Purchasing a new GPU (graphics processing unit) can be quite an exhilarating journey, but it’s essential to grasp how it operates first. Although people often mix up the terms GPU and graphics card, they are not technically synonymous. In this guide, we will break down the intricacies of these components in a way that is easy to comprehend. Keep reading to enhance your understanding.
A standard graphics card houses a GPU chip, commonly referred to as a GPU die, which is soldered onto a circuit board and flanked by memory modules. Together, these components are encased within a heat sink, cover, and fan, forming the complete graphics card assembly.
Major manufacturers like AMD and Nvidia produce different models within each GPU generation by utilizing multiple GPU dies, allowing for tailored specifications. The identification of a GPU die usually follows a system where a lower number signifies a more powerful die. For instance, Nvidia’s AD102 die powers the high-end RTX 4090, while the RTX 4080 and RTX 4080 Super are based on the AD103 die. Newer models tend to use refined manufacturing processes, resulting in enhanced performance without increasing physical size.
In technical terms, the GPU refers to the graphics processor itself. It becomes a graphics card only when combined with other components.
IDG
The processing units within a GPU are often called cores, which may sound confusing since they don’t directly compare to the physical cores found in a CPU. Instead, GPUs feature thousands of smaller processing elements grouped into clusters, known as compute units.
Each manufacturer has its own terminology for these compute units—AMD refers to them as compute units, Intel calls them Xe cores, and Nvidia uses the term stream multiprocessors. Generally, faster GPUs have a higher number of CUDA cores (Nvidia) or stream processors (AMD) than their slower counterparts. However, it’s vital to note that these specifications are only comparable among models from the same brand and generation.
Most PC users will recognize VRAM (video RAM) as a critical specification. VRAM serves as the graphics card’s memory, responsible for storing the frame buffer, texture data, and other graphical information.
If the settings of your games and video resolution exceed your GPU’s VRAM capacity, this could lead to graphical glitches and decreased performance. Beyond mere size, factors like memory type (for instance, GDDR6 or GDDR6X), memory bus width (commonly 384, 256, or 128 bits), and memory bandwidth (measured in GB/s) are all key considerations.
Graphics cards are intricate devices with numerous specifications. Understanding GPU terminology is crucial for making meaningful comparisons.
IDG
Similar to CPUs, GPUs have clock rates for both their cores and memory. The core clock rate indicates how swiftly a GPU can execute tasks, while the memory clock rate determines the speed at which the VRAM communicates with the GPU. As a rule of thumb:
Higher clock rates for both the core and memory translate to better performance. You can even adjust these clock rates through a process known as overclocking.
A GPU also has a maximum power draw known as TGP (Total Graphics Power). This figure, measured in watts, indicates the maximum amount of power a graphics card can consume during use. TGP is a good indicator of how power-hungry a card is, as well as what temperatures to expect under heavy workloads like gaming.
FP32, or Floating-Point Single Precision, is a method of gauging a GPU’s theoretical performance. It reflects the number of floating-point operations a GPU can handle per second and is quantified in TFLOPS.
This metric is helpful for comparing different GPUs, even across varying generations, but it doesn’t encompass all performance factors. Other critical specifications include the supported PCIe generation and compatibility with graphical APIs like DirectX, Vulkan, and OpenGL—important for creatives and developers alike.
When it comes to software features, technologies like Nvidia’s DLSS, AMD’s FSR, and Intel’s XeSS differ in performance and support for gaming applications, so make sure to verify which upscaling technology your GPU supports and its respective version.
The graphics memory (shown here in green) is strategically placed next to the GPU die to minimize communication delays.
IDG
When selecting a new GPU, it’s wise to rely on genuine benchmarks and real-world tests from credible sources. Pay attention to performance metrics in games and applications you plan to use. Exploring reviews on platforms like Reddit, YouTube, and reputable tech publications can also inform your decision-making process.