View Single Post
  #3  
Old May 12th 04, 01:23 AM
R. Marshall Chew
external usenet poster
 
Posts: n/a
Default

In article ,
says...
In article , R. Marshall
Chew wrote:

Installed two drives on Windows XP Pro PC (P4C800 Deluxe) using the
onboard Intel ICH5 SATA controller (not the Promise controller). Didn't
have to do the "F6" routine on install as XP recognized the SATA drives.
Here is my question: In device manager, C (WD740 raptor boot drive) and
D (Seagate 160) both show as 3rd and 4th master and there are two
instances of "eide controllers" (DVD and CDRoms are on the regular ide
controllers 1 and 2). In the BIOS the drives show as UDMA-6 (on boot-
up), but windows shows them as UDMA-5. I am not running RAID, just 2
SATA drives. Are the drives running at the full 150 speed or are they
being limited to regular Ultra ATA drives because they are connected to
the ICH5 controller? I get a different answer from everyone I ask.
Should I move the drives to the Fastrack controller to "insure" I'm
getting the right performance or leave them as is? Again, no RAID.
Thanks for anyone who can give me the answer.


If you download the ICH5R datasheet, for the other I/O devices on
the Southbridge, it says:

"Note that most transactions targeted to the ICH5 first appear
on the external PCI bus before being claimed back by the ICH5.
The exceptions are I/O cycles involving USB, IDE, SATA, and
AC97. These transactions complete over the hub interface
without appearing on the external PCI bus."

I interpret this to mean, that storage devices plugged into the
SATA or IDE ports, are not restricted by PCI bandwidth concerns.
There are still bandwidth issues however - the IDE ports have
a max write speed of 88.9MB/sec and max read of 100MB/sec
(i.e. ATA100). The write speed is limited by a 22.5ns write
strobe chosen for the implementation (=44.4MHz x 16 bits).

The SATA drives, the IDE drives, and the PCI bus all share a
266MB/sec hub interface from the Northbridge. Two SATA drives
bursting from controller cache can saturate the bus by themselves.
On sustained bursts from large files, the head data rate of the
disk will be the limiting factor. With Raptors, maybe this would
be about 70MB/sec or so on each drive (I didn't verify this
number on storagereview.com).

As with any PCI bus, all devices on the PCI bus are limited to
an aggregate of about 100MB/sec practical bus bandwidth. (You
can turn up the burst length on a card and improve this a
bit, at the expense of fairness problems with other PCI cards,
like PCI sound cards.)

Total bandwidth 266MB/sec

Thanks Paul! Guess I'll just leave it as is for now.

Rob C.