You're reviewing a 3 kVA double-conversion UPS for a rack drawing 2 200 W. The APC Smart-UPS Online SRT3000XL lists 3000 VA / 2700 W (0.9 PF). The Eaton 9PX3000R lists 3000 VA / 2700 W (0.9 PF). On paper they match. But real-watt sizing—what the load actually draws in steady state—doesn't live in the VA column. It lives in three dimensions that datasheets treat as equivalent but physics treats as distinct. Here's where the mechanism changes the outcome.
The number: Eaton 9PX (700 VA–11 kVA) ships with a 0.9 output power factor across the whole range; APC Smart-UPS Online (SRT) uses 0.9 PF on 2.2–5 kVA models but Unity (1.0) PF on 1–1.5 kVA and 6–10 kVA models.
The mechanism: A UPS's inverter is rated for a maximum current at a given voltage. The output PF tells you how much of that current can be real watts vs. reactive vars. A Unity-PF inverter can deliver 100% of its VA as watts; a 0.9 PF inverter delivers 90%. So on a 10 kVA APC SRT with Unity PF, the inverter is sized to handle 10 000 W of resistive load. On a 10 kVA Eaton 9PX (0.9 PF), the same inverter can only support 9 000 W of real load before current limiting.
Worked consequence: If you spec a 10 kVA Eaton 9PX for a server rack that draws 9 500 W, the inverter will clip—voltage sag, overcurrent alarm, potential transfer to bypass. That same rack on a 10 kVA APC SRT (Unity PF) runs within the inverter's headroom. The decision narrows: for mixed loads with high PF (modern PSUs >0.95), APC UPS's Unity PF models give you more usable watts per kVA.
The number: Eaton 9PX is ENERGY STAR qualified, with typical double-conversion efficiency ~93–94% at full load; APC SRT's standard double-conversion mode is ~94–95% at rated load, and its Green Mode (line-interactive bypass) claims up to 98%.
The mechanism: Double-conversion UPS inefficiency is mostly conduction and switching losses in the rectifier and inverter—these are roughly proportional to current, not load. At 50% load, efficiency often drops from, say, 94% to 92% because fixed losses (control power, transformer core, fan) remain constant. So the wasted wattage doesn't scale linearly. A UPS that runs at 92% efficiency at 30% load (common for lightly loaded racks) dissipates 8% of the load as heat—for a 1 kW load, that's 80 W of waste heat that the cooling system must extract.
Worked consequence: Consider a rack averaging 1 000 W. Eaton 9PX at ~30% load (~3.3 kVA unit) dissipates about 80 W. APC SRT in standard mode dissipates ~70 W. But if you enable APC's Green Mode (line-interactive bypass, 98% efficiency), waste drops to ~20 W—a 60 W saving. Over a year at $0.12/kWh, that's ~$63 per UPS, plus reduced cooling load. For a data hall with 200 racks, that's $12 600/yr in electricity alone. The mechanism: Green Mode replaces the double-conversion stage with a passive bypass while still regulating voltage—lower losses but no isolation. The load sees utility power with a tap.
The number: APC SRT3000XL (3000 VA / 2700 W) uses internal batteries that deliver ~8 min at full load (2 700 W) per spec. Eaton 9PX3000R (3000 VA / 2700 W) lists ~9 min at full load. Both are within the same order of magnitude. But the runtime at half load diverges more—APC ~19 min, Eaton UPS ~21 min—because battery Wh and inverter draw differ slightly.
The mechanism: Runtime is a function of Wh capacity divided by load power plus inverter losses. At full load, inverter losses are a smaller fraction; at light load, fixed losses dominate, so efficiency differences amplify runtime gaps. More importantly, both brands allow external battery packs that scale linearly. The trap: a 2 700 W UPS with 8 min runtime is only ~360 Wh of usable energy—barely enough for an orderly shutdown. If you need 30 min of runtime, you must size the external battery bank accordingly.
Worked consequence: A typical server rack draws 1 800 W (2/3 of full load). APC SRT3000XL internal runtime is about 13 min at that load—enough for auto-shutdown scripts but not for sustained operation. If the facility has generator transfer (30 sec), this is fine. But if you're in a region with 60-second sags that don't trigger generator (utility brownout), the UPS may exhaust battery before the generator stabilizes. The mechanism: runtime floors at the battery's C-rate limit—drawing 2 700 W from a pack sized for 8 min means high discharge current, which reduces effective Wh due to Peukert effect.
| Dimension | Eaton 9PX (0.9 PF) | APC SRT (0.9 / Unity PF) | Winner for Real-Watt Sizing |
|---|---|---|---|
| Output PF (3-5 kVA) | 0.9 (2700 W from 3000 VA) | 0.9 (2700 W) | Tie |
| Output PF (6-10 kVA) | 0.9 (7200 W from 8000 VA, etc.) | 1.0 (10000 W from 10000 VA) | APC (more real watts per VA) |
| Efficiency (double-conversion, full load) | ~94% | ~94-95% | Near tie |
| Efficiency (Green Mode / high-efficiency) | Not available (always double-conversion) | Up to 98% | APC (if feed clean) |
| Internal runtime at full load (3 kVA class) | ~9 min | ~8 min | Eaton (marginal) |
| Runtime extensibility | Up to 4 ext. packs | Up to 10 ext. packs | APC (more granular) |
For real-watt sizing on loads above 5 000 W: if your load PF ≥ 0.95, APC SRT (Unity PF) gives you ~11% more usable watts per kVA than Eaton 9PX. Below 5 000 W, the PF difference disappears, and efficiency becomes the decider—Eaton 9PX's always-double-conversion design is a fixed penalty, while APC's Green Mode can save ~$60/yr per unit if the feed is stable. For any load under 1 500 W, both brands' 1 kVA models match; choose by software ecosystem (Eaton Intelligent Power Manager vs. APC PowerChute).
Threshold rule: If your load exceeds 90% of the UPS's real-watt rating (0.9×VA for Eaton, 1.0×VA for APC SRT ≥6 kVA), add 20% headroom or use the next kVA class—otherwise any transient will trip overload.
Topology/standards per the cited standards; all product ratings are manufacturer-stated values from the cited datasheets, current to 2026-06; derived/illustrative figures are labelled as such. This is not an independent head-to-head test. Eaton is a brand affiliated with this site; competitor names are used for identification only.