In principle, I'd have to agree with most of what russki said here.
RAID (all levels, not just RAID 5) has one primary purpose, that is availability. This is the capability to keep the system up and operational when a hard drive fails. This is a feature that servers need, not desktops.
RAID is not backup. It is not designed nor is suitable to replace a true off-line backup. The purpose of a backup is to save your data from any destructive cause -- OS corruption, malware/virus, accidental deletion/format, or to just plain go back to an earlier version of a document. RAID cannot protect against any of these things. The only destructive force that RAID can protect against is hard drive failure.
So since RAID isn't a substitute for backup, you're going to have to put a backup plan in place anyway. That fulfills protection from hard drive failure as well as everything else, leaving RAID as an availability solution only. Since a desktop seldom needs the availability of a server, you have to ask yourself what you're installing a RAID for?
Further, there is a problem with RAID on the desktop from a reliability standpoint. Enterprise-level RAID cards, like those from 3Ware, LSI, and Adaptec are robust and reliable. Do a quick search on this forum and you'll find dozens of threads that start out with statements similar to the following:
"I was running RAID on my desktop Intel chipset, and all of a sudden the array is listed as degraded ..."
"My nVidia RAID isn't listed in my bootable devices ..."
"My RAID array won't rebuild ... it keeps saying the new drive is bad ..."
Almost all of the desktop RAID controllers seem to have reliability problems of this type. An unreliable RAID that drops out the array or won't rebuild it when a drive goes bad is useless.
About the only other theoretical thing that RAID can give you is higher data transfer rates. But the question is, is that useful for your application? There are a few applications that benefit greatly from high transfer rates (Photoshop scratch disks, DVD authoring, media file creation and video editing applications, professional audio programs like ProTools). Then there are applications that do not benefit at all from sequential transfer rate. Most people here and benchmarks by several web sites offer evidence that gaming falls into this category - i.e. maybe a few percent decrease in load times, if that.
Thus, my conclusion is that unless you have a specific application that can benefit from RAID's high sequential transfer rates, or you have a specific application that needs the availability, the risk, complexity, and cost of using RAID on the unreliable desktop RAID chipsets is too high to justify the small performance benefits.
There is one situation not mentioned here where RAID can benefit, that is where you need to store a quantity of data that is beyond currently available drive sizes, and particularly if you have no means to back up that data. For example, if you want to build a media server to store audio and movies, and need to store 3 TB of data. In this case, RAID 5 would be very worthwhile, because it gives you at least some protection against hard drive failure, and is the only way to get a volume size >1TB right now (1TB hard drives are the largest available).
However, in this case, I definitely would not use a desktop chipset controller. An Enterprise RAID card is definitely warranted here due to the reliability and need to make sure that your data isn't compromised, especially since you can't back it up.
Furthermore, Enterprise level RAID cards perform much better than desktop chipsets -- not due to the parity calculation capability, but predominantly due to the large write cache that can do delayed writes. This prevents a lot of the write-read-calculate-write cycles that cause slow RAID 5 writes on desktop chipsets.
For a typical desktop these days, I like a standard 2-drive setup. Use a very fast drive for C: (Raptor, Velociraptor, or high-end 7200 RPM drive like Seagate 7200.11 or WD Caviar Black), and then a large 7200 RPM drive for D:. Install OS and programs on C:, data on D:. Invest in Acronis True Image, Symantec Ghost, or some other imaging program, and image C: to a file on D: once a week.
For me, that makes the fastest, cheapest, and most robust desktop system. Obviously, this can be modified to suit particular applications.