SYSTEM MEMORY

How much RAM does a home server need?

Enough for the operating system, every service that can run at the same time, filesystem caching and some headroom. There is no honest universal number because a Home Assistant hub and a virtualisation server are different machines.

Start from the applications

Read the requirements and observed behaviour of the software you plan to run together. OCR, file indexing and other jobs can create short bursts even when the server is usually quiet.

Containers do not remove memory requirements

Docker makes deployment convenient but the applications still use real memory. Ten lightweight containers may fit comfortably while two memory-heavy services do not.

Virtual machines add overhead

If each service lives in a separate VM, you reserve memory for multiple operating systems as well as the applications.

File servers benefit from spare RAM

Operating systems use unused memory for filesystem caching. Running constantly at the limit gives the server less room to absorb bursts.

Local AI is different

For local LLMs, system RAM and GPU VRAM both matter. Do not treat ordinary RAM as equivalent to dedicated VRAM. Use the local AI memory calculator for that workload.

Measure before upgrading

Watch real memory use during backups, OCR, indexing and other busy periods. Upgrade because the workload needs it, not because a generic checklist says so.