WebGL 2 / Visual performance lab
Three deterministic shader scenes. One fixed 720p workload. A transparent result you can repeat in the same browser.
Benchmark protocol
Standard run · 1280×720 · 30 sec
Sequence progress
0%
Elapsed
0.0s
Live FPS
0
Scene controls
Keep this tab visible. Switching tabs or hiding the canvas invalidates the timed run.
WebGPUStress Journal · Graphics desk
The Shader Desk
Browser graphics · Field edition · Free to run
A shader is the invisible instruction behind every pixel on this page. Our test turns that hidden work into a visible report: three scenes, one controlled sequence, and enough evidence to tell speed from consistency.
The subject
This is a visual WebGL 2 benchmark. Instead of hiding the workload off-screen, it lets you watch increasingly complex fragment shaders while the browser records how quickly and consistently each frame arrives.
Why it matters
GPU model names and peak numbers say little about browser rendering today. A live run can expose slow shader execution, uneven pacing, visual corruption, and results that refuse to repeat.
The readers
It is built for people who need a useful visual signal before installing a full benchmark suite—or before deciding whether a graphics problem deserves deeper investigation.
Our reporting method
Neon Contours, Chrome Signals, and Raymarch Relay each receive ten seconds at the same internal resolution. Their frame records are then combined without letting one unusually easy scene dominate the headline score.
Editor’s timing note
The standard run is the fast, comparable edition. Continuous mode is the follow-up investigation for visible artifacts, heat-soak behavior, and the consistency of a longer session.
Method, not mystery
The workload stays fixed so repeated runs are useful. Browser, driver, power mode, and display refresh rate still affect the result.
01
Ten seconds of layered procedural lines and wave fields at a fixed internal resolution.
02
Ten seconds of reflective lighting, ray intersections, and procedural environment detail.
03
Ten seconds of iterative signed-distance rendering with a fixed 72-step ceiling.
Read the signal
Overall throughput in each fixed scene. Higher is faster under the same conditions.
The slowest sustained slice of the run. A large gap from average FPS suggests uneven pacing.
Average relationship between 1% low and average FPS across all three scenes.
Geometric mean keeps one unusually easy or difficult scene from dominating the result.
Yes, within the browser. It measures a controlled WebGL 2 shader workload, not full game or compute performance.
Browser rendering is synchronized with the display. The heavier scenes create separation by missing frame deadlines under load.
Yes, but use the same browser version, power mode, display conditions, and 720p standard run.
Stop the run. Repeat with stock clocks and current drivers; persistent corruption can indicate instability.
Local / No account
Up to 20 reports, stored only in this browser. A pinned baseline is retained. Imported reports are unverified external data.
Imported report text is shown in its original language.
No saved reports yet. Run a test or import a report.