Most likely compatibility is more stable across clients

GPUs aren't universally supported by browsers or available on client devices, and implementing a user-agent/telemetry based response adds overhead.