← Back to writing

August 8, 2026

A Practical Way to Explain Pixel Cost

A long-form technical note on profiling, visual evidence, and explaining rendering trade-offs to artists.

Start with the question, not the capture

A profiler capture is evidence, not an explanation. The first paragraph should tell the reader what decision the capture is meant to support.

A repeatable capture matrix

Record resolution, hardware, quality tier, camera distance, lighting condition, and whether the frame is representative. Without those conditions, two millisecond values can look comparable while describing different work.

Separating symptoms from causes

A material may be expensive because of texture bandwidth, divergent branches, overdraw, or a downstream pass that the material triggered. The portfolio explanation should show how the hypothesis changed after each capture.

capture.ts
const capture = {
  resolution: "1920x1080",
  gpu: "integrated",
  pass: "material-evaluation",
  frameMs: 2.1,
  textureSamples: 18,
};
External embed used to test long-form media handling.

What changed after the experiment

The final choice preserved the visible cue that artists used to judge the material. A smaller number is useful only when the reader understands what was kept, what was removed, and why.

Notes for future captures

  • Keep a stable scene and a representative scene.
  • Capture both the final view and at least one debug channel.
  • Report a range when scene variance matters.
  • Write the trade-off in plain language before listing implementation detail.

Related work

Applied in context