CPU Talent Processor, GPU Into AI Chip Synonym

AI boom is delayed, Nvidia surpasses Intelide Data CenterchipDuring the faucet, the GPU is synonymous with AI chip, and the difference between the CPU has also received attention.

Analysts saidThe CPU is a "talent type" processor, but it is difficult to fight a lot of trivial work. The GPU can handle a large number of simple work at the same time, which is more suitable for AI application situations.

The central processor (CPU) and the drawing processor (GPU) are processors, which are key operational engines that can process data.However, the architecture of the CPU and GPU is different and for different goals.

CPUIt is by millions晶体管It may have multiple processing cores, which are usually called the brain of the computer. They are responsible for executing the instructions and programs required by the computer and the operating system. Theoretically, it can complete any operational work. It is a "talent type" processor.

GPUIt is composed of many smaller and more professional cores. It is good at handling simple special tasks. It is mainly used in computer image processing. It is a "specialty type" processor.

Zhong Yingting, a municipal adjustment agency, said that the two handles messages in different ways. The CPU is a serial type, and the GPU is parallel to computing, suitable for different application situations.The so -called serial processing refers to completing one job at a time; parallel operations divide one job into many different steps and allocate them to multiple cores to accelerate the instruction cycle.

The CPU frequency speed is high and can handle very complicated computing instructions, but when a large number of trivial tasks are delivered, although it can be processed, it will waste too much time.For example, the Amazon online service company is like a chef in a restaurant that can roll over hundreds of burgers, but it takes a lot of time; if you hand over the task to the assistant to have many hands, that is, the GPU, you can quickly complete it.Essence

Analyst Wang Zhaoli said that the current AI application is mainly based on deep learning. The algorithm will use a large number of parallel operations. It is a more suitable application situation of GPU. Therefore, in the AI era, the importance of GPU will be much too CPU.

Intel and AMD are the main suppliers of the CPU, and Nvida is the GPU leader.As Microsoft and other large manufacturers scramble to grab the AI field, Nvidia is considered the biggest beneficiary of this wave of AI boom.

Wang Zhaoli pointed out that because the GPU is not cheap and the energy efficiency is poor. If the application scenarios are clear and the demand is large, the special application chip (ASIC) that can develop a relatively simple development function will be more power -saving and the efficiency will be better.

With the expansion of AI applications, including smart factories, smart cars, security, etc., Wang Zhaoli expects that the future AI application will also expand to terminal devices such as computers and smartphones.专用AI chips will increase.

Because AI chips are mostly advanced semi -conductive system technology, the cost is high, and once the algorithm changes, the chip may not be applicable, with high risk.Wang Zhaoli believes that this is a place where manufacturers who intend to develop chips need to evaluate consideration.