If you are buying a new Graphics Processing Unit (GPU), you must have a clear understanding of how everything works. Although the terms GPU and Graphics Card are often used interchangeably, they are not the same thing technically. We will review all the technical information and break it down in an easy-to-understand way for you. Keep reading to learn more.
More reading: 4 Things to Consider Before Buying a Graphics Processing Unit
A typical graphics card consists of a GPU chip soldered to the circuit board surrounded by memory modules. This is integrated into the heat sink, cover, and fan, resulting in the actual graphics card.
Companies like AMD and Nvidia use multiple GPU templates in each generation to create different models and customize specifications as needed. Often, assigning a GPU template follows a system where the lower number in the specifications sheet refers to a larger and more powerful template. For example, Nvidia’s RTX 4090 flagship graphics card uses the AD102 GPU, while the RTX 4080 and RTX 4080 Super use the AD103 GPU. The latest GPUs use smaller manufacturing processes to deliver more performance in the same physical space.
The computing units in a GPU are often referred to as cores. However, this is misleading, as GPU cores do not correspond to the physical cores of a Central Processing Unit (CPU). GPU cores consist of thousands of small cores or processing elements, which are assembled into groups and then into computational units.
These computing units are classified differently by each GPU manufacturer – Compute Units (AMD), Xe Cores (Intel), and Streaming Processors (Nvidia). Faster GPUs have more Cuda cores (Nvidia) or Stream Processors (AMD) compared to slower models. However, these specifications are only comparable within the same generation and company.
Video Random Access Memory (VRAM) is a specification familiar to almost every PC user. VRAM is the memory available on the graphics card to store frame buffers, texture data, and other graphic information.
If your game settings and video resolution use up all the VRAM on your GPU, it can lead to graphic errors and performance drops. In addition to VRAM size, there are other crucial specifications such as memory type (e.g. GDDR6 or GDDR6X), memory bus width (384, 256, or 128 bits), and memory bandwidth (in gigabytes per second).
Just like a Central Processing Unit (CPU), a Graphics Processing Unit also has a clock rate for both the GPU and memory. The core clock rate refers to how fast the GPU processes instructions. On the other hand, the memory clock rate determines how fast the VRAM sends or receives data to the GPU. In general:
The higher the core clock rate and memory clock rate of a graphics card, the better the performance. You can also adjust the core clock rate and memory clock rate, known as overclocking.
Similar to a CPU, a GPU has a maximum power consumption, referred to as Total Graphics Power (TGP). TGP indicates the maximum power consumption of the graphics card during operation and is measured in watts. TGP is an indicator of how much power your graphics card is using and also provides information about the expected temperature of the GPU during intense gaming or other GPU-intensive tasks.
FP32, or Floating-Point Single Precision, is a mathematical method to measure the theoretical performance of a GPU. It refers to the number of floating-point operations a GPU can execute in one second and is measured in TFLOPS.
FP32 is useful for comparing different GPUs, even across generations, although the method does not cover all aspects of performance. Other specifications like supported PCIe generation and compatibility with graphics APIs like DirectX, Vulkan, and OpenGL are also important, especially for designers or developers.
Software features such as Nvidia’s DLSS, AMD’s FSR, and Intel’s XeSS vary in performance and game support. So, you should check which upgrade technology and version your GPU supports.
When buying a new GPU, always consult tests and realistic standards from reliable sources to make an informed decision. Check out performance numbers in games and applications you intend to use. Consider looking at Reddit posts, YouTube benchmarks, and reviews from trusted publications.
This article originally appeared on our sister publication PC-WELT and has been translated and localized from German.
