top of page

Terms and Definitions

GPU terminology:

Anti-Aliasing: Technique reducing jagged edges in graphics by smoothing pixel colors along object edges.

API (Application Programming Interface): A set of functions and protocols allowing software applications to communicate with the GPU. Widely used APIs include DirectX, Vulkan, and OpenGL.

Async Compute: Feature enabling simultaneous execution of compute and graphics tasks, enhancing GPU efficiency.

Bit Depth: Number of bits used to represent color in an image, affecting color accuracy and quality.

CUDA (Compute Unified Device Architecture): NVIDIA's parallel computing platform and API, enabling developers to harness GPU power for general-purpose computations.

CUDA Cores: Individual processing units within an NVIDIA GPU, performing parallel calculations for enhanced performance.

Deep Learning: Subset of machine learning employing neural networks to analyze and learn from large datasets.

DirectX: Collection of APIs by Microsoft, allowing software, especially games, to interact with GPUs for rendering and multimedia.

Driver: Software facilitating communication between the operating system and the GPU, crucial for efficient GPU operation.

FLOPS (Floating-Point Operations Per Second): Measurement of GPU's floating-point performance, indicating its computational speed.

Frame Buffer: Segment of GPU memory storing the current frame displayed on the screen.

Geometry Shader: Shader processing geometric data, generating new vertices for creating intricate shapes and effects.

GPU (Graphics Processing Unit): Specialized electronic circuit accelerating graphics rendering through complex calculations.

HBM (High Bandwidth Memory): Memory technology with high data transfer rates and low power consumption, found in modern GPUs.

Integrated GPU: GPU integrated into the CPU chip, suitable for basic graphics tasks with lower power consumption.

Jitter: Random variations in GPU operation timing, potentially causing visual artifacts or reduced performance.

Luminance: Brightness of an image, processed by GPUs to create well-lit scenes.

Mipmap: Set of progressively smaller pre-rendered textures optimizing rendering performance at various distances.

Multi-GPU Configuration: Utilizing multiple GPUs in tandem for increased graphics processing power.

Noise: Random data used in procedural generation, creating natural textures and patterns in graphics.

Overclocking: Increasing GPU clock speed and voltage for improved performance, requiring proper cooling solutions.

PCIe (Peripheral Component Interconnect Express): High-speed interface connecting GPUs and expansion cards to motherboards.

Pixel: Smallest unit of a digital image, processed by GPUs to render graphics.

Rasterization: Process converting vector graphics to raster images composed of pixels for real-time rendering.

Ray Tracing: Technique simulating light behavior for realistic graphics by tracing light rays interacting with objects.

Rendering: Creating images from 3D models, textures, and lighting, GPUs excel at rendering due to parallel processing.

Shader: Program executed on the GPU, responsible for tasks like vertex manipulation, pixel shading, and geometry processing.

Tensor Cores: Specialized hardware for accelerating machine learning tasks through efficient tensor calculations.

Texture Mapping: Technique applying images (textures) to 3D surfaces for realistic and detailed visuals.

Unified Memory: Architecture allowing GPU and CPU to access the same memory pool, simplifying data management.

VRAM (Video Random Access Memory): Dedicated GPU memory storing textures, frame buffers, and other data for rendering.

V-Sync (Vertical Synchronization): Syncing frame rate with monitor refresh rate to prevent screen tearing and produce smoother visuals.

Warp: Group of threads in a GPU executing instructions concurrently, optimizing parallel processing.

X-axis: Horizontal dimension in a coordinate system, used for object positioning and manipulation.

Z-buffering: Technique managing object depth in 3D rendering, ensuring accurate scene representation.

Zero-copy Memory: Memory management technique sharing GPU and CPU memory space, reducing data transfers and improving performance.

bottom of page