Tensor Core

NVIDIA detailed its new "Rubin" GPU architecture, which delivers 50 PetaFLOPS at sparse NVFP4 operations, features 224 streaming multiprocessors, 288 GB of HBM4 memory, 336 billion transistors, and 89…

Researchers at arXiv analyzed the performance of Med-DDPM, a 3D generative diffusion model for MRI synthesis, across three generations of NVIDIA GPUs. They identified inefficiencies in memory access a…

Loop unrolling, a classic compiler optimization, is experiencing a resurgence in the machine learning era as a critical technique for maximizing throughput on modern compute architectures like SIMD ve…