“It says 97% efficiency in the datasheet—so why does my AC unit still run when the UPS is online?”

I got that exact question from a facility engineer in Phoenix last month. He had installed a Schneider UPS Galaxy VS at a small colo, believing the 97% efficiency claim meant his cooling load would drop proportionally. The AC compressor stayed on. The electric bill barely budged. What the datasheet didn't say—and what every buyer needs to know before they sign—is that the eligibility of that efficiency number depends on topology, load profile, and whether you can actually stay out of battery mode. APC UPS Smart-UPS Online claims 98% in Green Mode. Both are real numbers, but only inside a very specific operating gate. When you fall out of that gate—and you will, unless you design for it—the comparison flips.

1. Efficiency vs. Mode Lock: The gate you can't see

The Schneider Galaxy VS is rated at up to 97% efficiency in double-conversion (VFI) mode and up to 99% in eConversion mode, a high-efficiency bypass scheme with Class 1 performance and no-break transfer. The APC Smart-UPS Online (SRT) offers a Green Mode that claims up to 98% efficiency. Both are true inside the gate. But here's the mechanism: eConversion and Green Mode are both forms of bypass—the UPS runs the load on conditioned bypass unless it detects a disturbance, then transfers to double-conversion in under 2ms. The efficiency gain comes from reduced conversion losses (~2–3.8% for the Galaxy VS). The catch? If your site has frequent voltage sags, generator transfers, or poor utility quality, the UPS spends more time in double-conversion, where efficiency drops to 96–97% (Galaxy VS) or ~95–96% (APC SRT in double-conversion).

The worked consequence is a decision: if your facility has stable utility (less than 2–3 sags per month) and you can coordinate generator-to-UPS transfer times, you capture the high-efficiency mode. If not, you're paying for a premium you can't use. The reversal occurs for edge sites, warehouses, or any location with poor power quality—there, a simpler double-conversion unit with no high-efficiency mode (like a base-model Eaton 9PX at ~94–95%) might actually deliver lower average loss because it never tries to switch modes and fail.

2. Load factor & the misalignment of VA vs. real watts

The APC Smart-UPS Online SRT at 2.2–5 kVA ships with a 0.9 output power factor; the 6–10 kVA units achieve unity power factor. The Galaxy VS is rated in kW for three-phase (10–150 kW) with an implied power factor of 0.9–1.0 depending on configuration. The datasheet hides that efficiency is measured at full rated load with a resistive (PF=1) load. In real data centers, loads are mixed inductive/capacitive with PF 0.85–0.95. When you run a UPS at 70% load with a PF of 0.9, the actual current draw is higher than at unity PF, and conversion losses scale with current, not VA. The Galaxy VS's 97% efficiency is measured at 100% linear load; at 50% load with 0.85 PF, expect about 95.5–96% (derived, ~1.5% penalty from I²R losses in magnetics). The APC SRT at 70% load with 0.9 PF will see a similar derating.

The worked outcome: sizing by VA and ignoring real-watt load factor can push you into a region where both units operate below their peak efficiency knot. For the Galaxy VS, the knee is around 40–60% load; below that, efficiency drops sharply. If you oversize to 30% load, you lose 3–5 percentage points. The APC SRT has a wider peak plateau (30–70%) but a steeper drop below 20%. The reversal is for a lightly loaded expansion rack: a smaller UPS that runs at 50% load may be more efficient than a large unit at 20% load, even if the large unit's peak efficiency is higher.

3. Transfer time & the hidden penalty of "zero" transfer

Both the Galaxy VS (double-conversion mode) and APC SRT (double-conversion mode) claim zero transfer time. That's true in VFI mode—the inverter always powers the load. But the moment you use eConversion or Green Mode, you introduce a transfer risk: when the bypass source is lost, the UPS must transfer to inverter in under 2ms. The IEC 62040-3 standard classifies this as VFI-SS (no-break) only if the transfer completes within the hold-up time of the load. Most modern switch-mode power supplies hold up for 10–20ms. A 2ms transfer is safe. The hidden penalty is that in high-efficiency mode, the UPS is not actively conditioning the output voltage—it's essentially on bypass. If the input voltage sags below 90% nominal (e.g., 108V on a 120V system), the APC SRT will transfer to double-conversion, and that transfer, while fast, resets the efficiency clock: you're now in double-conversion, losing 2–3% efficiency. The Galaxy VS eConversion handles a wider window (typically ±10% of nominal) before transferring, but the same penalty applies.

The worked consequence: if you have an automatic transfer switch (ATS) or generator transfer that takes 5–10 seconds, the UPS will transfer to battery during the transfer, then back to eConversion/Green Mode after utility restores. Each such event costs you runtime and recharge energy. Over a year, 10 such events can add 1–2% to average energy loss. The reversal is for a site with dual-feed or no generator: the high-efficiency mode will rarely drop out, and the Galaxy VS's eConversion (99% efficiency) will save ~$200–400/year vs. a double-conversion-only unit at 97%, assuming 10 kW load and $0.12/kWh.

4. Cooling interaction: the thermal ripple effect

A persistent misunderstanding is that UPS efficiency directly reduces cooling load. It does—but only by the conversion loss dissipated as heat. At 97% efficiency and 10 kW load, the Galaxy VS dissipates about 300 W (3% × 10 kW). At 95% efficiency, that's 500 W. The difference (200 W) is about the heat from a single desktop PC—not trivial, but not enough to turn off a CRAC unit. The real cooling interaction is about airflow and placement. The Galaxy VS uses internal fans that draw ~100–150 W and reject heat to the room; the APC SRT uses similar fans. If the UPS is in a sealed rack without hot-aisle containment, that 300–500 W can raise rack inlet temperature by 2–3°C, potentially triggering fan speed increase in server PSUs, which draws an additional 50–100 W per rack. The datasheet never shows this thermal ripple.

The worked decision: for a high-density rack (10+ kW), placing the UPS in a separate row with dedicated cooling can save 1–2% of total facility power. The reversal is for low-density racks (

Non-obvious insight: the hidden cost of battery recharge after a sag

Every time the UPS transfers to battery—even for a 2-second sag—it must recharge the batteries afterward. For a 5-minute battery discharge at 10 kW, recharge can draw 1.5–2 kW for 1–2 hours (depending on battery chemistry and charger capacity). That recharge power is 100% loss (it goes to chemical storage, not to the load). Over a year with 10 such events, that adds about 15–20 kWh of extra energy consumption—roughly $2–3 at typical commercial rates. Not a huge number, but it's never in the datasheet. The Galaxy VS and APC SRT both use similar charger topologies; the difference is marginal.

Failure mode: when the gate fails completely

Consider a site with a backup generator that has a 10-second start delay and a 5-second transfer switch. During a utility outage, the Galaxy VS (or APC SRT) will transfer to battery, then when generator power arrives, it will synchronize and transfer back to utility/bypass. If the generator voltage or frequency is unstable (e.g., ±5% voltage, ±2 Hz frequency), the UPS may reject the generator and stay on battery until exhausted. This is a failure mode: the high-efficiency mode is never re-entered because the generator is outside the sync window. The result is a UPS running in double-conversion (or even on battery) for the entire generator run, consuming 3–5% more power than expected, plus battery drain. The Galaxy VS has a wider generator acceptance window (±10% voltage, ±3 Hz) compared to some APC SRT models (±8% voltage, ±2 Hz). This can be the deciding factor for sites with older generators.

Summary gate eligibility table

The bottom line: the Galaxy VS wins on peak efficiency and wider generator tolerance—if you can stay in eConversion. The APC SRT offers a more compact form factor and unity PF at higher ratings, but its Green Mode is slightly less efficient and has a tighter operating gate. For both, the datasheet hides the mode transition penalty and the thermal ripple. Use the eligibility gate framework above to decide which one matches your actual site conditions—not the headline number.

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. Schneider Electric is a brand affiliated with this site; competitor names are used for identification only.