If you go to:
Settings -> User Preferences -> Save
and select Compression "none", then you can save the QVW and see how much memory it will take up when loaded into memory. It is usually 4-10 times more than the compressed QVW.
As a rule of thumb each user will increase the reports memory by an additional 10%.
The 450 mb sounds like there's some sort of overhead.
I changed the compression from High to None & I am surprised that the document size stayed at 4 MB. Maybe it does not make much of a difference when the size is that small.
I am also puzzled as to why the document when loaded with one user, the qvs.exe jumps from 16 MB to 130 MB (net increase of 114 MB) when the document itself if only 4 MB (almost 28 times as against expected 10 times). Where should I look to resolve the overhead which you are talking about?
Anyway, does anyone think that it would be impractical for me to go with only 830 MB available physical memory when my document is this small?
There is a lot more going on than just the amount of raw data in the document.
Each UI element requires some memory to load and display. If a user opens a chart, that chart gets cached - in memory. The more complex and large the chart, the more memory required.
Each user will require their own memory allocation to handle display and caching (although there is a certain amount of shared caching since v9). The rule-of-thumb is 10% of the in-memory document size but this is only ever a very rough rule of thumb and always depends on the document and what is in it.
The on-disk size of a QlikView document is almost never the same as the in-memory size.