Most UPS sizing advice starts with “how long do you need it to run”, and that is the wrong end of the problem. Work the arithmetic through and the runtime question answers itself, while the question that actually decides your purchase is the one nobody leads with.
Here is the whole thing in three steps.
Step 1: measure the load, in watts
Not the power supply rating, which is what the supply can deliver rather than what your machines take. Not a processor’s TDP. Put a plug meter on the whole lab and read what it draws, and take the reading under load rather than at rest, because that is the condition the UPS has to carry.
If you have not measured yet, the whole approach to standing draw is worked through in what a homelab really costs in electricity, and the same meter reading serves both purposes.
Step 2: convert watts to volt-amps, which is what the label means
This is the step that catches people, and it is the reason a UPS marked 1500 cannot necessarily carry 1500 watts.
A UPS is rated in volt-amps, which is apparent power: volts times amps, ignoring whether the two are in step. Your bill and your equipment care about watts, which is real power. The ratio between them is the power factor of the load, and a UPS has to be sized on the larger number.
Modern computer power supplies with active correction sit near 0.95, which is close enough that the difference is modest. Older equipment does not, and this is where the surprises live:
| Power factor of the load | VA needed for 300 W |
|---|---|
| 0.60 | 500.0 VA |
| 0.70 | 428.6 VA |
| 0.80 | 375.0 VA |
| 0.90 | 333.3 VA |
| 0.95 | 315.8 VA |
| 1.00 | 300.0 VA |
The same 300 watts of equipment needs anywhere from 300 to 500 VA depending on what is drawing it. If a UPS specification quotes only a VA figure and no watt figure, it is quoting the flattering number and assuming a power factor you may not have.
Step 3: add headroom, then read off the size
Do not run a UPS at its rating. Eighty percent is a sensible ceiling: it leaves room for the machine you add next year, keeps the unit out of its least efficient region, and means a startup surge does not trip it.
At a power factor of 0.95 and eighty percent headroom:
| Lab load | Load in VA | Buy at least |
|---|---|---|
| 50 W | 52.6 VA | 65.8 VA |
| 100 W | 105.3 VA | 131.6 VA |
| 150 W | 157.9 VA | 197.4 VA |
| 200 W | 210.5 VA | 263.2 VA |
| 300 W | 315.8 VA | 394.7 VA |
| 500 W | 526.3 VA | 657.9 VA |
Round up to whatever the market sells. The UPS sizing calculator does all three steps against your own power factor and headroom, and reports the watt figure alongside the VA so you can check both against a specification.
Now the runtime question, which turns out not to be one
Work out what you actually need. A UPS in a home lab is not there to ride out an outage. It is there to give your machines the seconds they need to shut down cleanly instead of corrupting a write in progress, which for most setups is two to five minutes with a margin.
At 150 watts:
| Shutdown window | Energy required |
|---|---|
| 2 minutes | 5.0 Wh |
| 5 minutes | 12.5 Wh |
| 10 minutes | 25.0 Wh |
| 15 minutes | 37.5 Wh |
Now compare that with what a UPS in the size class you just calculated actually carries. A single 12 V 9 Ah battery is 108 Wh nominal. After a realistic depth of discharge and inverter losses that is about 78 Wh usable, which at 150 W is 31 minutes:
| Lab load | Runtime on one 12 V 9 Ah battery |
|---|---|
| 40 W | 116.6 minutes |
| 60 W | 77.8 minutes |
| 100 W | 46.7 minutes |
| 150 W | 31.1 minutes |
| 200 W | 23.3 minutes |
| 300 W | 15.6 minutes |
You needed 12.5 Wh and the smallest thing that can carry your load has 78.
That is the whole point. Size for the VA and the runtime arrives as a side effect, several times over. People buy far larger units than they need because they are thinking in minutes, and minutes were never the binding constraint.
Two caveats on that runtime column, and they both point the same way. It is an upper bound: a battery delivers less than its rating when discharged quickly, an effect named after Peukert, and a battery a few years old delivers less again. And the vendor’s own runtime chart is the authoritative number, because they have measured their unit and this is arithmetic.
When you should buy more than the arithmetic says
- You want to ride through short outages rather than shut down. That is a different purchase and it is priced accordingly. Be honest about which one you are making.
- Something in the lab must not go down, like networking that a remote shutdown depends on. Put that on its own small unit rather than oversizing everything.
- You are about to expand. Headroom you planned for is cheaper than a second UPS.
- The batteries will age. They are consumable, typically on a three to five year cycle, and capacity falls the whole time. Buying with margin means the unit stays adequate rather than becoming marginal.
What the UPS costs to own
Two running costs, both real and both usually left out.
Electricity. The conversion is not free and the unit has its own standby draw, continuously, forever. On a lab of around 100 W the addition is tens of dollars a year, which is the running cost of the protection rather than an argument against it. Cost yours with the power cost calculator.
Batteries. They are a consumable on a multi-year cycle, and a UPS with dead batteries is a power strip that beeps. Whatever you buy, check the replacement batteries are available and priced sanely before you commit, because that decision recurs and the purchase does not.
The thing a UPS does not do
It does not protect your data. It protects one specific failure mode, which is losing power partway through a write, and that is worth having: for a parity array, it is the difference between a clean shutdown and a stripe left inconsistent with no record of which half is stale. The RAID explainer covers why that particular failure is nastier than it sounds.
But a UPS sees your deletions, your filesystem bugs and your ransomware exactly like everything else on the shelf. Three copies, two kinds of media, one of them somewhere else.
Finally, the shutdown has to actually be configured. A UPS with no USB or network connection to the machines it is protecting will hold them up for thirty minutes and then drop them just as hard as the outage would have. The cable and the daemon are the half of this that does the work.
FAQ
How many VA do I need for a 200 W server?
About 263 VA with a power factor of 0.95 and twenty percent headroom, so the next size up that anyone sells. If your equipment is older and its power factor is lower, that number climbs: at 0.7 the same 200 W wants closer to 350 VA. The sizing calculator takes the power factor as an input for that reason.
What is the difference between VA and watts on a UPS?
Volt-amps are apparent power and watts are real power, and the ratio between them is the load’s power factor. A UPS is rated in VA because that is what its components have to handle, and the watt figure is always equal to or smaller than the VA figure. A specification that quotes only VA is quoting the larger number, so find the watt rating before assuming a unit will carry your load.
How long should a home lab UPS run for?
Long enough to shut down cleanly, which is usually two to five minutes with a margin. At 150 W, five minutes is 12.5 Wh, and any UPS large enough to carry 150 W has several times that in its battery already. If you want to ride through outages rather than shut down, that is a deliberately different and more expensive purchase.
Should I put my NAS and my server on the same UPS?
Usually yes, since sizing on the combined load is exactly what the arithmetic above does, and one unit is cheaper and simpler than two. The exception is anything whose job is to keep working while the rest shuts down, such as networking a remote shutdown depends on, which is better on its own small unit than as a reason to oversize the big one.
Do I need a pure sine wave UPS?
For a computer power supply with active power factor correction, it is the safer choice. Simulated sine wave output and active correction interact poorly, and the failure can look like a machine that refuses to switch to battery. Check what your power supplies expect, since this is a compatibility question rather than a quality one.
How often do UPS batteries need replacing?
Typically every three to five years, and it is worth treating as scheduled maintenance rather than a fault. Capacity falls the whole time, so a unit that gave you thirty minutes when new gives you noticeably less by the end. Test it: a UPS whose batteries died quietly is indistinguishable from a working one until the moment it matters.
Sources
- The relationship between watts, volt-amps and power factor is definitional. Every VA, watt and runtime figure on this page is computed by the same code that runs the UPS sizing calculator.
- Loads, power factors, battery capacities, depth of discharge and inverter efficiency here are illustrative inputs, not measurements of any product. Nothing on this page was measured on hardware. Take the load from a plug meter and the battery figures from the unit’s own specification.
- Runtime from battery capacity is an upper bound. Real delivery falls as discharge current rises, an effect described by Peukert’s law, and falls again as a battery ages. The manufacturer’s runtime chart is the number to trust over any arithmetic, including this page’s.