Yes and no. Some type of deferred structure is almost always part of a GPU renderer as it's necessary for batching and reordering which help performance tremend

Yes and no. Some type of deferred structure is almost always part of a GPU renderer as it's necessary for batching and reordering which help performance tremendously. Sometimes this is entirely an internal system behind an imperitive API, though. Like skia only offers an imperitive API even though it builds a deferred rendering structure from that under the covers. So you end up doing a scene graph to imperitive API to internal renderer scene graph in some UI toolkits. In others they may share the same scene graph structure, applying those optimizations directly from the initial graph. Although all that said a "pure" scene graph is often overkill and slow. They were all the rage in the early 2000s but less so these days. QML's QtQuick looks like the primary remaining example?