PSU Efficiency Curve Drop-Off at Low Idle Loads Under 20W

This article explores the sharp decline in Power Supply Unit (PSU) efficiency when benchmarking computer hardware at very low idle loads below 20 watts. It explains the underlying electrical reasons for this steep drop-off, how traditional efficiency certifications overlook ultra-low load scenarios, the impact of high-capacity power supplies on idle power measurements, and how modern standards like ATX 3.0 address sub-20W efficiency for accurate hardware testing.

The Mechanics of the Sub-20W Efficiency Drop

Power supply efficiency is the ratio of DC power delivered to the computer components compared to the AC power drawn from the wall outlet. Most modern PSUs peak in efficiency between 40% and 60% of their maximum rated load, often reaching 90% to 94% efficiency. However, when a modern system enters a deep idle state—often drawing less than 20W—the PSU's efficiency curve plummets dramatically, frequently dropping below 60% or even 50%.

This severe drop-off happens because a PSU requires a fixed amount of baseline power just to operate its internal components. This is known as fixed loss or parasitic power consumption. Regardless of whether the PC demands 500W or 15W, the PSU must power its internal control ICs, supervisory circuits, voltage regulators, and active protection mechanisms.

When the system load is under 20W, the PSU's fixed internal power consumption becomes a massive percentage of the total power draw. For example, if a PSU requires 8W simply to stay active and the computer hardware requests 12W at idle, the total AC draw from the wall is 20W. This results in a dismal 60% efficiency. As the hardware load drops further toward 5W, the efficiency curve falls almost vertically.

Switching Losses and Component Sizing

Inside a PSU, transformers, inductors, and MOSFETs are optimized to handle hundreds of watts. At loads under 20W, these large components operate far outside their optimal electrical range.

The primary cause of inefficiency in this state is switching loss. Every time a transistor switches on and off to convert power, a tiny amount of energy is lost as heat. At high loads, this loss is negligible compared to the massive power flowing through the system. At under 20W, switching losses dominate. While modern PSUs use "burst mode" or "skip mode" to reduce switching frequency at low loads, the efficiency still degrades rapidly compared to normal operating loads.

The Over-Provisioning Problem in Benchmarking

In hardware benchmarking, reviewers frequently use high-wattage power supplies (such as 1000W to 1600W units) to ensure the test bench can handle any flagship CPU or GPU without swapping components. While convenient, this ruins low-load idle benchmarks.

Efficiency scales with percentage of maximum capacity, not absolute wattage. A 20W idle load represents a 4% load on a 500W PSU, which might manage 75% efficiency at that level. However, that same 20W load represents a mere 1.25% load on a 1600W PSU. At 1.25% load, the 1600W unit's larger internal components and higher baseline power draw will cause its efficiency to crash, artificially inflating the "at-the-wall" idle power numbers recorded during a benchmark.

For years, the industry relied on the 80 Plus certification standard, which historically ignored low-load states. Standard 80 Plus ratings (Bronze, Silver, Gold, Platinum) only measure efficiency at 20%, 50%, and 100% of the PSU's rated capacity.

Even the highest tier, 80 Plus Titanium, only tests down to a 10% load. For an 850W Titanium PSU, a 10% load is 85W—more than four times higher than a modern PC's sub-20W idle state. Because manufacturers were not graded on sub-20W performance, they did not optimize their circuits for it, leading to massive efficiency drop-offs in real-world idle scenarios.

How This Skews Hardware Benchmarking

When benchmarking CPUs, motherboards, or GPUs, measuring power at the AC wall outlet is common practice. If the test system idles at 18W DC, an inefficient PSU operating at 55% efficiency at that load will pull 32.7W from the wall.

If a reviewer tests a more power-efficient CPU that drops the system DC load to 14W, the PSU's efficiency might drop further to 50%, resulting in a 28W draw at the wall. The PSU's collapsing efficiency curve masks the true power savings of the hardware. To benchmark sub-20W loads accurately, reviewers must measure DC power directly at the cables (using tools like NVIDIA PCAT or inline clamp meters) rather than relying on AC wall meters, which are heavily distorted by the PSU's low-load drop-off.

Modern Standards: ATX 3.0 and Cybenetics

The industry has recently begun addressing the sub-20W efficiency drop-off. Intel’s ATX 3.0 specification introduced strict new requirements for low-load efficiency, specifically targeting the 2% load mark and "Alternative Low Power Modes" (ALPM). Under ATX 3.0, PSUs are required to maintain at least 60% efficiency at a 2% load (which equals 20W on a 1000W PSU), with a recommended target of 70%.

Additionally, the Cybenetics certification program—a modern alternative to 80 Plus—tests power supplies across a much wider range of loads, including ultra-low loads and vampire (standby) power on the 5VSB rail. Hardware benchmarkers looking for accurate wall-power readings at idle now rely on ATX 3.0 compliant units and Cybenetics efficiency graphs to minimize the impact of the sub-20W efficiency drop.