The GPU was originally made to draw game graphics. Because it has to calculate the countless dots on a screen at the same time, it is strong at doing many simple calculations all at once.
AI training and inference also repeat the same kind of calculation an enormous number of times, so GPUs were a perfect fit. Today tens of thousands of GPUs built just for AI are gathered in one building to make large models. Nvidia is the best-known company.
ឧទាហរណ៍ប្រចាំថ្ងៃ
If one professor solving a hard problem is the CPU, thousands of students doing easy calculations at the same time are the GPU. For AI, the second is much faster.
ហេតុអ្វីវាសំខាន់
The AI race is also a race over how many GPUs you can get. Chip shortages, export controls and data center power problems are all tangled up in it.