CUDA
ParallelKernelBench Exposes LLM Weakness in Multi-GPU Kernels
ParallelKernelBench shows GPT-5.5 and peers struggle with multi-GPU CUDA kernels, solving less than 31% of tasks. Here's why it matters.
NVIDIA Introduces CCCL Runtime to Modernize CUDA Development
NVIDIA's CCCL Runtime brings modern C++ abstractions to CUDA, enabling safer, more efficient GPU programming for developers.
Nvidia's New MoE Kernels Promise 93% Speedup for AI Training
Nvidia unveils advanced MoE training kernels, boosting AI model throughput by up to 93% in GPT pre-training and redefining large-scale efficiency.
NVIDIA CUDA 13.3 Brings Tile Programming to C++
NVIDIA CUDA 13.3 introduces tile-based GPU programming in C++, optimizing Tensor Core use and simplifying kernel development.
NVIDIA Unveils CompileIQ to Maximize GPU Kernel Performance
NVIDIA's AI-powered CompileIQ optimizes GPU kernel performance using evolutionary algorithms, enabling up to 15% gains in critical AI workloads.
NVIDIA CUDA 13.3 Boosts GPU Programming with Tile C++ and Python
NVIDIA CUDA 13.3 introduces Tile C++ programming, Python updates, and CompileIQ, delivering up to 15% kernel speedups and enhancing GPU development.
NVIDIA Nsight Tools Slash Vision AI Decode Times by 85% in New VC-6 Batch Mode
NVIDIA's optimized VC-6 batch mode achieves submillisecond 4K image decoding, delivering up to 85% faster per-image processing for AI training pipelines.
NVIDIA Releases CUDA Tile for BASIC in April Fools Joke With Real Tech
NVIDIA's cuTile BASIC announcement showcases CUDA Tile's language-agnostic design while poking fun at legacy code. The underlying tech is genuinely significant.
NVIDIA CUDA 13.2 Update: Latest CUDA News Today (Ampere & Ada GPUs)
CUDA 13.2 extends tile-based GPU programming to older architectures, adds Python profiling tools, and delivers up to 5x speedups with new Top-K algorithms.
CUDA News Today: NVIDIA Brings CUDA to Third-Party Platforms
NVIDIA now allows developers to access CUDA via third-party platforms, simplifying software deployment and integration across various OS and package managers.
NVIDIA CCCL 3.1 Adds Floating-Point Determinism Controls for GPU Computing
NVIDIA's CCCL 3.1 introduces three determinism levels for parallel reductions, letting developers trade performance for reproducibility in GPU computations.
NVIDIA Brings CUDA Tile Programming to Julia with cuTile.jl Release
NVIDIA releases cuTile.jl, enabling Julia developers to write high-performance GPU kernels using tile-based programming with near-parity Python performance.
NVIDIA cuda.compute Brings C++ GPU Performance to Python Developers
NVIDIA's new cuda.compute library topped GPU MODE benchmarks, delivering CUDA C++ performance through pure Python with 2-4x speedups over custom kernels.
NVIDIA Integrates CUDA Tile Backend for OpenAI Triton GPU Programming
NVIDIA's new CUDA Tile IR backend for OpenAI Triton enables Python developers to access Tensor Core performance without CUDA expertise. Requires Blackwell GPUs.
NVIDIA CUDA 13.1 Drops CUB Boilerplate with New Single-Call API
NVIDIA simplifies GPU development with CUB single-call API in CUDA 13.1, eliminating repetitive two-phase memory allocation code without performance loss.
