Well that all depends on a few things. First off, RAID stands for Redundant Array of Inexpensive Disks. There are several types of RAID, and depending on what you want to do they may or may not be good for you.
RAID 0 (stripe) is taking 2, or more physicaly different disks, and "merging" them together to form 1 disk. The data is written to and read from all disks at the same time, thereby increasing speed. This would be helpfull in loading the levels in a larger game such as Crysis, etc. The downfall to that is, if 1 of the drives fail, you are screwed. All your data on the drives is lost. Restore from backup. The other "downfall" is that your RAID array is only readable from the particular controler that created the array. So you can't pop your drives into a different tower and expect them to be readable.
RAID 1 (Mirror) is just what it sounds like. You have 2 identical drives, and your controller reads and writes the same data to both drives. Bad thing, twice as expensive, good thing, if 1 drive fails, you're still able to work, game, etc. When you replace the drive that is failed, the controler will rebuild the array automaticaly.
RAID 5 Raid 5 is really the entry point of "Higher End" RAID systems. What Raid 5 does, is stipe data across drives, but it also includes Parity information on the other drives. What this parity data allows is for the controler to contain the data of 1 of the drives in the array. RAID 5 needs a minimum of 3 identical drives to function, and will allow you to operate even when 1 drive fails. If 2 drives fail, you are screwed as well. In a 3 disk RAID 5 Array you will have Disk1, Disk2, and Disk3. Disk1 will contain Disk 1 Data as well as parity information on the Disks 2 and 3. Disk 2 will contain Disk 2 Data, and parity for disks 1 and 3, and Disk 3 will contain Data from Disk3, and parity from 1 and 2. So you see how if one of the drives failes, the other 2 have enough information in Parity to not only keep the array functioning, but when you replace the disk, the other 2 drives will rebuild the failed disk.
RAID 5 will typically give you roughly 2/3 actual usable capacity compared to what the full drive count is. 3 100Gb disks will yeild approx 200GB useable drive space.
That is typicaly where the "average" user will stay in regards to RAID Levels, with RAID 5 being less likely due to the cost of the disks. Now if you were a home business for example, RAID 1 (Mirroring) would be good for you. But if you wanted the benefits of Data Redundancy AND speed, then RAID 5 would be your best bet.
As it pertains to Gameing, I would recomend just staying with the single drive, and always, always have a backup solution for any "critical" files. Be it, copying them to DVD/CS or something like that.
If you are looking for extra data security, RAID 1 would definately be worth it. It uses two HDs in case one should die. Pop a replacement in and the info from the surviving drive will be transfered back.
If you are looking for more speed out of your machine, RAID 0 is the way to go. It treats the drives as one logical unit. If you had 10 pieces of information to write to a drive, a RAID 0 would write 5 pieces to one and 5 to the other. One would expect twice the performance, but that is hardly the case. Back when RAID arrays became popular, consumers had to contend with significantly slower HDs. In recent years, the speed of the drive has increased to a point that the gain one would experience is very small.
In my opinion, if someone is looking to get faster performance, you should consider a 10K RPM drive instead of a RAID setup.
There is generally no real world(vs. synthetic transfer rate benchmarks) performance advantage to raid of any kind.
Go to www.storagereview.com at this link: http://faq.storagereview.com/tiki-index.php?page=Single...
There are some specific applications that will benefit, but
gaming is not one of them. Even if you have an application which reads one input file sequentially, and writes
it out, you will perform about as well by putting the input on one drive, and the output on the other.
An interesting set of tests, dirtmountain.
What interested me was the difference between the HDtach results showing 100% improvement, and the sysmark2004 tests which showed little to 5% improvement. The real world is not synthetic, but more like the sysmark2004 system. Even then, the tests use a limited amount of data for convenience. In the real world, a user will have filled up a significant part of the drives, making the access delays longer. This will hurt the raid-0 case more because raid-0 needs two accesses vs. one.
The failure in these tests was using a single drive for comparison. If one bought a second drive for raid-0, the alternative comparison should have split the application data across the two single drives. I have yet to see a raid-0 vs jbod benchmark that does this.