How Locking CPU Frequency Isolates IPC in Benchmarking

This article explains the methodology behind isolating a CPU's Instructions Per Clock (IPC) during hardware benchmarking by locking processor frequencies. By removing clock speed variations, dynamic boost algorithms, and thermal throttling from the testing equation, hardware reviewers can accurately measure and compare the raw architectural efficiency of different processors.

The CPU Performance Equation

To understand how locking frequencies isolates IPC, you must first understand the basic formula for CPU performance:

Performance = IPC × Frequency (Clock Speed)

  • Frequency refers to how many cycles a CPU executes per second, measured in Gigahertz (GHz).
  • IPC refers to how many tasks (instructions) the CPU can complete within a single one of those cycles.
  • Performance is the final output, such as a benchmark score or the time taken to render a video.

Because performance is the product of both IPC and frequency, you cannot measure IPC directly using a standard benchmark if the frequency is constantly changing. By locking the frequency to a static number across different CPUs, frequency becomes a constant variable. Consequently, any difference in the final performance score is strictly the result of IPC.

The Problem with Default CPU Behavior

Modern processors do not run at a single, fixed speed. Technologies like Intel Turbo Boost and AMD Precision Boost dynamically adjust clock speeds thousands of times per second. These adjustments are based on:

  • Current operating temperatures.
  • Power delivery limits (wattage).
  • The number of active cores being used.

If you benchmark two CPUs out of the box, one might run at 5.2 GHz while the other runs at 4.8 GHz. While this shows you which CPU is faster overall, it tells you nothing about which CPU has the superior underlying architecture (IPC).

How Benchmarkers Lock Frequencies

To conduct an "iso-frequency" (equal frequency) test, hardware testers manually override the CPU's default behavior through the motherboard's BIOS. The process involves several strict steps:

  1. Set a Fixed Multiplier: Testers choose a baseline frequency that all CPUs in the test group can stably achieve without overheating—typically between 3.5 GHz and 4.0 GHz.
  2. Disable Boost States: Turbo Boost, Precision Boost, and any motherboard-specific performance enhancers are turned off so the CPU cannot exceed the chosen frequency.
  3. Disable Power-Saving States: Features like Intel SpeedStep or AMD Cool'n'Quiet are disabled so the CPU does not drop below the chosen frequency when idle or under light loads.
  4. Isolate Threads and Cores: IPC is best measured in single-threaded workloads. If multi-threaded workloads are used, testers must disable extra cores so that all tested CPUs have the exact same core and thread count.
  5. Standardize Memory: Because IPC is heavily influenced by how fast the CPU can fetch data, testers lock the RAM speed and timings to identical values across all test systems.

Calculating the IPC Difference

Once the CPUs are locked to the exact same frequency and matched in core count and memory speed, the tester runs a benchmark.

Because both processors are ticking at the exact same rate (e.g., 4.0 billion cycles per second at 4.0 GHz), the clock speed variable is completely neutralized. If Processor A completes the benchmark 15% faster than Processor B under these locked conditions, the tester can definitively conclude that Processor A has a 15% higher IPC for that specific workload. This proves that Processor A's internal design is more efficient, regardless of what clock speeds the manufacturer chooses to ship the product with.