Comment by stochastic_monk

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?