Memory Bandwidth
Memory
DEFINITION
The rate at which data can be transferred between GPU memory and processing cores, measured in GB/s.
OVERVIEW
Memory bandwidth represents the maximum rate at which data can be read from or written to the GPU's memory. It's a critical specification that can bottleneck performance in memory-intensive workloads.
TECHNICAL DETAILS
Memory bandwidth is calculated as: Memory Clock Speed × Memory Bus Width ÷ 8. For example, a GPU with a 384-bit bus width and 19.5 Gbps memory speed has 936 GB/s bandwidth. Higher bandwidth allows the GPU to feed data to its cores faster, reducing idle time and improving utilization in workloads that require frequent memory access.
COMMON USE CASES
- Large batch training where data transfer is frequent
- High-resolution image and video processing
- Graph neural networks with large adjacency matrices
- Molecular dynamics simulations
- Real-time rendering with complex scenes