You’re not missing anything. The trap is that published efficiency numbers are eligibility metrics, not operating averages. They tell you what a UPS can do under ideal conditions — not what it will do on your floor, with your load, on your utility feed. This is the core of the eligibility gate: if you pick a UPS solely by its headline efficiency, you’re likely to keep far less than you think. Let’s unpack the three gates that separate brochure watts from kept watts.
The Schneider Galaxy VS advertises double-conversion efficiency “up to 97% at every load level”, with an eConversion high-efficiency mode that reaches up to 99%. That sounds definitive. But the eligibility gate is: can your load tolerate eConversion? eConversion is a high-efficiency mode that passes power through the bypass path while the inverter stands ready — Class 1 no-break transfer. For linear, non-critical loads, it’s fine. But if your downstream equipment includes motor drives, large inrush transformers, or any device sensitive to the sub-cycle voltage transient that occurs during a transfer to double-conversion, eConversion becomes a gamble. The vast majority of IT loads (switch-mode PSUs with active PFC) are tolerant; the reversal is in mixed-load racks with legacy hardware or medical imaging, where a single transfer event can cause a reset. Eaton UPS’s 9PX runs double-conversion continuously, with a high-efficiency operating mode that stays within VFI topology. You don’t get the 99% headline, but you keep the 95–96% real-world average without worrying about which loads are in the bypass path. The rule: if your load profile is >80% modern IT, eConversion is safe; if you have any “unknown” legacy gear, the kept efficiency is the double-conversion rating, not the promotional mode.
Efficiency is a ratio — but the denominator changes if the UPS can’t deliver full watts at the load’s power factor. The Eaton 9PX is rated at 0.9 output power factor across its entire range (700 VA–11 kVA). That means a 10 kVA unit delivers 9,000 real watts, and efficiency is measured at that real load. The APC Smart-UPS Online (SRT) — Schneider UPS’s direct competitor in this class — has a split output power factor: 0.9 PF on 2.2–5 kVA models, but Unity (1.0 PF) on 1–1.5 kVA and 6–10 kVA models. That sounds better — Unity means watts = VA, right? But here’s the mechanism: a UPS’s real-world load nearly always has a lagging power factor (typical server PSU: 0.95–0.98 lagging). At Unity PF rating, the UPS is only guaranteed to deliver full VA at PF = 1.0. If your load is at 0.95 PF, a 10 kVA SRT may be limited to ~9,500 W, not 10,000 W — but the inverter still dissipates the same reactive circulating current losses. The worked consequence: efficiency at the load’s actual PF is often ~0.3–0.5 points lower than the spec-sheet curve measured at nominal PF. For Eaton 9PX, the 0.9 PF rating means it’s guaranteed to deliver 9,000 W at PF = 0.9, which matches the typical load profile better. The reversal: if you run Unity-PF loads (e.g., resistive heaters, some LED lighting), the Unity-rated SRT has a slight edge because the inverter sees lower circulating current — but that edge narrows to your load.
| Gate | Eaton 9PX (10 kVA) | Schneider APC SRT (10 kVA) | What you actually keep |
|---|---|---|---|
| Topology mode | Continuous double-conversion (VFI) | Double-conversion + Green Mode | Eaton: no mode-switching risk; Schneider: 98% only if load tolerates Green Mode |
| Output PF rating | 0.9 PF across range | Unity on 6–10 kVA | Eaton: 9,000 W guaranteed at typical PF; Schneider: 10,000 W only at PF=1.0 |
| Efficiency at typical load (0.95 PF, 50% load) | ~95.5% (illustrative, based on ENERGY STAR qualified) | ~95.2% (illustrative, double-conversion at 50% load) | Nearly identical — but Eaton’s number is measured at 0.9 PF, so it holds up better |
Here’s the non-obvious insight that most spec-sheet comparisons miss. A 1% efficiency difference on a 10 kVA UPS at 50% load (5 kW) is only 50 W of extra heat — negligible. But the real heat penalty doesn’t come from efficiency; it comes from reactive power circulation in the inverter. When a UPS is running at low load (per the standard at nominal voltage and with a resistive load — but at 20% load on the SRT, real-world efficiency can drop to ~92–93% (illustrative, based on typical double-converter curves). The worked consequence: if you oversize “for future growth” and run at 25% load for two years, your kept efficiency is closer to 92% — meaning you’re dumping 400 W of heat (on a 10 kVA) instead of 250 W. That 150 W extra heat has a compounded cooling cost: ~0.4 kW extra for the CRAC, 24/7/365. Over two years, that’s ~7,000 kWh of wasted electricity — about $700 at $0.10/kWh — that you never see on the efficiency brochure. The reversal: if you right-size aggressively and run at >50% load from day one, the difference between a 95% and 97% UPS is ~1°C in aisle temperature — thermodynamically real but operationally invisible. The rule: efficiency counts most at the load you actually run, not the load you spec’d for.
This is the failure mode that flips the decision for anyone with backup gen sets. A UPS that corrects input voltage over a wide range (like Tripp Lite SU3000RTXL3U, which regulates input from 65 V to 150 V back to 120 V ±2%) looks like a feature — but the voltage correction happens via boost/buck transformer taps or inverter recharging current, both of which increase input current distortion. On a generator, high current distortion (THDi > 20%) can cause the generator’s AVR to hunt, leading to frequency drift and eventual transfer to battery. The Galaxy VS’s input power factor correction reduces this risk — but the generic line-interactive or small double-conversion UPS without PFC (like some APC SMT series) can cause a genny to re-sync repeatedly, forcing the UPS onto battery even when the generator is running. The efficiency you keep during a 4-hour genny run is zero if the UPS is cycling on battery. The 9PX’s online topology inherently presents a clean, PFC-corrected input to the generator because the rectifier draws sinusoidal current. The gate: if your site has a generator, check that your UPS has active input PFC and a wide input frequency window (≥ ±5 Hz). Otherwise, the efficiency number on the brochure becomes irrelevant as soon as the genny fires up.
Final rule (executable threshold): Don’t compare efficiency numbers until you’ve normalized for three things: (1) the operating mode you’ll use >90% of the time (double-conversion, not eco-mode); (2) the actual power factor of your load (use a PF meter for one week); and (3) your generator’s voltage regulation bandwidth. If the gap between two UPS options after those three gates is 1%, buy the better gated efficiency, but only after confirming it holds at your load profile.
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.