Thanks! I've written code for discrete GPUs but not integrated, and so I'm unsure what the advantages would be. Would there be lower latency for transfers, then, if not higher throughput?