No, he is correct. The GPU on the other side of the PCI is a remote computer.
Huge difference over operating over a PCI bus with gargantuan bandwidth and near zero latency, and working over random network connections.
Certainly a huge difference, but if you want good performance you need to treat this as a remote buffer management problem in both cases.
Sorry, but I can’t take the suggestion that the PCI bus and the internet should be treated the same seriously. You’re telling 4 or 5 orders of magnitude difference. Maybe more on some specs.
It’s like saying you should use the same file access algorithms for a RAM disk and punch cards. No you shouldn’t!
GPUs are highly complex and performance under generic wrappers.