AIO Benchmarking: Pump Speed vs. Fan Speed Optimization

Benchmarking All-In-One (AIO) liquid coolers requires a precise balance between pump speeds and radiator fan speeds to accurately measure thermal performance and acoustic efficiency. While the pump circulates coolant to transfer heat away from the CPU, the fans exhaust that heat from the radiator into the surrounding air. This article explores the mechanics of pump and fan speed optimization during hardware benchmarks, detailing how testers isolate these variables, apply noise-normalization, and identify the point of diminishing returns for maximum cooling efficiency.

The Mechanics of AIO Heat Transfer

To understand optimization, you must first understand the two-stage heat transfer process of an AIO cooler.

  1. Heat Absorption (The Pump): The CPU generates heat, which passes through the cold plate into the liquid coolant. The pump pushes this heated liquid through tubes to the radiator.
  2. Heat Dissipation (The Fans): The radiator disperses the liquid's heat into a dense array of metal fins. The fans blow air through these fins to exhaust the heat away from the cooler.

In benchmarking, if the pump runs too slowly, the liquid stays on the CPU too long, causing temperature spikes regardless of how fast the fans are spinning. Conversely, if the fans run too slowly, the liquid inside the radiator never cools down, eventually saturating the loop with heat no matter how fast the pump is working.

Isolating Pump Speed

To benchmark the impact of the pump, hardware testers lock the radiator fans at a fixed RPM (usually 100% or a fixed decibel level) and run a constant thermal load on the CPU (such as Blender or Prime95) until the liquid temperature reaches a steady state. The tester then adjusts the pump speed in increments—typically from 50% to 100%.

During this phase, benchmarks consistently reveal a point of diminishing returns:

  • Too Slow (Under 50%): Coolant flow is restricted, resulting in significantly higher CPU temperatures.
  • The Sweet Spot (70% to 85%): The coolant moves fast enough to prevent heat soak at the cold plate, achieving near-maximum thermal transfer.
  • Maximum Speed (100%): Running the pump at max speed often yields only a 0.5°C to 1°C improvement over 80% speed, while drastically increasing high-pitched pump whine and mechanical wear.

Isolating Fan Speed

Once the pump’s behavior is mapped, testers lock the pump at a constant speed (usually 100% to eliminate it as a bottleneck) and test the radiator fans across their RPM range.

Unlike the pump, fan speed scales much more linearly with thermal performance. Radiators have high fin density, which restricts airflow. Fans must generate enough static pressure to push air through this resistance. In benchmarks, increasing fan speed from 1000 RPM to 2000 RPM will drastically lower CPU temperatures. However, turbulence and motor noise increase exponentially with fan RPM, making fans the primary source of noise in an AIO system.

Noise-Normalized Benchmarking

Running both the pump and fans at 100% only measures maximum theoretical cooling capacity, which is often intolerably loud for daily use. The gold standard for modern hardware benchmarking is noise-normalized testing.

In this methodology, testers use a sound meter positioned at a fixed distance to lock the cooler's total noise output to a specific target, such as 35 dBA or 40 dBA. Optimizing pump vs. fan speed is critical here:

  1. Prioritizing the Pump: Because pumps are enclosed in the CPU block (or radiator) and move smaller blades, they generate less audible noise than 120mm or 140mm fans. Testers will set the pump to the highest possible speed that remains inaudible below the target dBA threshold (often 80% to 100%).
  2. Tuning the Fans: With the pump maximizing coolant flow silently, the tester then raises the fan speeds until the sound meter hits the exact target decibel level.

By optimizing the pump to do the heavy lifting quietly, the fans do not have to spin as fast to achieve the same thermal result, leading to a much higher score in noise-normalized benchmarks.

Applying Benchmark Data to Real-World Curves

The data gathered from pump and fan benchmarking directly informs how users should configure their systems in the BIOS or control software. Because liquid takes much longer to heat up and cool down than air, benchmark data proves that tying your pump speed to fluctuating CPU temperatures is inefficient and causes unnecessary audible revving.

The optimal configuration derived from hardware benchmarking is straightforward:

  • Pump Configuration: Set the pump to a static, constant speed (usually between 75% and 85%). This ensures maximum flow rate during sudden CPU load spikes while avoiding the acoustic penalty of running at 100%.
  • Fan Configuration: Set the fans to a dynamic curve based on coolant temperature (if the AIO has a liquid temp sensor) or a smoothed CPU temperature delay. This keeps the system silent during idle tasks and only ramps up airflow when the liquid is actually saturated with heat.