Everything else on this site assumes the power stays on. The surge protector under the desk is insurance against a spike arriving down the mains, and it is worth having, but it has no answer at all to the opposite problem: the voltage going away. When that happens the monitors blink out, the tower drops mid-write, and the only question that matters is whether anything was holding the machine up for the thirty seconds it needed to close its files. A UPS is a battery and an inverter in a box, and on a desk its job is almost never to let you keep working through an outage. It is to turn a hard crash into an orderly shutdown.
The number on the front of the box is the one you should pay least attention to. Every unit here is sold on its VA rating — 1500VA, 850VA, 425VA — and VA is apparent power, not the watts your equipment actually draws. The watt figure is always lower, and the ratio is not a constant you can assume: CyberPower rates the CP1500PFCLCD at 1500VA / 1000W and the CP1500AVRLCD3 at the same 1500VA but only 900W. Two units with identical VA numbers, a hundred watts apart in what they can really carry. Find the watt rating before anything else and size against that, because the watt number is the one your load runs into first.
Then work out what your desk actually draws, which is less than you think. A desktop power supply is labeled for its peak capacity, not its consumption — a 750W unit in a machine doing spreadsheets and video calls is pulling a fraction of that at the wall. Monitors are modest too: a 27-inch panel is typically in the tens of watts rather than the hundreds. The honest way to size a UPS is to measure the real draw with a cheap plug-in meter, add the screens, and then leave headroom — both because a UPS at its rated limit has almost no runtime, and because batteries lose capacity with age. Sizing so your desk sits near half of the unit’s watt rating is the comfortable place to land. If you would rather not measure, bias upward: the cost of too much capacity is money, and the cost of too little is the thing you bought it to prevent.
The waveform is the specification that quietly decides whether this works at all. When a line-interactive UPS switches to battery it has to synthesize AC, and it does so either as a true sine wave or as a stepped approximation marketed as “simulated sine wave.” Most things do not care. Desktop power supplies with active power factor correction — which is most current ones, and effectively all of the 80 PLUS units in enthusiast builds — sometimes do: the failure mode manufacturers warn about is the power supply refusing the stepped waveform and dropping the load, or audibly buzzing, at the exact moment the UPS was supposed to save you. CyberPower sells an entire product line on this distinction and publishes the waveform for every model, which is why the sine wave units are ranked above the cheaper stepped one here even where the stepped one has a longer published runtime. If a desktop tower is plugged in, buy the sine wave.
Runtime specifications are real, but they are not about your desk.When CyberPower publishes “10 minutes / 2.5 minutes” for the CP1500PFCLCD, those figures are measured at half and full of the unit’s1000W rating — 500W and 1000W — not at whatever your setup draws. A typical desk sits below the half-load point, so it will run longer than the half-load figure, and that is the direction the error runs in your favor. What the numbers are genuinely good for is comparing units and resetting expectations: nothing in this class gives you an hour. Every one of these is a two-to-fifteen-minute device, which is ample for a scripted shutdown and nowhere near enough to finish a workday. If you need to keep working through a long outage, you are shopping for a portable power station, not a desk UPS.
Half the outlets on most of these have no battery behind them, and people get this wrong on the first install. A “12 outlet” UPS typically means six on battery plus surge and six on surge only, in two labeled banks. CyberPower spells out the split on every model and APC publishes it too. Plug the tower and the primary monitor into the battery bank, then put the printer, the desk lamp, the speakers and the phone chargers on the surge-only side, because anything you park on the battery eats runtime from everything else. One hard rule: a laser printer should never go on a UPS at all. Its fuser draws a large inrush that can overload the inverter outright, and printing is not something you need to finish during a blackout.
Finally, treat the battery as a wear item rather than a spec.Sealed lead acid cells in this class are generally good for a few years before capacity falls off, and the failure is quiet — the unit keeps showing a happy light and then delivers thirty seconds instead of ten minutes. Two things follow. Buy a unit whose battery is user-replaceable, so year four is a cartridge rather than a landfill decision. And check that the warranty covers the battery and not just the electronics: CyberPower states its three-year warranty includes the batteries, which is a materially better promise than a three-year warranty that excludes the one part guaranteed to wear out. Every unit below is ranked on published manufacturer specifications, and where a manufacturer does not publish a figure this page says so rather than filling the gap.
How to size and choose a UPS
Start with watts, not VA, and get the number from a meter rather than a guess. Plug a cheap inline power meter between the wall and your existing power stripand read it while you work normally, then once more with something demanding running. That second figure is the one to size against. Add the monitors if they are on a different strip. Then buy a unit whose watt rating is roughly double that number: a UPS running near its limit has a runtime measured in seconds, and the published full-load figures on this page — one minute for the CP1000PFCLCD, 2.5 for the CP1500PFCLCD — show exactly how fast that falls away.
Next, answer one question: is a desktop tower going on the battery? If yes, buy a true sine wave unit and stop considering the stepped ones, because a power supply with active power factor correction can refuse a simulated sine wave at the moment of transfer. That is the single reason the CP1500PFCLCD is ranked first and the cheaper, longer-running CP1500AVRLCD3 is ranked third. If no — if this is a laptop, a dock, monitors and network gear — the stepped units are good hardware at a lower price and the objection does not apply to you.
The in-between capacities are usually a trap worth naming. CyberPower also sells the CP1350PFCLCD at 1350VA / 880W, which sits between the 1000W and the 600W sine wave models. On paper it is a sensible middle rung; in practice it is priced close enough to the 1000W CP1500PFCLCD that you give up 120W of headroom and two minutes of full-load runtime to save very little. Check both prices on the day, but if the gap is small, take the bigger unit — headroom is the thing that ages well as the battery fades.
Plan where it physically goes before you buy, because a 1500VA UPS is a heavy box with a five-foot cord and it is competing for the same volume as a cable tray, a mounted towerand your knees. It belongs on the floor rather than in a tray — it is the heaviest thing in the setup and it needs air — and it wants to be near the wall outlet, not daisy-chained off a strip. That last point matters: a UPS plugs into the wall, and the strip plugs into the UPS, never the other way around.
Then wire it in the right order. Tower and primary monitor into the battery bank. Dock, router and modem into the battery bank if there is room, because an internet connection you can still use during a brief outage is most of the value. Everything else — second monitor if you are tight, speakers, lamps, chargers — onto the surge-only side. Laser printer onto the wall, never the UPS. Finally, install the manufacturer’s shutdown software and actually configure it: CyberPower ships PowerPanel and APC ships PowerChute, and an unconfigured UPS gives you a beeping box and a dead machine ten minutes later instead of a clean shutdown. Set it to hibernate or power down at a battery threshold you choose, then pull the plug once to watch it work. Testing it on a Tuesday afternoon is how you find out it was never armed.
