What is memory bandwidth?

Definition

Memory bandwidth measures how much data can move between memory and a processor in a given time. During artificial intelligence inference, model weights and intermediate values must be supplied to compute units quickly. If those units wait for data, adding more arithmetic capacity will not produce a proportional speed increase.

Large language models can be especially bandwidth-sensitive during token-by-token generation because substantial model data may need to be accessed for relatively little computation. Quantization, caching, smaller active parameter sets, and hardware with faster or unified memory can reduce this bottleneck. Capacity and bandwidth are different: a machine can hold a model in memory yet still run it slowly if transfer rates are limited.

Acronyms and aliases

RAM bandwidth variant

Frequently asked questions

Why does memory bandwidth matter for artificial intelligence inference?

Inference repeatedly moves model weights and intermediate values to processors. Low bandwidth can leave compute units idle while they wait for those values.

Is memory capacity the same as memory bandwidth?

No. Capacity determines how much data fits in memory, while bandwidth determines how quickly that data can be transferred to and from the processor.

Videos explaining memory bandwidth