As large-scale language models such as ChatGPT set off a new wave of AI changes, the shortage of AI chips is becoming increasingly serious. For example, ChatGPT developer OpenAI relies heavily on supercomputers provided by Microsoft and uses a large number of NVIDIA GPUs. Recently, it was reported that OpenAI is considering manufacturing or acquiring its own AI chips to solve the problem of high-performance and low-cost GPUs required for its business.
Related Articles
Today, let’s talk about what exactly is the GPU that OpenAI wants to join?
What is GPU
"A thousand mobile phones, there are a thousand gaming experiences." When we use different mobile phones to play games, the experience is different. In addition to the response speed, the exquisiteness and three-dimensionality of the game screen are also major differences, resulting in this One of the factors that make this difference is the performance of mobile phone GPUs.
GPU (Graphics Processing Unit), as a superhero hidden in mobile phones and computers, is not only a professional painter who can quickly draw colorful pictures but also a mathematician who can quickly complete a large number of tasks. Computational tasks.
Technically speaking, a GPU is a processor specifically designed to process graphics. It can handle a large number of graphics rendering calculations at a very fast speed and can handle multiple graphics tasks at the same time, thereby greatly increasing the Computing and processing speed of the computer.
GPUs were originally designed for graphics processing, but due to their parallel processing and high-speed computing capabilities, GPUs began to take on more and more important roles. GPUs are now widely used in scientific computing, machine learning, big data analysis, and other fields.
GPUs in computers can be divided into two types, integrated GPUs and independent GPUs.
Integrated GPU is small in size is generally built into the computer motherboard, and can even be integrated into the CPU. Integrated GPUs can make computers lighter and are often found in laptops.
The independent GPU is larger and is an independent component with a dedicated socket on the computer motherboard. It is more powerful than an integrated GPU and can be upgraded separately (replacing the graphics card). However, due to its larger size, it takes up more space on the computer, consumes more energy, and generates more heat when running.
Some laptops have both types of GPU. Generally, the integrated GPU is used to save energy and reduce heat generation. When more powerful graphics processing performance is needed, the independent GPU is switched to perform related tasks.
What are the differences between GPU and CPU
Seeing this, have you thought of another important role in the computer, the CPU (Central Processing Unit, central processing unit)? So, what is the difference between GPU and CPU?
Although both can perform computing tasks, their capabilities vary. If the CPU is a knowledgeable mathematics professor who can solve any problem, then the GPU is 10,000 elementary school students. With more people and more power, they can calculate simple mathematics problems extremely quickly.
In fact, before the advent of the GPU, basically all tasks were completed by the CPU. After the GPU became available, the two divided their labors. The following table lists the differences between the two.
Through the above comparison, we found that GPU and CPU have their own advantages. In mobile phones and computers, they cooperate with each other, work together, and serve us together.
GPU is more suitable for AI
Through the previous introduction, we know that GPU is very suitable for large-scale parallel computing. The training in AI (Artificial Intelligence, artificial intelligence) involves a lot of data processing, especially in the field of deep learning. Network models usually have millions or even billions of parameters and need to be trained with a large amount of data to obtain accurate results. prediction, so GPU is very suitable for AI algorithms.
Parallel processing capability
GPUs have a large number of cores and high-speed memory and are good at parallel computing. In the field of AI, the amount of calculation is very large, and GPU is just suitable for this scenario. For example, when a large number of simple math problems need to be calculated, 10,000 primary school students are definitely more suitable than one professor.
Greater memory bandwidth
Some common GPU memory bandwidths are around 400 GB/s, while the best CPU memory bandwidths are around 50 GB/s, so the GPU can fetch and access data in memory faster. In the field of AI, data generally occupies large contiguous memory spaces, and GPU is obviously more suitable.
High flexibility
GPU supports the use of programming frameworks and languages such as CUDA and OpenCL, allowing developers to easily utilize the computing power of GPU, highly customize GPU computing tasks, and provide support for different types of AI algorithms.
CUDA
Compute Unified Device Architecture, a general-purpose parallel computing architecture launched by NVIDIA, enables GPUs to solve complex computing problems.
OpenCL
Open Computing Language, an open design language, is an open standard for cross-platform parallel programming of various accelerators in supercomputers, cloud servers, personal computers, mobile devices, and embedded platforms.
Strong scalability
As the complexity of AI models increases and the amount of data grows, we can increase processing capabilities by adding more GPUs, just like adding more primary school students for calculations, so that the system can better cope with the growing calculations. need.
