[conspire] How to check if a mobo is compatible with

Rick Moen rick at linuxmafia.com
Sun Mar 30 15:35:15 PDT 2008

Quoting Mike Higashi (mhigashi at gmail.com):

> This isn't an area that I've throughly investigated, but I suspect
> that the situation is somewhat better than you believe, largely due to
> the universal adaptation of RAM-based buffers on board hard drives.

Yes, but that's just for a limited amount of cached data at any given
time, when you happen to have a cache hit:  I note that 8MB of
disk-based cache appears to still be typical, which really isn't very
much in an era when 100GB drives are considered tiny.  As you say,
disk-local cache improves sequential reads or writes to a certain extent
-- but pretty soon you're out of cache hits and are back to physical
disk access.

Also, you're actually a lot more likely (in *ix) to hit the OS's _own_
disk cache, because those typically have a great deal more RAM to play
with.  (That system-level cache is really a great deal more useful for
performance and for limitation of disk wear.  I frankly wish HD
manufacturers would just skip the HD cache and keep their prices lower,
which would let me apply the same money to a little more system RAM:
Hard drives are a suboptimal place for disk caching.)

(As an aside, it's not unknown for manufacturers to gimmick their speed
testing scenarios to hit disk-based cache more than is statistically
likely in the real world.  This allows them to claim unrealistic speed
results.  A cynic would suggest this is one of the reasons for
disk-local cache RAM being present in the first place.)

More information about the conspire mailing list