.jpg)
Overview of core hardware GPU
→ Introduction to CPU and GPU
CPU Central Processing Unit
Good at focusing on precise calculations in a timely manner
(Just like a PhD student can solve complex equations, but if he needs to solve 100 sets of simple addition, subtraction, multiplication and division, he is not as good as 100 elementary school students who can solve them simultaneously.)
The cores are used to handle common computer operations.
GPU Graphics Processing Unit
Good at computing large amounts of data in parallel
(Just like 100 elementary school students calculating 100 sets of simple addition, subtraction, multiplication and division at the same time, they will be faster than 1 doctoral student)
Multi-core parallelism, good at image processing (simultaneously calculating the generation of a large number of pixels) and AI computing (constantly distinguishing between a large number of correct and incorrect)
Detailed video explanation_ 3D animation reveals how the GPU of the graphics card works:
https://www.youtube.com/watch?v=6RFTS340tkk
→ GPU DGX H100
Conceptual structure diagram:
Entity structure diagram:
DGX H100 Overview:
Designed for performance, DGX H100 features separate connections for compute and storage fabrics
For AI training
→ (Invitrogen) GPU Model Introduction
Mainstream Model Evolution
Invincible GPU has undergone eight iterations of architecture adjustments from 2008 to 2022. The V100 with Tensor core has a mixed-precision computing speed that is 9 times faster than the P100, and the fourth-generation Tensor core launched by Invincible has a FP8 performance that is 16 times higher than Ampere, and in terms of AI large language model reasoning, the performance is 30 times higher than Ampere.
Cost-effectiveness calculation of various types of GPUs
Comparison of various GPU capabilities
Model 800 is a domestically produced light and fine cutting version of Model 100.
H20 is a domestically produced heavily shorn version of H200
H800 = 2.5 A800 A800 = 2.5 V100
→ Introduction to Data Center
Definition and Application
Supercomputer
Definition:
Refers to a type of computer system with powerful computing power and highly parallel processing capabilities.
It usually consists of a large number of processors, high-speed network connections, and large amounts of storage.
Main features: Ability to quickly process large-scale data and complex computing tasks. It plays an important role in scientific research, engineering simulation and commercial applications.
Functions and industry applications:
Scientific research areas: weather forecasting, climate simulation, cosmic simulation, etc. Processing large amounts of meteorological data, simulating the Earth's climate system, and simulating the evolution and structure formation process of the universe.
Engineering fields: aerospace, automobile design, materials science, etc. Aerodynamics simulation, fluid mechanics simulation, simulation of material structure and properties, etc.
Business field: Widely used in financial risk management, supply chain optimization, data analysis, etc. Risk simulation and high-frequency trading, simulation and optimization of supply chain operations, etc.
Artificial Intelligence Computing
Definition:
Intelligent computing refers to the process of using artificial intelligence technology for calculation and analysis.
It covers technologies such as machine learning, deep learning, and natural language processing.
Core idea: Simulating the process of human intelligence to enable computers to process and analyze complex data and problems.
Functions and industry applications:
In the fields of image recognition, speech recognition, natural language processing, recommendation systems, etc.: Intelligent computing can use deep learning algorithms to identify and classify objects and features in images, convert and understand human speech, and analyze user behavior and preferences.
In the fields of intelligent driving, medical diagnosis, and financial forecasting: intelligent driving uses intelligent computing technology for environmental perception and decision-making; in medical diagnosis, intelligent computing can analyze and diagnose medical images; in financial forecasting, intelligent computing can assist in financial decision-making and risk management.
Ordinary computing server
Definition:
General computing servers are server devices used to perform general computing tasks.
It usually has some processing power, memory, and storage capacity, and can provide a stable network connection.
Main function: Used to run and manage various computing programs and applications, including office applications, web browsing, multimedia playback, etc.
Functions and industry applications:
Enterprise applications: General computing servers are widely used in enterprise environments to support various enterprise-level applications such as ERP, CRM, HRM, etc.
Virtual environment: As the infrastructure of the virtual environment, it provides services such as virtual machines and containers to achieve flexible allocation and management of resources.
Network services: website hosting, email services, file transfer services, etc. They can handle a large number of network requests and provide stable and reliable services.
Application ratio
Supercomputers (about 10%)
It mainly includes scientific research institutions, universities, national laboratories, corporate R&D departments and other institutions and enterprises that need to handle large-scale scientific computing and engineering simulation.
Artificial intelligence computing (estimated 35%)
Large corporate client groups, including Internet companies, financial institutions, manufacturing, healthcare and other industries: using intelligent computing to improve business efficiency, optimize decision-making and innovate products.
Research institutes: such as the Hong Kong Innovation Institute of the Chinese Academy of Sciences, the Hong Kong Applied Textiles Institute, and the Hong Kong Applied Science and Technology Research Institute, which promote the development and popularization of artificial intelligence capabilities through the development and cooperation of scientific research projects.
Local universities and colleges: such as the eight universities in Hong Kong, can leverage the strong scientific research and technological strength and academic leadership of local universities to develop and conduct industry research on artificial intelligence and big models.
Ordinary computing power server (about 55%)
The customer base of general computing is broad, including individual users, small and medium-sized enterprises, government agencies, etc., and it is used for general computing tasks and office applications.
Comments