What are CPUs?
A central processing unit (CPU) is a chip that acts like the brain of a computer, managing all of the tasks that keep a system going, from running programs to passing instructions to hardware. A CPU does this with a handful of “cores” — parts of the chip that quickly process data and handle instructions.
How are CPUs different from GPUs?
Graphics processing units (GPUs) were created to display 3D images in games or videos. Compared to CPUs, GPUs have less powerful cores but more of them, enabling parallel processing. That means GPUs can do a bunch of stuff at once, like drawing all the lines and pixels that make up one 3D graphic. These are the chips that have been in high demand for training AI.
Why are graphics chips used for AI training?
People figured out that a GPU’s parallel processing is really useful for other tasks, including AI, which requires a lot of connections between different data points to be made simultaneously. A CPU, on the other hand, would have to do those tasks one at a time.
So GPUs are better than CPUs for AI?
Yes, but you need both. A system can’t run without a CPU directing things, and a sluggish one won’t be able to handle the AI tasks a GPU throws at it. But some CPUs are being built with special processors that can handle the “inference” stage of an AI task — basically, the GPU handles most of the work, then tosses it to the CPU to give the final product to the user.