Knowledge about graphics cards and gpu

Graphics cards and GPUs

GPU is the core of the graphics card. The graphics card is composed of GPU, video memory, circuit board, BIOS firmware, etc. GPU belongs to the graphics card. The GPU is a core processing chip on the graphics card and the heart of the graphics card. It cannot be used alone as an external expansion card. The GPU consumes a lot of power due to its heavy parallel computing tasks and can only be used by welding on the circuit board of the graphics card.

There is a GPU on the graphics card, which is the most important component to distinguish the graphics performance. The graphics card is also called the display adapter , which is divided into independent graphics card and integrated graphics card on the motherboard . The independent graphics card is mainly composed of GPU, video memory and interface circuit. The integrated graphics card does not have independent video memory. It uses the memory on the motherboard . If the circuit board of the graphics card is regarded as the motherboard, the GPU is the CPU on the graphics card, that is, the central processing unit of the graphics card.

There are currently only two GPU development manufacturers, AMD(ATI) and NVIDIANVIDIA, with the latter accounting for the majority of the market.

GPU

GPU stands for Graphics Processing UnitImage Processing Unit. At the beginning, GPUs were mainly used for 3D game rendering, but now they are widely used to accelerate computational workloads, such as financial model calculations, scientific research, and oil and gas development. From an architectural point of view, the CPU is composed of several cores and many cache memories, so the CPU can process several threads in parallel. In contrast, a GPU is composed of hundreds of cores and can therefore process thousands of threads concurrently. Although the number of cores of the GPU far exceeds that of the CPU, the processing power of each core is much smaller than that of the CPU, and it does not have some features required by modern operating systems. The GPU is not suitable for processing ordinary calculations. They are used more for computationally intensive operations such as video processing and physics simulation.

GPU pass-through
is GPU pass-through. Simply put, GPU pass-through means that the virtual machine maps the physical graphics card directly to the virtual machine. In this way, the image information of your virtual machine is no longer handed over to the virtual graphics card, but to the virtual graphics card. Send it directly to your real graphics card for calculation, which is more efficient.

GPU virtualization
Virtualization is the process of slicing the graphics card and allocating these graphics card time slices to virtual machines for use. Some types of GPU devices can be virtualized into n vGPUs, and n virtual machines can directly use the GPU device at the same time. According to different GPU device models and virtualization strategies, GPU exclusive or transparent transmission can be designed accordingly. Each virtual machine can directly access some hardware resources of the physical GPU through the bound vGPU (all vGPUs can share time-sharing access to the 3D graphics engine and video codec engine of the physical GPU, and have independent video memory)

Video memory

Video memory capacity is also an important parameter of the graphics card. Conceptually, video memory capacity is the capacity of the local video memory on the graphics card, which determines the ability of the video memory to temporarily store data. Video memory is also called the frame buffer. To simply understand, when you play a game, you will first store the pictures and scenes in the video memory space.

amd graphics card

In 2006, AMD acquired ATI for US$5.4 billion. Since then, it has two good cards in CPU and GPU. Since then, AMD's share of graphics cards has continued to decline, and the gap with NVIDIA has continued to widen. The graphics card starting RXwith AMD, referred to as A card, may be more cost-effective than N card. But if you use your computer for productivity, such as graphics rendering and other graphics work mainly, and games as a supplement, then you can choose the A card.

The latest product for desktops is the AMD Radeon RX 7900 XTX desktop graphics card.

The latest notebook product is the AMD Radeon RX 7000M series mobile graphics card.

NVIDIA graphics card

N-cards starting with GTX or RTX have good support for games and have established the monopoly of high-end graphics cards in games.

The RTX30 series
GeForce RTX™ 3090 Ti and 3090 use the 2nd generation NVIDIA RTX architecture NVIDIA Ampere. This series of products is equipped with dedicated 2nd generation RT Core, 3rd generation Tensor Core, new SM multi-unit stream processor and amazing 24GB G6X video memory.

RTX40 Series NVIDIA® GeForce RTX™ 40 Series GPUs are powered
by the more efficient 3rd generation RTX architecture , delivering a leap in performance with AI-driven DLSS 3. NVIDIA Ada LovelaceThanks to the blessing of panoramic ray tracing technology, a lifelike virtual world is created. Additionally, Max-Q technology optimizes system performance, power consumption, battery life and sound for peak efficiency. Ada GPU can achieve 2 times the performance improvement of traditional rasterization games and up to 4 times the performance improvement of ray tracing games. NVIDIA has three GPUs based on the Ada Lovelace architecture, including GeForce RTX 4090 in 24GB version and GeForce RTX 4080 in 16GB and 12GB versions.

Specifications

	GeForce RTX 4090 laptop GPU	GeForce RTX 4080 laptop GPU	GeForce RTX 4070 laptop GPU	GeForce RTX 4060 laptop GPU	GeForce RTX 4050 laptop GPU
NVIDIA CUDA Core number of cores	9728	7424	4608	3072	2560
Acceleration frequency	1455 - 2040 MHz	1350 - 2280 MHz	1230 - 2175 MHz	1470 - 2370 MHz	1605 - 2370 MHz
Memory capacity	16 GB	12 GB	8 GB	8 GB	6 GB
Video memory type	GDDR6	GDDR6	GDDR6	GDDR6	GDDR6