1. First of all, AI chips have advantages in algorithmic operations. AI tasks are characterized by a large amount of computation, which requires frequent invocation of neural network models for computation. Compared with traditional processors, AI chips can perform parallel operations, greatly improving the efficiency of the algorithm. At the same time, for the specific algorithm model, AI coreThe chip architecture can be customized to further improve the efficiency of the algorithm.
2, performance and traditional chips, such as CPU, GPU is very different. Faster and more energy efficient when executing AI algorithms. The speed of ordinary chips is slow, the performance is low, can not be practically commercial. In the process of uploading and downloading ordinary chips, it is entirely possible to have data leakage problems. The ai chip can calculate on the mobile phone terminal, without the need to upload to the cloud, avoiding the risk of data leakage.
3, computing power and energy efficiency ratio: AI chips in the processing of large-scale data and complex calculationsTasks (such as deep learning algorithms) have higher computational performance. Often, multiple high-performance computing units, such as Gpus (graphics processors) or specialized neural network processors (Npus), are integrated to meet specific needs such as fast implementation of matrix multiplication and vector addition to better support machine learning and deep learning algorithms.
4, ordinary chips: may include general purpose computing units such as cpus and Gpus, but not necessarily optimized for AI tasks. AI chips: Due to their dedicated hardware, they are usually able to provide efficient computing at lower energy consumption, which is useful for mobile devices andServers that need to run for a long time are very important. Normal chips: May be less energy efficient for AI tasks because they use more general purpose computing resources.
5, if the CPU is used for training, the CPU core is small, the training time is long; The multi-core advantage of the GPU comes into play at this time. Therefore, those who play deep learning, when training, borrow the advantages of multi-core, parallel processing of Gpus, and use Gpus in non-graphics fields. Fpgas also have the advantage of parallel processing and can also be designed as hardware with multi-core characteristics.
This will directly affect the domestic computing infrastructure support, as well as the Internet, cloud services, games and other fields. The National Supercomputing Center, known as the "Mount Everest of computing," has ultra-high computing power, much of it powered by Nvidia. Many Internet vendors use Nvidia and AMD Gpus.
This behavior will affect the normal science and technology research and development of domestic science and technology enterprises, and may even affect the actual progress of cutting-edge technology. Strictly speaking, the United States does not restrict the export of only two chips, but restricts Nvidia and AMDThe company's high-end chips.
For the global chip market, almost all chips need to be supplied by Nvidia and AMD. In this case, if Nvidia and AMD stop exporting high-end chips to China, it basically means that the middle and high-end areas of many of our industries will be affected accordingly, and our domestic GPU chips will be difficult to achieve technical blockade in a short time. Nvidia and AMD stopped exporting high-end chips to China.
The United States has asked AMD and Nvidia to cut off chip supplies to China. This is a bang for the chip industryMoving news, because the United States requires AMD and Nvidia to cut off high-end chips to China, so this matter has directly led to a corresponding negative impact on many industries in China.
1. AMD is a manufacturer of CPU, motherboard chip, graphics card chip and other products. AMD CPU must use AMD chip motherboard. The merchant was right. Whether the motherboard and CPU are compatible, in addition to looking at the chip, it also requires the CPU interface to be consistent with the CPU socket of the motherboard. Your CPUModel number? New motherboard model? No matter which brand of motherboard, there are only three chips: Intel, AMD, NVIDIA. There are only two major brands of cpus: Intel and AMD.
2, AMD special bar is the walled bar, the old bad PCB version and memory particles are assembled by themselves, called AMD special bar because Intel's memory detection is strict, such a memory bar is generally not supported, and AMD's not strict, you can use. There is only one useful information in these photos, which is the big Intel on the CPU radiator, indicating that it is Intel's CPU andMotherboard, can not use the AMD special bar.
3, "A" is the first letter of "AMD", AMD company is still called "AMD" in the Chinese world, nicknamed "massage shop" (take the initial letter of pinyin). Strictly speaking, "Crossfire (CF)" refers to the A card, and "Speed force (SLI)" refers to the N card, both of which are technologies that allow more than two independent graphics cards to process graphics data in parallel, thereby improving graphics processing performance.