I don't know which company is the hottest this year, but the hottest in these two months should be NVIDIA. Others play with AI to train large models, and he quickly sells the market value to trillions of dollars by selling shovels that train large models. This article will take you and myself to break the stereotype that NVIDIA only does game graphics cards, and re-acquaint yourself with this legendary company~
Knowledge refinement
GPU: Abbreviation for Graphics Processing Unit, simply put, graphics card. The original graphics card only did the role of image output, and in 1999, the name of GPU officially appeared, and gradually began to develop into a processor for image processing and high-performance computing.
CPU: The central processing unit is the core of computer computing and control, and the initial graphics processing function is also completed by the CPU.
DPU: Short for Data Processing Unit, that is, Data Processor. CPUs are mainly used for general-purpose computing, GPUs are used to accelerate calculations, and DPUs are used for data processing.
T&L: Short for Transform and Lighting, that is, Light and Shadow Transformation. This technology allows us to see dynamic and realistic three-dimensional stereoscopic images. In 1999, NVIDIA first separated the technology from the CPU and integrated it into the GPU.
3dfx: is a company founded in 1994 to develop graphics cards and chips, and its graphics cards have spawned classic games such as Need for Speed and Diablo. It was acquired by NVIDIA in 2000.
CUDA: The computing platform launched by NVIDIA in 2006 makes it easier and more efficient for developers to program on the basis of GPUs to solve problems in various fields such as image processing and computing.
RTX: The image processing technology launched by NVIDIA, which integrates ray tracing, rasterization, artificial intelligence and other technologies, is used in a large number of graphics cards.
Ray tracing: is an algorithm in computer graphics that simulates the reflection and refraction of light to make objects more realistic. Widely used in games.
Rasterization: Similar to ray tracing, rasterization is also a type of image rendering algorithm. It mainly generates images by projecting the three-dimensional model into two dimensions, which can create richer visual effects.
DLSS: Short for Deep Learning Super Sampling, or Deep Learning Super Sampling. Simply put, it renders low-resolution pictures to higher resolution through AI algorithms, and games such as "Cyberpunk 2077" and "Fortnite" use this technology.
Rendering: refers to the process of generating digital images from models in a three-dimensional scene according to set parameters such as light and shadow and materials.
NVLink: A communication protocol developed by NVIDIA that can connect multiple CPUs or GPUs. Can be used in data centers or supercomputers.
A100: The full name is NVIDIA A100 Tensor Core GPU, which is the core product of NVIDIA's data center. ChatGPT uses more than 10,000 A100 graphics cards.
H100: is an upgraded version of the A100 graphics card, released in 2022, the performance is a step up compared to the previous generation. The single sheet sold for more than $40,000.