Run a real WebGL/WebGPU rendering workload on your graphics card and watch live FPS, frame time, and stability — honest, measured numbers only. No fabricated temperatures, clocks, or utilization percentages.
A sustained GPU workload increases power draw, fan noise, and heat. Keep intake vents clear, run laptops on AC power, and click Stop at any time if the chassis becomes excessively hot or the system feels unstable. The test never auto-starts on mobile.
Rolling rendering frequency during the run
Browser security sandboxes block web pages from reading low-level hardware sensors. We never fabricate these values — any browser tool that claims to show them without a native desktop companion is guessing.
How this browser test works, what the numbers mean, and where the browser's honest limits are.
A GPU stress test places a sustained, computationally demanding 3D rendering workload on your graphics processor. It exercises vertex transformations, rasterization, and fragment-shader math to verify thermal stability, frame pacing, and hardware resilience under heavy graphical load — the kind of pressure a demanding game or 3D application would apply over time.
The tool uses WebGL 2.0 (with a WebGL 1.0 fallback) to render thousands of instanced 3D objects with a procedural noise fragment shader. Every frame, your GPU transforms geometry and shades pixels through genuine hardware-accelerated draw calls. Higher stress levels raise the instance count and shader iteration depth, so the workload scales with real GPU effort rather than a cosmetic animation.
Timing comes from requestAnimationFrame deltas and performance.now(). From the interval between rendered frames we derive current FPS, average FPS, minimum and maximum FPS, per-frame time in milliseconds, and a stability percentage (the share of frames that land within ±20% of the average frame time). These are all measured on your machine — nothing is estimated or invented.
Frames per second is how many complete images your GPU draws each second. Higher is smoother. By default the browser synchronizes rendering to your monitor's refresh rate (VSync), so a 60 Hz display typically caps at ~60 FPS and a 144 Hz display at ~144 FPS even when the GPU could render faster. Watch how steadily the number holds, not just its peak.
Frame time is how long a single frame took to render, in milliseconds. It is the inverse of FPS: 16.7 ms equals 60 FPS, 6.9 ms equals 144 FPS. Frame time is often more revealing than FPS because occasional long frames (stutters) are visible as spikes even when the average FPS still looks healthy.
No. Browser security sandboxes prohibit web pages from reading GPU or motherboard temperature sensor diodes. Any website that displays a GPU temperature without a native desktop companion app is fabricating the value. This tool honestly reports temperature as not exposed by the browser.
No. Graphics APIs available to web pages do not expose an operating-system GPU utilization percentage such as “98% load.” What we can honestly report is the rendering throughput your GPU actually delivers — FPS and frame time under the configured workload — not a synthetic load meter.
Not reliably. WebGL provides no API to query total physical video memory or current system-wide VRAM allocation. It can report driver-reported limits such as maximum texture dimensions and viewport bounds, which is what the GPU Information card shows — real capability data, not a memory gauge.
As the GPU sustains heavy calculations, temperatures rise. To stay within a safe thermal and power envelope, modern GPUs automatically lower their clock frequencies — thermal throttling — which shows up as a gradual FPS decline during a long run. A healthy, well-cooled card holds its frame rate far more steadily than one with failing fans or dried thermal paste.
A benchmark measures peak performance over a short run to produce a comparable score. A stress test instead applies a sustained load to expose stability, thermal behavior, and frame-pacing consistency over time. This tool is a stress test: it reports live measured metrics and deliberately does not assign an arbitrary comparison score.
Under normal conditions a stress test will not damage a working graphics card — firmware protections throttle or shut down before dangerous temperatures are reached. Still, keep vents clear, run laptops on AC power, and stop the test if the chassis becomes excessively hot or the system feels unstable. On mobile devices, prefer Light or Medium; the test never auto-starts there.
Common questions about GPU stress testing, thermal behavior, and browser limits.
