What determines motherboard's ultimate I/O performance?

[Bert Hyman]

External things being equal, is the I/O performance of a motherboard
mostly determined by the chipset, or can the maker impact performance
(intentionally or otherwise) through other circuitry?

If it's entirely or primarily the chipset, which of Intel's current
products appears to be best at SATA I/O?

If it's seriously affected by the maker's implementation, do any of
them consistenly do better than the others?

 

[Paul]

External things being equal, is the I/O performance of a motherboard
mostly determined by the chipset, or can the maker impact performance
(intentionally or otherwise) through other circuitry?

If it's entirely or primarily the chipset, which of Intel's current
products appears to be best at SATA I/O?

If it's seriously affected by the maker's implementation, do any of
them consistenly do better than the others?

Here is a recent article. The charts were prepared in this case,
with what looks like a 4 drive RAID0 array.

http://www.tomshardware.co.uk/2007/...battle_uk/page11.html#raid_0_transferdiagrams

There are other aspects to RAID controllers, and sometimes the issues
don't show up, until something happens to the array. Over on 2cpu.com,
they have a few stories about handling large arrays, and waiting days
for a large array to rebuild. So STR is only one measure of a RAID.
How it performs when the array is busted, is important too. And very few
review articles spend the weeks necessary, to test behavior under
maintenance operations.

Of course, with a RAID0, there is no maintenance. It just breaks
Then you make a new one

Paul

 

[Bert Hyman]

Here is a recent article. The charts were prepared in this case,
with what looks like a 4 drive RAID0 array.

http://www.tomshardware.co.uk/2007/01/03/the_southbridge_battle_uk/p
age11.html#raid_0_transferdiagrams

Thanks; the search terms I was using were returning entirely too much
unrelated info.

 

[ICH9]

External things being equal, is the I/O performance of a motherboard
mostly determined by the chipset, or can the maker impact performance
(intentionally or otherwise) through other circuitry?

If it's entirely or primarily the chipset, which of Intel's current
products appears to be best at SATA I/O?

If it's seriously affected by the maker's implementation, do any of
them consistenly do better than the others?

here's another article at Tom's that addresses the history and new
heavyweight champ for chipsets.

http://www.tomshardware.com/2007/05...ies_chipsets_with_fsb1333_and_ddr3/index.html

What out for that virtualization layer though