pls confirm or debunk my theory on ide/sata deployment

hi all

i have a 32-bit motherboard that supports two ide channels and two sata
ports with raid functions. i never use raid. i have two disks, one for
systems and one for data. the systems disk boots up running the raid port.
the data disk uses one ide channel.

the data disk needs to be upgraded in the near future. i know sata2 disks
run faster than ide disks. if i change the data disk from ide to sata (my
mobo can only run sata1), i fear that putting two sata disks would
overwhelm the sata controller while leaving the ide controllers with little
to do, except to look after the dvd drives and the occasional backup.

my theory is: a slow ide drive on a slack ide controller should run faster
than a fast sata drive on a busy sata controller. this is according to
common sense. whether it's technically correct, i wouldn't have a clue.

thanks much

sharonpst wrote:
> hi all
>
> i have a 32-bit motherboard that supports two ide channels and two sata
> ports with raid functions. i never use raid. i have two disks, one for
> systems and one for data. the systems disk boots up running the raid port.
> the data disk uses one ide channel.
>
> the data disk needs to be upgraded in the near future. i know sata2 disks
> run faster than ide disks. if i change the data disk from ide to sata (my
> mobo can only run sata1), i fear that putting two sata disks would
> overwhelm the sata controller while leaving the ide controllers with little
> to do, except to look after the dvd drives and the occasional backup.
>
> my theory is: a slow ide drive on a slack ide controller should run faster
> than a fast sata drive on a busy sata controller. this is according to
> common sense. whether it's technically correct, i wouldn't have a clue.
>
> thanks much

They test a single drive, and then a pair of SATA drives in RAID0,
on different chipsets. The two disks get about double the bandwidth
of one disk, which says that two SATA ports can operate simultaneously
without interfering with one another.

http://www.tbreak.com/reviews/articl...1&pagenumber=4

So without getting into specifics involving bugs in hardware or
driver design, I don't see a reason to think that way.

Here, they test the Gigabyte iRAM storage device, a SATA disk which
uses DRAM chips to store the data. It has virtually zero seek time,
but is still limited by the SATA 150MB/sec cabling. For the Southbridge
used, they didn't have a problem with two of them operating
on the SAYA ports at the same time.

http://www.tomshardware.com/reviews/...te,1111-9.html

Paul

On 2008-08-21, sharonpst <sharonpst@dontspamme.com> wrote:
>
> i have a 32-bit motherboard that supports two ide channels and two sata
> ports with raid functions. i never use raid. i have two disks, one for
> systems and one for data. the systems disk boots up running the raid port.
> the data disk uses one ide channel.

Knowing what mobo would be useful here so we we'd know what disk
controller you have and whether it is a PCIe type.

> the data disk needs to be upgraded in the near future. i know sata2 disks
> run faster than ide disks. if i change the data disk from ide to sata (my
> mobo can only run sata1), i fear that putting two sata disks would
> overwhelm the sata controller while leaving the ide controllers with little
> to do, except to look after the dvd drives and the occasional backup.
>
> my theory is: a slow ide drive on a slack ide controller should run faster
> than a fast sata drive on a busy sata controller. this is according to
> common sense. whether it's technically correct, i wouldn't have a clue.

In practice the speed advantages of SATA are yet to be realised,
since the transfer rates of hard drives are slower than either
connection.

What you suggest basically doesn't hold. SATA has a separate
channel for each device - bandwidth isn't shared between the two
devices. There is a potential pinch point in the disk controller
and its interface to the machine, though. This is where details
help. However, on a typical desktop machine even if there is
contention between the two devices (and that's a big if) there will
usually be very little real world performance hit since it is
comparatively rare for both disks to be accessed simultaneously.
Typically such activity would be associated with heavy load or a
disk-disk copy.

If you are concerned about performance under load I'd suggest going
for SATA II drives with NCQ - you'll have to read the technical
specs for that because it is hidden away and isn't universal among
SATA II devices. You won't get any perfomance benefit with the
existing controller but if you later upgrade that, either via an
expansion card or new mobo, they come into their own. In terms of
benchmark-style throughput figures it makes no difference but
traditional ATA devices collapse into a heap the minute they are
placed under load. NCQ devices keep on going - pretty much what
SCSI has been doing for years but at much lower cost.

--
Andrew Smallshaw
andrews@sdf.lonestar.org

"sharonpst" <sharonpst@dontspamme.com> wrote...
>
> i have a 32-bit motherboard that supports two ide channels and two sata
> ports with raid functions. i never use raid. i have two disks, one for
> systems and one for data. the systems disk boots up running the raid port.
> the data disk uses one ide channel.
>
> the data disk needs to be upgraded in the near future. i know sata2 disks
> run faster than ide disks. if i change the data disk from ide to sata (my
> mobo can only run sata1), i fear that putting two sata disks would
> overwhelm the sata controller while leaving the ide controllers with little
> to do, except to look after the dvd drives and the occasional backup.
>
> my theory is: a slow ide drive on a slack ide controller should run faster
> than a fast sata drive on a busy sata controller. this is according to
> common sense. whether it's technically correct, i wouldn't have a clue.

AFAIK, SATA controllers devote full bandwidth to each SATA port.

However, I doubt you will run into that problem anyhow, because your 2 HDs will
not saturate even a single SATA port. Any 7200 RPM HD will be limited by the
physical data density on the HD and the rotational speed. You will be unlikely
to get over 80 MBps from a single HD even at peak throughput (except possibly
when the cache is filling), and only when copying data directly from one HD to
the other would you get sustained simultaneous demand from both of them. Even
then, one would be only reading while the other writes.

The new perpendicular media HDs are faster than the old ones, but they still
don't make up for raw spindle speed (e.g., WD Raptor and VelociRaptor).