WebGPU Test
WebGPU Test is a free online tool that checks if WebGPU works in your browser, and shows everything it reveals about your graphics card. It runs a live 3D rendering test with up to a million spinning cubes, runs compute shaders that verify the results of the GPU and measure its speed in GFLOPS and its memory bandwidth in GB/s, and lists every supported feature and every limit of your GPU.
Use it to find out if a WebGPU game, a 3D viewer or an AI model running in the browser will work on your computer, to see why it does not, to compare browsers and graphics cards, or to make sure the right GPU is being used on a laptop with two of them. Everything runs inside your own browser, nothing is installed and nothing is sent to a server.
How to use this online tool?
Just open the page: the test starts by itself. The checklist below shows each step, from finding the WebGPU API to drawing on the screen. If everything is green, WebGPU is working. Then press Run the compute tests to measure your GPU, and move the slider of the rendering test to add more cubes and see how the frame rate holds up.
Testing WebGPU...
Secure page (HTTPS)
WebGPU is only available on HTTPS pages.
WebGPU API
The browser exposes navigator.gpu.
Graphics adapter
The browser gives this page access to a GPU.
GPU device
A device is opened on the adapter.
Rendering
A 3D scene is drawn on a canvas.
Compute
Compute shaders run and return correct results.
GPU preference:
On a laptop with two graphics cards, the GPU preference asks the browser for the dedicated or the integrated one. Some browsers and systems ignore it, and always give the same GPU.
What is WebGPU?
WebGPU is the modern way for web pages to use the graphics card. It replaces WebGL, which was based on the old OpenGL, with an API designed after the current native ones: Direct3D 12 on Windows, Metal on Apple devices and Vulkan on Android and Linux. Besides drawing 3D graphics faster and with less work for the CPU, it adds compute shaders, which let a page use the GPU for general calculations. That is what makes it possible to run AI models, image and video processing, physics simulations and scientific computing directly in the browser.
Which browsers support WebGPU?
| Browser | WebGPU support |
|---|---|
| Google Chrome | Enabled by default since version 113 on Windows, macOS and ChromeOS, and since version 121 on Android 12 or newer. On Linux it may still have to be enabled by hand. |
| Microsoft Edge | Same as Chrome, since version 113. |
| Opera, Brave, Vivaldi | They use the same engine as Chrome, and follow it. |
| Mozilla Firefox | Enabled by default since version 141 on Windows, and since version 145 on Macs with Apple silicon. On other systems it can be turned on in about:config. |
| Safari | Enabled by default since Safari 26, on macOS Tahoe, iOS 26, iPadOS 26 and visionOS 26. |
Support grows with every release, and it also depends on your graphics card and its driver: a browser may support WebGPU and still refuse to use a GPU it knows to be unreliable. This page is the quickest way to know what your own browser does today.
How to enable WebGPU
- Update your browser and your graphics driver. Most problems disappear with a recent version of both.
- Turn on hardware acceleration. In Chrome and Edge it is in Settings → System → "Use graphics acceleration when available". Without it, there is no GPU for WebGPU to use.
- Chrome and Edge: open chrome://gpu (or edge://gpu) to see the WebGPU status and the reason it is disabled. If your system is not supported yet, enable chrome://flags/#enable-unsafe-webgpu and, on Linux, chrome://flags/#enable-vulkan, then relaunch the browser.
- Firefox: open about:config, set dom.webgpu.enabled to true and restart Firefox.
- Older versions of Safari: enable the Develop menu in Settings → Advanced, then turn on WebGPU in Develop → Feature Flags (called Experimental Features in older versions).
Flags are called "unsafe" because they skip the list of graphics cards and drivers the browser considers safe. They are fine for testing, but if the browser crashes or draws garbage, turn them off again.
Why is WebGPU not working?
| Problem | What to do |
|---|---|
| The page is not secure | WebGPU only exists on HTTPS pages (and on localhost, for developers). On a plain HTTP page, navigator.gpu is simply missing. |
| No graphics adapter | The browser has WebGPU but did not give a GPU to the page. Hardware acceleration is off, or your GPU or driver is on the browser's block list. Update the driver, and check chrome://gpu or about:support for the reason. |
| Old system or GPU | WebGPU needs Direct3D 12 on Windows, Metal on Apple devices or Vulkan on Android and Linux. Very old graphics cards and systems do not have them. |
| Remote desktop or virtual machine | There is often no real GPU inside, so you get no adapter, or only the slow software (fallback) adapter. |
| The GPU device was lost | The GPU took too long, or the driver crashed or restarted, and the browser reset it. Use fewer cubes and press "Run the test again". |
| The wrong GPU is used | On a laptop with two graphics cards, choose "High performance" above. On Windows you can also force it in Settings → System → Display → Graphics, by setting your browser to "High performance". |
| Low frame rate with few cubes | Check that you are not on the fallback adapter, that battery saver is off, and that the laptop is plugged in. Many laptops slow the GPU down on battery. |
What do the results mean?
- Frame rate: how many images per second the rendering test draws. It is capped by the refresh rate of your screen, so the interesting number is how many cubes it takes for it to drop.
- Correctness: the GPU adds two lists of 4 million numbers, and every result is checked on the CPU. A failure means the GPU, the driver or the browser is producing wrong numbers, and anything that relies on WebGPU computations may misbehave.
- GFLOPS: billions of 32 bit floating point operations per second. An integrated GPU usually gives hundreds of GFLOPS, and a gaming graphics card many thousands (several TFLOPS).
- Memory bandwidth: how many gigabytes per second a compute shader can read and write. It matters most for AI models and image processing, which move a lot of data.
- Fallback adapter: a GPU emulated on the CPU. It works, which is better than nothing, but it is many times slower than real hardware.
WebGPU vs WebGL
WebGL is supported almost everywhere and is still what most web 3D content uses, but it only draws graphics, and its design comes from the OpenGL of the 2000s. WebGPU talks to the GPU the way modern games do: it prepares work in advance instead of checking the state on every call, it has proper compute shaders, and it gives access to newer hardware features. The result is better performance, less load on the CPU, and things WebGL simply cannot do, like running large AI models in a web page. You can see the WebGL details of your browser in Check Browser Info.
Is my data safe?
Yes. This page talks to your graphics card directly in your browser, through the WebGPU API, and only shows the results on the screen. Nothing is recorded, nothing is stored and nothing is sent to any server. The copy and download buttons only create a file on your own computer.