Using IOZONE... you can easily test disk IO, and set block size and file size. If you want to test with noatime off or on, you can easily do that too - and know how much it helps on your system.
http://www.iozone.org/
You can also use SYSBENCH - use that to do OLTP benchmarks, test memory speed, and cpu speed. We have seen incorrectly matched memory cause as much as a 30% slowdown in the new Intel QUADQUAD and QUADHEX Dunningtons.
http://sysbench.sourceforge.net/
We have several servers... with different RAID 10 ARRAYS, based on SATA drives, SAS (SCSI)drives, and even SSD (Solid State Memory) (Intel X25-E Extreme) drives. Some are 4 drive RAID 10, some 6 drive, and some are RAID 1. We don't use RAID 5.
I have a file somewhere where I did a bunch of comparisons but I can't find it right now. Sorry.
In a nutshell, a RAID 10 array of 4 SAS 300GB 15K RPM drives results in ~ 46,000 on the Random write test, while a RAID 10 array of 12 SAS 300GB 15K RPM drives results in ~ 300,000.
The RAID 10 12 Disc Array is even faster than a single SSD drive.
We have found the SSD drives in RAID don't do very well - seems the controllers are not really up to it yet, even the latest Adaptecs.
We use the Adaptec \ 51645 \ SATA/SAS RAID controller in all our servers. Big onboard cache, fast co-processor, and 8x PCI.
Warm regards.