EN
TestYourSetup / PC and printer / Volume shader test

Volume shader test: a raymarched 3D fractal to benchmark your GPU

Render a rotating 3D fractal cloud on your graphics card and watch the frame rate. Raise the quality and resolution until it stops being smooth.

Quality
Resolution
More options: graphics API

Weak devices can freeze: raymarching is heavy. On an older laptop or phone start with Low quality at 50%, and close other tabs. The test stops when you press Stop, after 10 minutes, or when you switch tabs.

Nothing runs until you press Start.

Runs only on your device, only after you press Start. Nothing is uploaded and nothing is installed.

A volume shader test renders a 3D shape made of density instead of surfaces, a glowing fractal cloud, by marching a ray through it for every pixel on the screen. It is one of the heaviest things a graphics card can draw, which is why volume shader demos, often called "volume shader BM", became a popular quick GPU benchmark in the browser. This page is our own version, written from scratch for this site: a 3D Mandelbulb fractal raymarched as a glowing, rainbow-coloured volume through WebGPU or WebGL2, four quality levels from 48 to 384 steps per ray, resolution from 50% to 150%, and a live frame rate with average, frame time and 1% low. Nothing runs until you press Start, and it stops by itself when you switch tabs. For a single score across several kinds of work, use our GPU benchmark; for minutes of full load and throttling, the GPU stress test.

How does the volume shader test work?

For every pixel, the shader starts a ray at the camera and takes a fixed number of small steps through the scene, the "steps" in the quality setting. At each step it runs up to 7 iterations of a 3D fractal formula, a power-8 Mandelbulb, to find how dense the cloud is at that point, adds a little colour for that density, and dims what lies behind it. The colour is a rainbow hue that shifts with the position and the viewing angle. Where the ray first meets the surface, 4 more fractal evaluations work out the light and the sheen. When the ray has passed through or the cloud has become opaque, the pixel is done. The fractal turns slowly, so every frame is new work. Press Save image while it runs to download the frame with your frame rate.

Worked example: how much work one frame is

At 1920 x 1080 there are about 2.07 million pixels. At Medium quality each pixel takes 96 steps plus 4 for the lighting, and each runs up to 7 fractal iterations, so one frame is up to 2.07 million x 100 x 7, about 1,452 million iterations. Rays that miss the fractal, escape early or hit opaque density stop sooner, so the real number is lower. The table shows how the quality setting scales that work:

QualitySteps per rayFractal iterations per 1080p frame (at most)Relative work
Low48755 million1x
Medium961,452 million1.9x
High1922,845 million3.8x
Ultra3845,632 million7.5x

Resolution scales the work by the pixel count: 50% renders a quarter of the pixels of 100%, and 150% renders 2.25 times as many. So Ultra at 150% is about 17 times the work of Low at 100%, and about 67 times Low at 50%.

How to run the volume shader test: step by step

Step 1: Pick a quality and a resolution

Medium at 75% suits most laptops. Low at 50% is safe for phones and older PCs. High and Ultra at 100% or 150% are for gaming graphics cards.

Step 2: Press Start

The cloud starts rotating and the frame rate appears after a second. Keep the tab in front: switching tabs stops the test.

Step 3: Change settings while it runs

Raise the quality or resolution until the frame rate drops below what you consider smooth, usually 30 or 60 fps. Full screen renders at your whole screen size.

Step 4: Stop and read the result

The summary gives the average and 1% low frame rate at the settings you finished on.

What fps is good in the volume shader test?

There is no single good number, because the settings change the work by up to about 67 times. As a rough guide for any setting: under 15 fps is very heavy for your GPU, 15 to 30 is playable but not smooth, 30 to 55 is smooth enough for most people, and 55 or more is smooth with room to raise the quality. The most useful way to read it: find the highest quality and resolution where you still get 30 or 60 fps, and compare that with another device or a friend at the same settings. If your monitor runs at 60 Hz, the frame rate cannot go above about 60 however fast the GPU is; the refresh rate test tells you your limit.

Volume shader BM and cznull

Many people search for "volume shader BM" or "cznull volume shader". The original is an open-source WebGL demo by the developer cznull that raymarches a fractal volume and shows the frame rate, and it spread as an easy way to compare phones and graphics cards. We did not copy it: our shader, colours, settings and measurements are written for this site, with quality levels, resolution scaling, a 1% low figure, a result image and an automatic stop. Numbers from the two tests are not comparable with each other.

Why does the volume shader test lag or freeze my device?

Raymarching multiplies work by the step count for every pixel, so the load climbs much faster than in an ordinary game scene. A phone or an older laptop can need more than a second per frame at High or Ultra, and while the GPU is busy the whole screen can stop responding. Start at Low and 50%, and step up from there. If a frame takes more than 5 seconds, the test stops by itself. On a laptop with two graphics chips, make sure the browser uses the dedicated one (Windows: Settings, System, Display, Graphics, set the browser to High performance).

Volume shader test vs GPU benchmark vs GPU stress test

  • Volume shader test (this page): one heavy scene, settings you control, live frame rate. Good for a quick feel of how strong a GPU is, and for comparing devices at the same settings.
  • GPU benchmark: four different scenes, including this volume, run at a load set for your GPU, ending in one score on a fixed scale. Better for comparing over time.
  • GPU stress test: up to 10 minutes of fixed full load with a chart, to see whether the card slows down when hot.
  • GPU test: a 20 second health check that the card renders correctly and smoothly.

Volume shader on phones, Macs and Chromebooks

The test runs in any browser with WebGL2 or WebGPU: Chrome, Edge, Firefox and Safari on Windows, macOS, Linux, ChromeOS, Android and iOS. Phones have very high pixel densities, so 100% on a phone can mean more pixels than a laptop screen; 50% is a fairer starting point. If the page says it cannot start, the WebGL test shows what your browser supports and how to turn on hardware acceleration.

Why we built this volume shader test

Volume shader demos became a popular way to check a graphics card in the browser, because one heavy scene shows quickly whether a GPU is strong or weak. The pages that host them often make that check harder than it needs to be. Nitin Raja, who builds this site, tried a lot of volume shader pages before building this one: many start drawing as soon as the page opens, run at one fixed setting and show a single fps number. On a phone or an older laptop that can freeze the screen before you have read anything, and one number at one setting says little when you want to compare two devices.

So we wrote our own shader, a power-8 Mandelbulb volume, from scratch instead of copying the well-known cznull demo, and built the page around what people need: a Start button, settings that change the load by known amounts, a frame rate that follows the GPU, and a stop that happens by itself when something goes wrong.

What this volume shader test measures

The shader and the work per frame are explained above in how the volume shader test works. In short, this is what the page reads and computes:

  • Inputs. Quality (48, 96, 192 or 384 steps per ray), resolution (50% to 150% of the test area times your screen's pixel ratio, at most 3,840 pixels wide) and the graphics API: WebGPU when the browser offers an adapter, otherwise WebGL2.
  • Frame time. Every frame is timed in a requestAnimationFrame loop. A fence (fenceSync) in WebGL2 and queue.onSubmittedWorkDone() in WebGPU tell the page when the GPU has finished its work.
  • Frame rate. Once a second: frames divided by their total time. The average is the mean of those values since you last changed a setting. The 1% low is 1,000 divided by the 99th percentile frame time in milliseconds. The verdict bands are under 15, 15 to 30, 30 to 55, and 55 fps or more.
  • Renderer name. Read from WEBGL_debug_renderer_info or the WebGPU adapter info, only to warn you when it is a software renderer such as SwiftShader or llvmpipe.

What it does not measure: GPU temperature, clock speed, power draw or memory, which browsers do not expose. Its frame rates are not a score on the scale of any other benchmark.

How accurate is the frame rate?

The page times every frame it hands to the graphics card and keeps at most two frames queued, so the frame rate follows the GPU rather than the browser. Three limits apply:

  • Refresh rate cap. A browser never shows more frames than your display's refresh rate, so a fast GPU at light settings reads 60, 120 or 144 fps and no more.
  • Browser and driver. WebGPU and WebGL2, and different browsers, can differ by 10% or more on the same PC. The result names the API it used.
  • Heat and power. Laptops on battery and hot GPUs run slower; the stress test shows that over time.

To know which card you are testing, see what GPU do I have; to see where it ranks, the GPU tier list.

Why this volume shader test is different

Each point below can be checked on this page or in your browser's developer tools.

  • Runs only on your device. The page sends nothing to a server: the result is text you copy or an image you save.
  • Software rendering is flagged. If the browser draws with the processor instead of the graphics card, the status line says so, so a slow result is not blamed on the GPU.
  • One plain sentence as the result. Average and 1% low, the quality, the exact resolution in pixels, the graphics API and how many seconds were measured.
  • A result image. Save image downloads the frame on screen with its frame rate and settings line, drawn in your browser.
  • 15 languages. The same test and the same explanations in English and 14 other languages.

This volume shader test vs other volume shader pages

This testTypical volume shader page
StartsOnly when you press StartOften straight away on page load
Quality and resolution4 quality levels, 50% to 150%, full screenVaries; often one fixed setting
Frame rateNow, average, frame time and 1% lowUsually fps only
Graphics APIWebGPU with WebGL2 fallback, selectableUsually WebGL only
Stops by itselfOn a hidden tab, after 10 minutes or a 5 second stallRare
ShaderOur own, explained on this pageOften a copy of the same demo

Sources

Volume shader test FAQ

What is a volume shader test?

A volume shader draws a 3D object made of density, like smoke or a cloud, by stepping a ray through it for every pixel and adding up what it passes. It is one of the heaviest kinds of shader for a graphics card, so people use volume shader demos as a quick GPU benchmark in the browser.

What is the volume shader BM?

Volume shader BM is the name people use for a browser benchmark of this kind; the best known is an open-source WebGL demo by the developer cznull that renders a fractal volume and shows the frame rate. This page is our own version, written from scratch, with quality and resolution settings, 1% low frame rate and a Stop button.

What is a good fps in the volume shader test?

It depends entirely on the settings. As a rough guide, 30 fps looks fluid and 60 fps smooth. If you get 60 fps at High and 100%, your GPU has plenty of headroom; integrated laptop graphics usually manage Low or Medium at reduced resolution.

Why does the test freeze my phone or laptop?

Each pixel runs hundreds of fractal steps per frame, which can take a weak GPU longer than a second. Start at Low quality and 50% resolution. If the device still freezes, close the tab; the test also stops itself if no frame arrives for 5 seconds.

Is the volume shader test safe for my GPU?

Yes. It loads the graphics chip like a demanding game for as long as you let it run, at most 10 minutes. GPUs slow down by themselves if they get too hot.

Why is my score different from someone else's?

The resolution, the quality, the browser, the graphics API (WebGPU or WebGL2), the driver and the power mode all change the frame rate. Compare runs only at the same settings, and ideally the same browser.

Related tools