The three RAIDs are on an ICH9R; as the OS (and it's corresponding swap file, living on the 'opposite' pair)
are on RAID0's, I treat them as 'completely expendable', likely to be 'lost' at any moment (though, in practice, I've
never lost a RAID, of
either type,
even after the failure of an RE3) - when I'm 'happy with' an OS, and the 'state of its installation', I boot to a different OS (to 'unlock' the system files), and image it to a backup drive... My partition structure, master boot record, and extended master boot record, are all backed up to a bootable floppy, which also happens to contain the installer for my boot manager. My current BIOS, its installer, and my working set of CMOS parameters are backed up to a second floppy - which can 'blind flash', in case of total disaster!
Next workstation rebuild
will contain a 'pricey' RAID card (Areca 1680IX), but that's for a completely 'dedicated' thing - trying to optimize OpenCL with
huge data sets... Most RAID troubles are simple: don't try to do RAID5 on a desktop, 'FakeRAID' platform; keep an eye on the 'active' driver situation (once in a rare while, there's a bad one); treat any RAID0 setup as 'expendable'! That's, pretty much, IT! In the year an a half or so, that I've been using/specifying GB boards, I have
yet to see an actually bad MOBO! I (again,
very rarely) see an actual defect answering questions here, occasionally; but, far more common, are people RMAing boards (and, sometimes, two or three times!) due to their inability to get something
working correctly - because they don't know
how to 'get it working'... I, some months back, spent several hours walking somebody through getting an Areca card (say, $800!) working on his system, who
didn't know what it was! Had, I think, three other cheap drive controller cards, and, apparently, every hard drive he had
ever owned 'slopped into' the thing!
"I have" a ZERO tolerance if a MOBO fails to function as it is spec'ed; no one should accept any faulty Expensive {MOBO} hardware.
Not accepting failure is a
good policy - but, one must be armed with reasonable expectations, and understanding of 'what you're getting' - face it - the
reason for the existence of that Marvell controller is so the manufacturer of the board can print
SATA3 in BIG LETTERS, on the side of the box - take it for what it's worth (in essence,
nothing!), disable it, saving a PCIe lane, and a couple interrupts and addressing blocks in the process, and put anything that 'matters' on the ICH. Look at it this way: Intel has had half a decade to reiterate and refine both the ICH architecture, and driver set - and that's only if you consider the first 'modern' ICH to be the ICH7R (which was the first 'iteration' to have both architectural and driver support for AHCI);
more that a decade, if you're not that 'fussy'! Once Marvell has, say, three or five years' development under
their belt - it
may become worth something, too - but -
my prediction: by that time, Intel (though, much to my surprise, there's no current 'timeline for this)
will have a PCH/ICH platform that will
do SATA3,
and, they will have an SSD controller chipset that will
also support those kinds of speeds! (There is, however, a hell of a lot of potential in the new SandForce controllers, but, they too, just don't have enough 'field time' in yet to have achieved serious firmware 'tweaking'...