Eaton vs APC UPS: Which One Sizes by Real Watts – and Why It Matters

Wednesday 17th of June 2026 · Jane Smith · Blog
by Mike Holt · Sizing methodology · 4 min read

You are looking at a 1.8 kW IT load – twelve PoE switches drawing 150 W each. The APC Smart-UPS Online SRT shows 2 kVA / 1.8 kW on its spec sheet. The Eaton 9PX shows 2 kVA / 1.8 kW. Both fit on paper. But what happens when the load isn't purely resistive? One of these will run hotter, lose runtime, or force an early upgrade. Here is the real-watts sizing breakdown.

1. Output Power Factor – the Lever That Changes Everything

The Eaton 9PX series operates at a fixed 0.9 output power factor across its range (700 VA – 11 kVA). The APC Smart-UPS Online SRT also claims 0.9 PF on the 2.2–5 kVA models, but switches to Unity (1.0) PF on the 1–1.5 kVA and 6–10 kVA frames. On paper, a 2 kVA unit from either delivers 1.8 kW – identical. But in physics, the power factor tells you how much of the VA the inverter can actually turn into real watts delivered to the load. A 0.9 PF means the inverter is rated for 90% of the nameplate VA as continuous watts; Unity means 100%.

Worked consequence: Assume a 1.6 kW load with a power factor of 0.85 (typical for modern server PSUs under 80 PLUS Silver). On the APC SRT 2.2 kVA (0.9 PF), the unit can handle up to ~1.98 kW real – you are well within. On the Eaton 9PX 2 kVA (0.9 PF), same story. But consider the APC SRT 1.5 kVA frame: at Unity PF, the real-watts limit is 1.5 kW. That 1.6 kW load exceeds real-watts capacity, even though the VA rating suggests "close enough." You would pick the next size up (2.2 kVA) and suddenly the APC UPS has a lower PF – the same real-watts as the Eaton 2 kVA. So the Eaton UPS stays in a smaller VA frame for the same real load, saving rack space.

Real-world rule: For loads with PF ≤ 0.9 (most of IT gear), the Eaton's fixed 0.9 PF means the VA-to-watts ratio is consistent. APC's split-PF strategy creates a trap at the 1.5 kVA boundary – you need to verify the real-watts limit, not just VA. If your load PF is above 0.95 (e.g., Unity PF server), APC's Unity models win on paper, but few edge devices hit that.

When this flips: If your load is nearly resistive (PF > 0.95, e.g., resistive heaters, some LED lighting drivers), the APC Unity frame gives you 100% of VA as real watts – the Eaton's 0.9 PF caps out at 90%. For a 1.4 kW resistive load, the APC 1.5 kVA unit (1.5 kW real) fits; the Eaton 2 kVA (1.8 kW real) also fits but costs more per VA.

2. Efficiency – How Much of That Real Power Stays as Heat

Eaton 9PX is ENERGY STAR qualified, with typical double-conversion efficiency in the 94–96% range at typical loads (illustrative, about 95% at 50–100% load). APC Smart-UPS Online SRT offers "Green Mode" up to 98% efficiency, but in standard double-conversion its efficiency is ~94–95%. The magnitude here: at 1.8 kW load, a 95% efficient UPS burns 90 W of heat. A 94% efficient unit burns ~108 W. The difference (18 W) is roughly the heat from a small GPU at idle – negligible for a data centre, but for a network closet with no dedicated cooling, it means a 5–7°F temperature rise difference over a day.

Worked consequence: In a sealed 42U rack with 2 kW of load, a 1% efficiency delta changes internal temperature by about 1.5°F (assuming 200 CFM airflow; illustrative). Over a year, that is ~$12 in extra cooling cost (at $0.12/kWh). Not massive, but when you scale to 10 kW in a row of racks, the 1% delta becomes ~$120/year per rack. The Eaton's consistent high efficiency across load range (vs APC Green Mode which only activates under strict conditions and has a transfer event) means you don't need to manage the mode.

When this flips: If your load is highly variable (cycles between 20% and 100% every few minutes) and you can tolerate the brief battery transfer of Green Mode (typically 2–4 ms), APC Green Mode can yield 1–2% efficiency advantage over Eaton's standard double-conversion. But for stable IT loads, the Eaton's fixed high efficiency avoids the complexity of mode switching.

3. Runtime Curves – the Real-Watts Sizing Trap

UPS runtime curves are non-linear – double the load, cut runtime by ~60–70%. The Eaton 9PX 2 kVA unit (1.8 kW) typically delivers ~10 minutes at 1.8 kW and ~22 minutes at 900 W (half load). The APC SRT 2.2 kVA unit (1.98 kW real) delivers ~8 minutes at 1.8 kW and ~18 minutes at 900 W. Here is the magnitude trap: because the APC's real-watts capacity is slightly larger (1.98 vs 1.8 kW), you might think you can run a 1.8 kW load longer. But the APC's internal battery pack is sized for the VA rating, not the real-watts difference – the UPS uses the same battery string across the range. At 1.8 kW, the APC runs 2 minutes shorter than the Eaton. That 20% runtime gap (2 min out of 10) can mean the difference between a graceful shutdown and a crash during a 9-minute brownout.

Worked consequence: For a 1.7 kW load that needs 12 minutes of runtime for graceful shutdown of a SAN, the Eaton 2 kVA fits (10 minutes at full load, ~13 min at 1.7 kW). The APC 2.2 kVA at 1.7 kW gives ~9 minutes – too short. You would need the APC 3 kVA frame, which is physically larger (3U vs 2U). The Eaton saves rack space because its real-watts efficiency and battery sizing align better at the boundary.

When this flips: If your load is very light (

4. Management Software – Sizing After Installation

APC bundles PowerChute Business Edition and Network Shutdown for graceful integration. Eaton uses Eaton Intelligent Power Manager (IPM) and a network card slot. The magnitude of sizing impact: Both can send shutdown signals based on runtime. But the Eaton 9PX includes load segment control (two switchable outlets) as standard on most models, allowing you to shed non-critical loads during a power event – effectively extending runtime for critical kit by 30–50% without buying more battery. APC offers this only on the higher-end SRT models with optional load-bank cards. Worked consequence: In a 2 kW load, if you can shed 500 W of non-critical PoE cameras after 2 minutes of outage, the critical 1.5 kW load gets ~14 minutes of runtime instead of ~10 minutes. That is a 40% increase without capital spend. Eaton's standard implementation makes this available on the 9PX; APC requires an accessory card.

Non-obvious insight: The best sizing move is not buying more VA – it's buying a UPS that lets you redefine the load in real time. Eaton's load-segment granularity on mid-range units can make a 2 kVA unit behave like a 3 kVA unit for the critical path during an outage. APC's approach forces you to size for the worst-case peak, costing more upfront.

Failure mode: If you never configure load shedding (most sites don't), the advantage is zero. Both units will alarm and then shut down at the same voltage threshold. The Eaton's extra outlets are wasted if not programmed.

Quick Comparison at the Sizing Boundary (1.8 kW IT Load)

ParameterEaton 9PX 2 kVA (1.8 kW)APC SRT 2.2 kVA (1.98 kW)
Real-watts limit @ 0.9 PF1.8 kW (fixed)1.98 kW
Runtime at 1.8 kW~10 min~8 min
Rack height2U2U
Load segment controlStandard (2 groups)Optional (accessory card)
Efficiency (double-conversion, typical)~95%~94%

The rule: If your load PF is ≤ 0.9 and you need real-watts density and runtime consistency, the Eaton 9PX lets you size one VA class lower than APC for the same real load, saving space and cost. If your load is very efficient (PF > 0.95) and you want the smallest VA frame, APC's Unity models give a marginal advantage at the 1.5 kVA boundary – but that advantage disappears as soon as you need >1.5 kW real. Choose by real watts at your load's PF, not by VA nameplate.


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.

Leave a Reply