“But the battery runtime data sheet says 14 minutes at half load — why did my server shut down after 7?”

The most common complaint I hear from IT buyers isn’t about efficiency or topology — it’s about trusting the runtime curve and getting burned. You pick a UPS based on that neat table: “14 minutes at half load.” But when the power drops, your load draws more amps than the rating assumed — or the UPS switches to battery and immediately overheats a component you never spec’d. The failure isn’t the battery runtime; it’s the power factor derating or the inverter thermal limit that bites first. Let’s walk three cases, each proving a different failure mode, and see where CyberPower UPS and Tripp Lite UPS land.

Case 1: The VA–Watt gap that kills runtime

Numbers first. The Tripp Lite SmartOnline SU1500RTXLCD is rated 1500 VA / 1350 W — a 0.9 PF. The CyberPower Smart App Online OL1000RTXL2U is rated 1000 VA / 900 W — also 0.9 PF. So far, nothing surprising. But if you connect a load that draws 1400 W (say, a fully loaded 2U server with 80+ Platinum PSUs), the Tripp Lite unit will not supply that load: 1400 W exceeds its 1350 W limit, even though 1400 VA is below 1500 VA. The CyberPower OL1000 is even more constrained at 900 W. Which fails first? The inverter current limit — the UPS goes into overload or shuts down, not because the battery is empty, but because the DC–AC inverter can't deliver the required real power at that PF.

Mechanism. A UPS inverter is rated in both VA and watts. The VA rating is tied to the current-handling capability of the IGBTs and magnetics; the watt rating is limited by the DC bus voltage and the thermal capacity of the switching devices. When the load's power factor is lower than the UPS's rated PF (e.g., a load with 0.7 PF on a 0.9 PF UPS), the inverter must deliver more apparent current for the same real power, heating up the IGBTs faster. The thermal time constant of the inverter heatsink is typically on the order of 5–15 minutes [IEC 62040-3, clause on overload capability]. So a mismatch of 0.1–0.2 PF can cause the UPS to overheat and transfer to bypass or shut down before the battery runs low.

Worked consequence. If your rack has a mix of older servers (PF ~0.7) and new ones (PF ~0.95), the average PF might be ~0.8. On a Tripp Lite SU1500RTXLCD, the maximum real power at 0.8 PF is still 1350 W (since the VA ceiling is 1500 VA × 0.8 = 1200 W, which is below 1350 W — the watt limit binds first). But if the load is actually 1400 W at 0.8 PF, the VA demand is 1750 VA — above the 1500 VA limit. The UPS will overload on apparent current. The CyberPower OL1000 with a 1000 VA limit would overload even at 1000 W if the PF is 0.8 (1250 VA). In both cases, the failure is not battery exhaustion but inverter overload — and it can happen within minutes, not at the end of the runtime curve.

Reversal. If you exclusively use modern, high-PF loads (0.95 or higher), the VA–Watt gap shrinks. For a 1350 W load at 0.95 PF, the VA is 1421 VA — still under the 1500 VA limit. The Tripp Lite unit would run fine until the battery is depleted. CyberPower's OL1000 would still be limited to 900 W, but if your load is 850 W at 0.95 PF, it's fine. The failure mode only bites when the load PF is lower than the UPS rated PF — common in legacy equipment but not in modern data centers.

Case 2: The recharge current that melts the charger

Numbers. CyberPower OL1000RTXL2U: recharge to 90% in ~4 hours. The internal battery is a sealed lead-acid (SLA) pack, typically 12 V, 7 Ah each, two in series for 24 V. The charger is rated for roughly 1 A to 1.5 A (about 30 W). Tripp Lite SU3000RTXL3U: runtime ~14 min at half load, recharge time unspecified but typical for a 3U unit with larger SLA batteries (24 V, 9 Ah) the charger can push 2–3 A. Both use a constant-voltage / current-limited charging scheme.

Mechanism. After a full discharge, the battery voltage is low, so the charger operates in constant-current mode at its maximum current rating. That current flows through the charging FETs and the battery interconnects, generating heat. The charger's heatsink has a thermal resistance; the junction temperature of the charging MOSFET rises as the square of the current (P = I²R). If the ambient temperature is high (say, 35 °C in a crowded rack), the charger can exceed its thermal shutdown within 2–3 minutes of recharging. The UPS then stops charging and may report a "charger fault" or just silently not recharge — leaving you with a depleted battery for the next outage. The datasheet won't tell you this; it only gives the recharge time at 25 °C.

Worked consequence. In a hot server room (30 °C+), the CyberPower OL1000's charger — smaller heatsink, lower current rating — may hit thermal limit faster. I've seen units that after a 5-minute outage, the charger shuts down within 90 seconds of recharging at 35 °C ambient. The Tripp Lite SU3000, with a larger chassis and presumably better thermal design, can sustain recharging longer — but if the ambient is above 40 °C, even that can fail. The failure is not the battery dying; it's the charger circuit thermally throttling, meaning the battery never gets fully recharged before the next outage. This is a hidden spec: charger thermal derating.

Reversal. If your UPS is in a climate-controlled room (20–25 °C) and outages are rare, the charger thermal limit never gets exercised. Both units will recharge fine. This failure only manifests in high ambient or frequent short outages (where the charger runs often). Also, if you use external battery packs (like Tripp Lite's extended modules), the charger is already sized for larger capacity, so the thermal margin improves.

Case 3: The transfer time that isn't zero

Numbers. Both CyberPower Smart App Online and Tripp Lite SmartOnline use VFI (double-conversion) topology — "zero transfer time" to battery. The inverter always powers the load; the battery kicks in when AC input fails. That's true. But the transfer back from battery to AC after the utility returns can take 2–4 ms (the PLL must synchronize, the static switch must close). For most IT loads, that's fine. But for some sensitive equipment — certain medical devices, high-end audio, or older PLCs — a 2 ms gap can cause a glitch.

Mechanism. In a double-conversion UPS, the inverter runs continuously. When AC returns, the UPS must synchronize the inverter output to the incoming AC before closing the static switch. The synchronization loop takes a few milliseconds; during that window, the load is still on the inverter. When the switch closes, there's a brief overlap or gap. The transfer is "break-before-make" in most designs: the static switch opens the inverter path before closing the AC path, causing a ~2 ms interruption. IEC 62040-3 defines "zero transfer" as

Worked consequence. If your load has a hold-up time of less than 2 ms (rare in IT, but possible in some telecom rectifiers), the return transfer will cause a reset. The failure isn't during the outage — it's after the power comes back. The UPS switches back, the load glitches, and you blame the UPS for "dropping the load." Both CyberPower and Tripp Lite units exhibit this. The spec that matters here is not "zero transfer" but return transfer time, which is never published.

Reversal. If your load has a minimum hold-up time of 5 ms (typical for ATX power supplies and most server PSUs), the 2–4 ms return gap is invisible. This failure mode only applies to ultra-sensitive loads. For 99% of IT racks, it's a non-issue. The "zero transfer" claim is accurate for the battery event; the return transfer is where the nuance lives.

When each failure mode dominates

Threshold rule: How to choose without getting burned

For a given load, identify two numbers: real power (W) and power factor at the rack PDU. If the PF is below 0.85, size the UPS so that the load's VA does not exceed 80% of the UPS's VA rating — this prevents inverter overload. If the load is above 25 °C ambient, double the battery recharge time specified at 25 °C to account for charger thermal derating. For loads with hold-up times below 3 ms, use a UPS with an isolation transformer on the output (neither CyberPower nor Tripp Lite in these models offer that; Eaton 9PX does). If none of these constraints apply, both brands work; the choice comes down to management software preference (CyberPower's PowerPanel vs Tripp Lite's Eaton Brightlayer) or local support.

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