Practical PowerPivot Limits?

They really tout PowerPivot as the future of self-service BI, but every time I try to use it on either our relational database or cube, I am wholly unimpressed.  It’s an absolute dog on our data.  I haven’t had the patience yet to load up our smallest segment of historical hospital data – Maine, with 5 million discharges, much smaller than say, Florida with 80 million discharges – into PowerPivot.

And if you attempt to load up from a cube, you have to specify an MDX query to tell it what data you want.  Having no experience with MDX, and the designer taking many minutes to refresh, I basically give up.

Is using PowerPivot practical against a large data warehouse?  (I'm assuming that 330 million rows in my fact table constitutes a large data warehouse, correct?)   Are there any practical limits to what PowerPivot will reasonably handle?  Is anyone else out there having performance issues with PowerPivot on large datasets?

My test computer is a developer laptop (better than what our clients use) connecting to our new servers over a fast internal network, btw.

Aha, one of my problems is a 32-bit machine, it seems.   Found this article, which describes exactly the error message I got, 20 minutes later, when I let it load up a dataset:

http://sqlblog.com/blogs/marco_russo/archive/2010/01/26/memory-considerations-about-powerpivot-for-excel.aspx

I guess I'll be experimenting with PowerPivot when I get a new development machine, which is not in this budget cycle, unfortunately.

Just as a point of reference, what kind of hardware are you running on exactly? How much memory would probably be most useful detail. I'm curious as to exactly how you setup your data and query since I've seen PowerPivot crunch 20+ million rows within seconds on a netbook. Another issue I've seen is exactly HOW you're trying to slice up your data. Have you tried looking at the exact same data via Analysis Services and seeing how quickly/slowly it loads there?