Technical Writing

Insights.

Engineering-led writing on performance systems. Frametime analysis, thermal management, configuration audits, and the physics behind what makes a system feel fast.

All Frametime Thermals Configuration Bottleneck Analysis Memory Streaming
Featured
Frametime Analysis

Why Your 1% Lows Are Ruining Your Game — And What FPS Can't Tell You

Average FPS is the most commonly cited metric in PC gaming, and the least useful for understanding actual system performance. This article explains why frametime consistency determines how a game feels — and what to measure instead.

The problem with average FPS

Ask anyone what their system performance looks like and they'll tell you their average FPS. 200 FPS in CS2. 144 in Valorant. 85 in Cyberpunk. These numbers are not useless — but they tell an incomplete story that frequently leads to the wrong conclusions.

Average FPS is computed over a fixed capture window. A system that delivers 180 frames for three seconds and then freezes for one second has an average FPS of 135. It also feels terrible. The average metric cannot distinguish between these two scenarios.

A system running 200 average / 80 one-percent-low feels worse than a system running 140 average / 120 one-percent-low. The first system has higher throughput. The second system has higher consistency. Consistency wins, every time.

What 1% lows actually measure

The 1% low figure represents the average frametime of the slowest 1% of frames in a capture window. If you're running at 144 FPS average and your 1% lows are 60 FPS, it means one in every hundred frames took more than 16.7ms to render — the threshold for 60Hz delivery.

This matters because human perception is highly sensitive to frame duration inconsistency. A consistent 100 FPS feels smoother than an inconsistent 144 FPS. The stutter, the momentary freeze, the input-to-response lag spike — these are all frametime events, not FPS events.

  • 1% low / average ratio above 0.7 — acceptable consistency
  • 1% low / average ratio 0.5–0.7 — perceptible instability
  • 1% low / average ratio below 0.5 — significant gameplay impact

Common causes of 1% low degradation

Frametime spikes that degrade 1% lows without affecting average FPS are almost always caused by one of four things:

  • CPU thread contention — background processes (Discord, streaming software, browser) competing for CPU cycles during a frame's preparation phase
  • Thermal throttling — CPU or GPU reducing clock speed under sustained load to manage temperature, creating periodic frame delivery delays
  • Memory bandwidth constraint — RAM running below rated speed (XMP/EXPO disabled) creates a bottleneck that manifests in 1% lows before average FPS
  • DPC/ISR latency spikes — device driver interrupt handling consuming CPU time at the wrong moment in the frame pipeline

How to measure it correctly

CapFrameX is the correct tool for frametime analysis. It captures per-frame delivery time data and computes 1%, 0.1% lows, and frametime variance alongside the average. It is free, open source, and used in professional hardware reviews.

To capture a valid baseline: run a repeatable sequence in your target game — the same route, same scene, same duration each time. Compare captures with identical parameters. Variation between captures is expected; large variation indicates instability worth investigating.

If you're not measuring frametimes, you're not measuring performance — you're measuring throughput. They're related but they're not the same thing.

FPS Engineering · Frametime Analysis · 2024
Run Diagnostic →
More insights
Configuration
XMP and EXPO: Why 68% of Systems Are Running Their Memory Wrong
JEDEC default speeds are a compatibility floor, not a performance target. Here's what XMP actually does and why not enabling it is one of the most common and costly configuration errors.
Configuration · 8 min read
Thermals
Thermal Throttling: The Invisible FPS Tax You're Probably Paying
A CPU at 95°C under load doesn't announce itself. It silently reduces its clock speed, frame delivery becomes inconsistent, and the symptom looks like a game engine problem — not a cooling problem.
Thermals · 10 min read
Bottleneck Analysis
GPU at 65% Utilisation: What It Actually Means
A GPU that isn't fully utilised while performance is limited is not an idle GPU — it's a symptom of a constraint upstream. This article explains how to correctly diagnose what's actually limiting your system.
Bottleneck Analysis · 7 min read
Streaming
The Streaming Tax: How Encoding Affects Gaming Performance
Running a game and encoding a stream simultaneously places specific and predictable demands on your hardware. Understanding where the load goes is the first step to managing it properly.
Streaming · 11 min read
Memory
RAM Speed vs Capacity: Which Matters More for Gaming in 2024
The answer depends entirely on your current configuration. If XMP is disabled, you're running at 2133MHz regardless of what your sticks are rated for — and the correct fix is free.
Memory · 6 min read
Configuration
Power Plans and Gaming: Why Windows Balanced Mode Costs You Frames
Windows power plans affect CPU core responsiveness in a way that directly impacts frame pacing. The correct power plan for gaming is not the default — and changing it takes thirty seconds.
Configuration · 5 min read
Start Here

Run the free diagnostic first.

Before reading further — check whether your system has measurable constraint. The diagnostic takes two minutes.

Run Free Diagnostic