RAM

RAM (Random Access Memory) is a type of computer memory that provides fast read and write access to data currently in use. Unlike storage drives, RAM is volatile, meaning its contents are lost when the system is powered off. It serves as the primary workspace where the CPU and GPU access data during computation.

In machine learning and AI, RAM is a critical resource. During training, datasets, model parameters, gradients, and optimizer states all reside in RAM or GPU VRAM (Video RAM). A model with billions of parameters requires substantial memory - for example, a 7-billion parameter model in 16-bit precision needs roughly 14 GB just for its weights, with additional memory required for activations and gradients during training.

The amount of available RAM directly constrains what models can be run locally. Techniques like quantization (reducing parameter precision from 16-bit to 4-bit), model sharding (splitting a model across devices), and offloading (moving layers between RAM and VRAM) help fit larger models into limited memory. Understanding RAM requirements is essential when selecting hardware for training or deploying AI models.