A home server runs continuously and spends most of its time idle. That profile makes it a very different purchase from a desktop, and it is why a repurposed old machine is usually the wrong choice — its idle power draw over a few years of continuous operation costs more than a modern compact machine that does the same work.
Sizing one correctly comes down to a handful of specifications, and processor performance is not the important one.
What changed in 2026
- Idle efficiency improved. Newer compact platforms drew meaningfully less power at idle, which is where these machines spend their time.
- Memory capacity ceilings rose. Support for larger memory configurations in compact form factors removed a previous constraint.
- Multiple storage slots became common. Two or more high-speed storage slots appeared in more compact machines, enabling separate system and data drives.
- Faster networking spread. Higher-speed network interfaces became more common, though slower ports persisted on budget models.
What to prioritize
| Specification |
Priority |
Reason |
| Memory capacity and upgradeability |
Highest |
The constraint for running several services |
| Idle power draw |
High |
Continuous operation; peak is rarely reached |
| Storage slot count |
High |
System and data separation |
| Network interface speed |
Moderate to high |
Depends on what you serve |
| Processor cores |
Moderate |
Most home workloads are light |
| Media decode and encode support |
Situational |
Essential for transcoding, irrelevant otherwise |
| Physical size and noise |
Situational |
Matters if it lives in a living space |
Memory is where people under-buy. Each service you run consumes some, and the total climbs faster than expected as you add things. Buying generously initially, and confirming the machine supports upgrades, avoids the situation where adding one more service means replacing the host.
Processor performance rarely constrains a home server. Most services are idle most of the time, and the occasional burst is handled fine by any modern processor. The exception is video transcoding, which needs hardware media support rather than raw processor speed.
Power and the old-desktop question
Idle power is the specification that determines running cost. A machine drawing a modest amount continuously costs a predictable amount per year, and the difference between an efficient compact machine and an old desktop can be substantial over the lifetime of the deployment.
Run the arithmetic for your electricity rate before repurposing old hardware. The old machine is free to acquire and not free to run, and over three or four years of continuous operation the difference frequently exceeds the purchase price of a new efficient machine.
That calculation also favours consolidating several always-on devices onto one host, which is one of the arguments for virtualization — see Proxmox homelab guide.
Storage and networking
Two storage slots is the practical minimum: one for the system and one for data. A single-slot machine forces everything onto one drive, which means no separation between the operating system and the data you care about.
For bulk storage, external enclosures or a separate storage device usually make more sense than trying to fit many drives into a compact chassis — the considerations are in NAS buying guide.
Check the network interface speed. Some compact machines still ship with slower ports, which caps throughput regardless of what the storage can deliver. If you serve large files over the network, this becomes the bottleneck.
Verify media hardware support if you transcode video. Some compact processors include capable media engines and some do not, and the difference between hardware and software transcoding is enormous for that workload.
Common mistakes
- Under-buying memory. The usual constraint, and sometimes not upgradeable.
- Repurposing an old desktop. Power cost exceeds a new machine over a few years.
- Single storage slot. No separation between system and data.
- Ignoring network speed. Caps throughput regardless of storage.
- Buying processor performance. Rarely the constraint for home workloads.
- Not checking media support for transcoding. A specific requirement, easily missed.
FAQ
How much memory do I need?
More than you initially plan. Count the services you intend to run, estimate generously, and confirm upgradeability.
Do I need error-correcting memory?
For data integrity with a checksumming filesystem, it helps — the discussion is in ZFS explained. Many compact machines do not support it.
Can one machine handle everything?
For typical home workloads, comfortably. Separating storage onto a dedicated device is common as things grow.
What about noise?
Passively cooled models exist and are silent. Actively cooled ones vary; check reviews if it lives in a living space.
Where to go next
For virtualization, read Proxmox homelab guide and Docker vs VMs. For storage, NAS buying guide, and for thin endpoints, thin clients explained.