Backends

Classical backend dropdown in import settings now has three options:

Automatic

Automatic mode means that if GPU is available on machine generation will run on GPU, if no will run on CPU. Read more about them in corresponding sections

<aside> ℹ️

Automatic backend is default, and after updating to 1.3.x all your textures with “GPU Legacy” option will be “Automatic”

</aside>

GPU

Default GPU backend that was there from the beginning. Fastest of both, its still preferable to use it in any case. CPU was added mainly for build machines and WebGL support, but whenever GPU is available us it.

<aside> ℹ️

If this backend selected explicitly but no GPU will throw error

</aside>

CPU

CPU backend was made as fallback for GPU for cases where GPU is unavailable. Main use cases:

Build machine without GPU. Previously resulted in empty texture without errors Fixing build machine issue.
WebGL build. Runtime generation in WebGL build is not working with GPU since some of the Graphics functions is not accessible, therefore need to fallback for CPU generation. This is valid only for runtime generation, not editor.

<aside> ℹ️

If generating in runtime in WebGL build CPU backend will be used, however Burst don’t support WebGL, so all generation will be single-threaded, which makes it pretty slow.

</aside>

CPU backend is based on Jobs/Burst compiler setup, and mimics the same jump flood fill algorithm as GPU backend does. It is pretty fast due to parallelism, but GPU is still around 10 times faster.