How this browser stress benchmark works
Unlike basic WebGL visualizers that merely rotate static textured polygons, this benchmark applies sustained computational load across both graphics processing pipelines and multi-threaded CPU architectures:
- Procedural Shader Compute: The WebGL canvas runs deep raymarching loops with trigonometric noise octaves per fragment, forcing GPU stream processors to calculate complex floating-point geometry without relying on local asset caches.
- Dynamic VRAM Allocation & Buffer Sampling: The tool allocates and cycles through blocks of uncompressed 32-bit floating-point textures (up to 2048 MB under 98% load). Reading and overwriting these memory addresses during every frame ensures the GPU memory bus remains saturated, simulating heavy gaming textures.
- Multi-Core Worker Pool: The app detects your logical CPU core count (
navigator.hardwareConcurrency) and launches an array of background Web Workers executing recursive SHA-256 digest hashing and matrix calculations, putting every CPU thread under real load. - Dynamic System RAM Allocation: The tool actively reserves chunks of memory buffers directly in client RAM, tracking live consumption alongside browser heap metrics.
How pure VRAM texture allocation saturates graphics memory
Modern graphics cards feature high-bandwidth video RAM (GDDR6 / GDDR6X / HBM) to store mipmaps, render targets, and geometry data. In the Pure VRAM Test, the application loads high-resolution uncompressed texture buffers into WebGL texture memory units and dynamically updates coordinates while rendering tens of thousands of active 3D particle objects simultaneously. This bypasses driver compression algorithms and provides a real-time estimation of VRAM saturation, frame timing jitter, and canvas backbuffer footprint.
Choosing the right stress level (40% to 98%)
Each load profile adjusts shader iteration depth, render target scaling, and CPU thread duty cycles:
- 40% (Moderate): Uses half-resolution canvas scaling and light raymarching passes. Designed for ultra-portable laptops, office workstations, and integrated Intel UHD or AMD Radeon APUs.
- 60% (High): Elevates shader passes and allocates moderate VRAM textures. Tests typical 1080p and 1440p gaming loads without forcing immediate thermal limits.
- 80% (Extreme): Expands shader loops to 120 iterations with high texture cycling and active CPU threads. Ideal for validating stability on dedicated desktop GPUs (NVIDIA RTX / AMD Radeon RX).
- 98% (Max Torture): Runs 160 raymarching steps, maximum VRAM allocation, and continuous background CPU worker operations. Designed to test fan curves, airflow limits, and thermal paste degradation.
Interpreting CPU performance: Normal, Struggling, or Choked
As graphics cards execute draw calls, the CPU must synchronize instructions and process event queues. The benchmark evaluates responsiveness through event loop latency jitter:
- Normal Performance:
- Main-thread event latency stays below 18ms with minimal jitter. The CPU processes background worker math and browser draw commands smoothly without dropping frames.
- Struggling:
- Main-thread latency spikes between 19ms and 49ms with noticeable micro-stuttering. The CPU is experiencing core thread contention or starting to throttle clock speeds under heat.
- Choked / Severe Drop:
- Event latency exceeds 50ms, accompanied by severe frame drops (over 45% below initial peak performance). The browser engine is dropping draw calls due to memory bus saturation or severe thermal downclocking.
Detecting GPU thermal throttling and VRAM saturation
When running the 1-minute, 5-minute, or 10-minute profiles, monitor the Live FPS and Frame Drop indicators. If your system begins at 75+ FPS but steadily declines to 35 FPS as fans ramp up, your graphics processing unit is actively thermal throttling. Modern graphics cards automatically downclock core frequencies and drop supply voltages when reaching silicon temperature thresholds (typically 83°C–89°C on desktop cards, and 90°C–100°C on gaming laptops) to prevent physical failure.
Frequently Asked Questions
Can a browser benchmark actually make my cooling fans spin up?
Yes. By saturating all available logical cores with Web Workers and executing deep raymarching fragment shaders at full resolution, power consumption and silicon temperatures rise rapidly, prompting cooling systems to respond.
Why does the test show VRAM as "Allocated Buffer"?
For security and privacy reasons, web browsers do not grant raw physical access to graphics memory addresses. Instead, the test allocates actual WebGL texture buffers in GPU memory and updates them actively, accurately testing VRAM throughput.
How long should I run the benchmark to test stability?
A 1-minute test verifies immediate boost clocks and initial stability. To evaluate thermal saturation, fan curves, and sustained cooling efficiency, run the 5-minute or 10-minute benchmark.