GRAPHICS PROCESSING UNIT

NEWS: GPUà essential for artificial intelligence, machine learning, and high-performance computing due to its unique architecture designed for massive parallel processing

About:

• A GPU is an extremely powerful number-cruncher built to perform many simple calculations simultaneously.
• Designed for fewer complicated tasks with fast task-switching
• A 1920×1080 screen has 2.07 million pixels per frame; at 60 frames/second
• Updates over 120 million pixels per second—each pixel’s colour depending on lighting, textures, shadows, and material properties.

Working

GPU processes through a four-step sequence called the rendering pipeline:

• Vertex Processing: Uses maths (matrices) to position 3D objects on screen.
• Rasterisation: Converts 3D triangles into pixels.
• Pixel Shading: Calculates final color of each pixel.
• Writing to Frame Buffer: Final pixel colors are stored in memory.

Shaders & VRAM

• Shaders = small programs that run in parallel on GPU.
  • Handle calculations for vertices and pixels quickly.
• VRAM = high-speed video memory for storing textures, models, and images.
  • Cache memory helps reduce repeated data fetching.
• GPU LocationàCan be a separate graphics card with its own VRAM or integrated with CPU (common in laptops and smartphones).