Graphics Processing Unit / GPU
The type of processor widely used in parallel calculation, artificial intelligence training.
Graphic Processing Unit (GPU) is a kind of processor that has become the indispensable equipment of artificial intelligence training today, designed to process images used in computer games and graphic rendering processes quickly. While CPUs are complex with a small number of powerful cores, successful in queued tasks, GPUs have thousands of simple draws, and these cores can simultaneously run in parallel with numerous data. This parallel structure is extremely convenient, since deep learning is formed from huge matrix and tensor palates on the basis; millions of bumps in a nervous network layer can be carried out simultaneously on the GPU.
The NVIDIA’s CUDA platform played a key role in making deep learning in the explosion in 2010s by enabling GPUs to program for general purpose calculation; success of leading models such as AlexNet has been largely possible with GPU acceleration. Today, the education of large language models requires thousands of GPUs to run months by connecting together in data centers, so the GPU supply and price have become a strategic bottle of the artificial intelligence industry. The chips such as NVIDIA H100 and A100 are considered the industry standard in the education of today’s large models.
The NVIDIA’s CUDA platform played a key role in making deep learning in the explosion in 2010s by enabling GPUs to program for general purpose calculation; success of leading models such as AlexNet has been largely possible with GPU acceleration. Today, the education of large language models requires thousands of GPUs to run months by connecting together in data centers, so the GPU supply and price have become a strategic bottle of the artificial intelligence industry. The chips such as NVIDIA H100 and A100 are considered the industry standard in the education of today’s large models.