GPU Benchmark Glossary

Average of worst 0.1% frame times, identifying extreme performance outliers. Critical metric for competitive gaming and professional applications requiring consistent frame delivery without severe drops.

Average of worst 1% frame times, representing minimum performance under load. More meaningful than average FPS for evaluating user experience. Low 1% FPS indicates performance dips that cause noticeable stuttering.

Dynamic technique that adjusts ray marching step size based on volume density. Higher density regions use smaller steps for accuracy, while empty spaces use larger steps for performance. Critical for optimizing volumetric rendering efficiency.

Lighting approximation computing how exposed each point is to ambient lighting. In volumes, calculated by casting additional sample rays to determine density accumulation. Adds realism but significantly increases computational cost.

GPU feature executing compute shaders concurrently with graphics workload. Can be used for volume data preprocessing or denoising while main rendering pipeline runs. Improves overall GPU utilization.

Physical simulation of light interaction with atmospheric particles. Creates realistic sky, fog, and volumetric lighting effects in games and simulations. Computationally intensive, requiring optimized volume rendering techniques.

Simulation and visualization of fluid flow using numerical methods. Results are volumetric scalar or vector fields requiring real-time rendering. GPU performance in volume shaders determines interactive exploration capabilities.

AMD GPU term for processing cluster containing multiple stream processors. Equivalent to NVIDIA's Streaming Multiprocessor (SM). Each CU executes shader warps in parallel. More CUs generally mean better volumetric rendering performance.

Technique using signed distance functions (SDF) to represent 3D geometry mathematically. Enables efficient ray marching by providing exact distance to nearest surface, allowing larger step sizes without missing details.

Hardware feature testing fragment depth before fragment shader execution. Pixels occluded by previously rendered geometry are discarded early, saving computation. Proper depth sorting maximizes early-Z benefits in volumetric scenes.

Optimization method that rapidly traverses empty regions in volumetric data without sampling. Uses auxiliary data structures to identify and skip regions with zero density, dramatically improving render times.

Statistical measure of frame rendering time consistency. High variance indicates stuttering or unstable performance. Volume Shader BM tracks variance to assess thermal throttling and driver stability issues.

Optimization discarding geometry outside camera view frustum before rendering. For volumes, bounding box tests determine if entire volume regions are off-screen. Reduces unnecessary ray marching in fragment shaders.

Percentage of theoretical memory bandwidth being used during rendering. Volumetric workloads are often memory-bound due to extensive 3D texture sampling. High utilization indicates bandwidth-limited performance.

Calculation of volume density rate of change in X, Y, Z directions. Used for normal vector estimation in volumetric lighting. Requires additional texture samples, increasing shader complexity and memory bandwidth demands.

Technique identifying surfaces of constant value within volumetric data. Common in medical imaging for visualizing organs and bones. Can be combined with ray marching for hybrid rendering approaches.

Strategy rendering objects with varying complexity based on distance or importance. Volumetric LOD adjusts sampling rate, texture resolution, or iteration count. Maintains visual quality while improving performance for distant volumes.

Application domain rendering CT or MRI scan data as 3D volumes. Requires accurate density mapping and interactive frame rates for clinical diagnostics. Volume Shader BM performance directly correlates to medical visualization capabilities.

Hierarchical spatial data structure dividing 3D space into octants. Enables fast culling of empty regions and adaptive level-of-detail rendering. Essential for real-time volumetric applications with large datasets.

Performance penalty occurring when shaders require more registers than available. Excess data spills to slower local memory. Complex volume shaders often trigger register pressure, reducing parallelism and throughput.

Duration required to translate and optimize shader source code into GPU machine code. Compilation happens at runtime in WebGL. Complex volume shaders may take seconds to compile, impacting initial load experience.

Ratio of active warps to maximum possible warps on a GPU. Higher occupancy improves latency hiding but requires balancing register usage, shared memory, and thread block size. Critical metric for optimizing volume shaders.

NVIDIA GPU architecture component housing multiple CUDA cores, shared memory, and scheduling units. Executes thread blocks in parallel. SM count and configuration directly impact shader performance in Volume Shader BM.

Light transport phenomenon where photons penetrate and scatter within translucent materials like skin or wax. Volumetric approximation techniques simulate this effect for realistic rendering of organic materials.

Reusing previous frame data to reduce current frame computation. Camera motion vectors reproject prior samples to new pixel positions. Enables high-quality volumetric rendering at reduced sampling rates.

Fast on-chip memory storing recently accessed texture data. Critical for volumetric rendering where 3D textures are sampled millions of times per frame. Cache hit rates significantly affect performance.

Mapping from scalar volume data values to visual properties like color and opacity. Essential in scientific visualization for highlighting specific density ranges. Transfer function design significantly impacts both visual quality and performance.

OpenGL/WebGL mechanism for efficiently passing constant data to shaders. Volume rendering parameters like view matrices and lighting coefficients are stored in UBOs for fast shader access across multiple draw calls.

Process of interpolating per-vertex data across triangle surfaces to generate per-fragment values. In volume rendering, ray entry/exit points are interpolated from cube vertices to initialize ray marching in fragment shaders.