Fakeraids don't have battery backed cache. So if power is lost, all write-behind cache data is lost, or controller itself does not go beyond the write speed capabilities of hard drives involved.
E.g., measure write IOPS for hard drives (should be 75-100 for 7200 rpm, 125-150 for 10000 rpm, 175-210 for 15000 rpms). Then test write IOPS for controller. If numbers go high and there's no RAM and battery on board - then it's kind of fakeraid that's unsafe to use.
Personally I recommend ZFS over fakeraid if you can afford to use FreeBSD/OpenSolaris, as it does not require hardware controller, supports bit rot prevention via checksums for user data and metadata, can read from other disk in the mirror if checksum fails (unlike any top-notch hardware RAID), can make materialized-on-disk copy-on-write snapshots, can use SSD as a level 2 cache to speedup read IOPS and has many more features.
You only need $1000 hardware RAID if hundred write IOPS is not enough for your highly-loaded database server. In terms of reliability, it's better to use ZFS and few additional HDDs.