Not same as PCI at all.
The main reason these graphics buses existed (vesa, agp..) was to provide a dedicated bus, PCI shares its bandwidth with all devices, these don't.
Then AGP 1.0 was already 2x faster than PCI.
Well AGP was based on PCI, appeared to the host largely as a second PCI bus that happens to have only one device, and PCI existed with 66 MHz and 3.3V variants which were widely used in high performance systems, just not in mainstream desktops. AGP did have AGP specific transactions, including split read transactions (which were also added to PCI as part of PCI-X) and to access the AGP memory window which let the card use system memory for texture storage avoiding the need for lots of onboard memory. But AFAIK, only a few cards made heavy use of that, as memory costs dropped and the performance advantages of having on board texture memory were too great to ignore.
Later revisions of AGP increased to speeds that PCI never supported, but it remained fundamentally based on PCI both at the physical and transaction model levels.