The question comes up from time to time, but is a poor question.
(which is why it doesn't have an answer).
To make a long story short, the resources consumed by extra users are minimal. The key question to ask is "how many requests per second". And the answer to that is complex.
To drive that number up
the following items play a role;
a) bandwidth (obviously)
b) Ram - the server can only access 2 gigs, but you want plenty for other
processes as well - like windows and Sql.
(On that note, the ideal is to have your Sql server on another machine) - or
if you're using TPS a fast hard drive.
We've got servers running comfortably on Windows 2008 with 512M of ram, the
server itself uses very little most of
the time, it's only when the thread-count rises that the Ram plays a part.
c) cores. NT is heavily multi-threaded so more cores means more performance.
servicing 100 requests per second is pretty easy - ad if you do the math
that translates to lots of individual users.
let's say a user makes a request every 30 seconds. And let's take 100 requests per second as a number. That translates into 3000 "similtaneous" users. But even that is really only a thumbsuck.
Let's assume a user is using the site (and doing something every 30 seconds) - but only interacts with the site for say 30 minutes in the day. (Typical banking app for example). And lets limit the day to 18 hours long (11pm to 5am is pretty quiet). so that's 6000 per hour, or 108000 per day.
All of this assumes of course you're limited to 1 server (and you're not really) and that your other resources will cope ok.
A safe answer to your original question -?
"thousands at a time, tens of thousands per day".
Cheers
Bruce