I assume he was making the common mistake of confusing "virtual memory" with "swap file", which is so ubiquitous these days that it's rarely worth correcting. It's obviously possible to have virtual memory without page faults, Windows CE for example does it across the board (it doesn't even support swapping).
Speaking of terminology.... "swap" is incorrect, too. There hasn't been a "swap file" in Windows since the 9x days. The terms "swap" and "page" are also not synonymous in Windows like they are in Unix-based systems -- in Windows parlance, it refers to moving the entire contents of a program's data to a disk file. In the pre-9x days, this was 386spart.par, and that file had to be contiguous on-disk and unchanging in size. With the advent of 386 enhanced mode support in Windows 3.1, it became possible to also do temporary, page-based storage in win386.swp, which had much different operating mechanics from segment-based memory swapping. Don't be fooled by the file name or the historical references, it's much more like pagefile.sys than 386spart.par.