With a definition such as:
- radial-gradient(closest-side, transparent 0%, transparent 75%, #...
The browser first parses the CSS, then generates the actual bitmap data for this definition in memory then finally paints the pixels into the correct places. The in-memory bitmap generation is far more taxing than simply tiling an existing bitmap.Then there is the issue of dynamic CSS which may cause reflow which will in turn probably require more dynamic bitmap generation.