Same here

Athlon 2500XP
512Mb RAM

WCGrid_Rosetta process
Memory usage : 6 332 Ko
Peak memory usage : 58 696 Ko
Virtual Memory size : 297 336 Ko

UD.exe (the interface)
Memory usage : 11 196 Ko
Peak memory usage : 16 012 Ko
Virtual Memory size : 9 828 Ko

Clearly, the software will be a drain for any machine below 256 Mo ram.

Reducing the memory footprint should be a priority or most users will experience slowdowns and simply give up.

I've actually been running it on a few of the used compy's we have at the shop lying around and even some p 166's with 64meg of edo seemed to handle it fine sure it takes longer to load but once it's done it kinda hides away

I noticed this too. As a professional software engineer, I don't know much about protein folding, but this seems like a stupidly large amount of memory. You can't say it has anything to do with the "code base" size, because this couldn't possibly be more than about 2 MB!

It is just another thing that will put people off using it. It has to have virtually no effect on the user's PC for them to consider using it.

Dito!

300MB increase in swap file. Its the LARGEST VM usage in the list - much greater than photoshop (which is only 50MB typically with no document open. It did go to 100MB once though) and any other program.

I'm using WinXP. In the task manager, it shows:
Mem usage: 5K (no problem there)
VM Size: 298,000K

Hurray, Rick!! Yes, I have noticed that the working page set is only a quarter as large as the VM allocated. So the only time this really affects the computer is when Rosetta is first loaded into memory, usually at boot when a lot else is also going on. But I can sympathize with the shock and denial that others seem to feel.

I am going to do the “Back when I was a GOOD programmer” routine so anybody who does not want to be terminally bored can mutter “WAS is the important word” and stop reading. I can still remember 30 years ago when I was working on a gamma-ray spectral analysis program and had to invert a matrix. It was with a sense of shock that is hard to understand now that I suddenly realized that it was an impossibly huge problem. Our new 1970s analyzer was just a small machine we put into a corner in an upstairs lab room, but it could produce much more data than a 1960s machine at about a tenth the price. But I could not analyze all those spectral lines in our computer. By dropping three quarters of them I could cut the matrix down to 4 megabytes in single precision. But this was not a sparse matrix, and paging in VM would be prohibitively time consuming. I worked out the time and space requirements for several different approaches and took them to our department head. Then I was assigned a new project. These days it is impossible to describe the visceral shock produced by a simple program that could require 5 MB to 129 MB, depending. But if you had spent as much time researching a $5,000 proposed expenditure for an additional 16 KB memory for our minicomputer, you might understand how Bill Gates would later describe 64 KB as enough for any home computer. Later, working on Numerical Analysis programs in Van Vleck Hall I would cynically keep in mind that the exercises in the text books were carefully chosen to be solvable in a tiny memory footprint. The real world required hundreds of megabytes that only a few National Laboratories could afford.