Loading timeline…
20062000s
GPU Computing for ML
Utilizing graphics processing units for massive parallel mathematical computation.
Why It Was Important
Neural networks require thousands to millions of simultaneous matrix multiplications—which CPUs handle sequentially. Nvidia's release of CUDA allowed researchers to write general-purpose code for GPUs, which have thousands of cores. This slashed model training times from months to days, making deep learning practical for the first time.
Who Invented It
NVIDIA / AI Researchers
Hardware engineers responding to the needs of the gaming and scientific communities.
Applications
- Model Training
- Scientific Simulation
- Cryptocurrency
- Generative AI
Key Papers
- Large-Scale Deep Unsupervised Learning Using Graphics Processors
Rajat Raina, Anand Madhavan, Andrew Y. Ng · ICML 2009
Videos
GPUs: Explained
IBM Technology
Stanford Seminar - NVIDIA GPU Computing: A Journey from PC Gaming to Deep Learning
Stanford Online