I was picturing the overheads of continuous RAID backups. The two
drives would both go thru the MOBO components. To me, intuitively, I
just figgered that double write would take more time than single
write. Could I be wrong?
It doesn't take more time (negligably so anyway, with two
drives it could be that either of them takes a slight bit
longer than the other at any particular operation but it's
not worth considering in most cases).
More significant is the CPU overhead, but often when
accessing a drive the CPU is waiting for files anyway, it
has the spare time for this.
More significant is how it's implemented. If a discrete
chip sitting on the PCI bus, it's inherantly bottlenecked by
that bus, even moreso if other moderate to higher bandwidth
devices are on the same bus. For example, a typical Silicon
Image 3112 chip and a gigabit nic chip used similtaneously
(like copying files over the GbE lan to a suitably fast
recipient) would result in both sharing a little over
100MB/s available throughput.
You don't mention which RAID level though, nor the goal of
this... nor the most demanding uses... so we can't really
estimate if the pros outweigh the cons. In general on a
semi modern board with the southbridge integral RAID, the
performance factor for having the RAID is not a problem,
even of benefit if you were considering RAID0.