GPU Architecture
Architecture
DEFINITION
The underlying design and organization of a GPU, defining its capabilities and performance characteristics.
OVERVIEW
GPU architecture refers to the fundamental design of the graphics processing unit, including the arrangement of cores, memory hierarchy, instruction set, and specialized units. Each architecture generation brings improvements in performance, efficiency, and new features.
TECHNICAL DETAILS
NVIDIA's recent architectures include Ampere (A100, A10), Hopper (H100), and Ada Lovelace (RTX 40 series). Each architecture introduces changes like new core designs, improved Tensor Cores, larger caches, faster memory interfaces, and new instructions. Architecture determines the supported features, precision formats, and maximum theoretical performance.
COMMON USE CASES
- Selecting GPUs for specific workload requirements
- Understanding performance differences between generations
- Planning hardware upgrades and migrations
- Optimizing code for specific architectural features
- Evaluating price-performance tradeoffs