Cache: CPU has a large number of cache structures. At present, mainstream CPU chips have four levels of cache, which consumes a lot of transistors and needs a lot of power when running. The cache of GPU is simple. At present, mainstream GPU chips have at most two layers of cache, and GPU is more efficient than CPU.

Response mode: CPU requires real-time response and high speed for single task, so multi-layer cache is needed to ensure the speed of single task. GPU arranges all tasks and then processes them in batches, which requires relatively low cache.

Floating-point operation mode: CPU is not only responsible for floating-point shaping operation, but also has many other instruction set loads, so the power consumption allocated by CPU to floating-point calculation will be less. GPU basically only does floating-point operations, so it has a simple design structure and can do it faster.

Application direction: CPU is good at operating system and other applications, and needs to respond to real-time information quickly and optimize the delay, so the control parts such as branch prediction, out-of-order execution and low-delay cache need the number of transistors and energy consumption. GPU is suitable for architecture operations with high predictability, large amount of similar operations, high latency and high throughput.

Introduction to gpu and cpu

GPU: It is called graphics processor, also called display core, visual processor and display chip. It is a microprocessor, which is specially used for image and graphics related operations on personal computers, workstations, game consoles and some mobile devices (such as tablets and smart phones). It can be vividly understood as 90% ALU (arithmetic unit), 5% control unit and 5% cache unit.

CPU: As the computing and control core of a computer system, it is the final execution unit of information processing and program running, which can be vividly understood as having 25% ALU (computing unit), 25% control unit and 50% cache unit.