Briefly,
I was able to install I-TASSER on Mandriva Linux under Virtual Box under Windows XP 32bit. Virtual Box and Mandriva Linux (just like majority of other Linux distributions) are free. Steps should be the same for other versions of Windows and for Mac.
Motivation: to get an idea how to install I-TASSER, while using familiar environment. This installation will not be usable (I-TASSER is prohibitively slow on emulated Linux for any real proteins), so if one plans to use I-TASSER suit, he will have to install Linux directly on hardware.
Steps:
(1) Download Virtual Box (VBox) https://www.virtualbox.org/
(2) Download 64bit Mandriva 2011 Linux CD image http://www.mandriva.com/en/downloads/ (I like the GUI, it is lightweight, yet powerful; VBox supports it naturally)
(3) Create virtual machine (64bit Mandriva Linux), add hard drive (I used 100 Gb, it might be possible to use less), add Linux distributive iso as CD-ROM; run the Linux machine; install Linux
(4) Install guest additions, add shared folders and add NTFS-3G drivers to support NTFS from Linux
(5) Download I-TASSER Suite package I-TASSER.tar.gz ,
sequence database nr.tar.gz (note that downloaders like Flashget seem to be blocked so if you get interrupt at 5.1G you will have to DL all the 5.2G again),
library files (namely PDB (~723MB), MTX (~260MB), DEP (~157MB), SIG (~169MB), CNT (~66MB), summary1 (~31MB),summary2 (~185MB))
(6) Move files to home/username/I-TASSER (using shared folders) and extract all these files; yes, the folder exists already, but you want to extract it here.
(7) Download(!) and install(!) PSIPRED version 2.6 or earlier (follow instructions in readme file).
(8) Create fasta sequence file (I used AAAAAAAAAAAAAAAAAAAA for testing purposes)
(9) Run job; it fails ... I had to install something else, but I don't remember what
(10) Search forum for the error that I-TASSER gives you and apply suggested fix.
(11) Rerun computation; wait for several hours .. I day .. get the result. Find it in the input folder. Open .pdb file and observe a helical structure, just as expected. The will be only one structure, since there is not many options for a short peptide.