Ubuntu 24.04 for Jetson AI Compute NVIDIA CUDA® 13.0.0 CuDNN 9.12.0 NVIDIA TensorRT™ 10.13.3.9 Graphics Vulkan 1.4 Vulkan SC…
Accelerating Long-Context Model Training in JAX and XLA | NVIDIA Technical Blog
…These collectives leverage SHARP, in-network reductions, and multicast acceleration features of NVIDIA NVLINK Switch to enable latency-optimized one-shot and throughput-optimized two-shot AllReduce algorithms. The underlying CUDA interface…