[Pc_Support] External SATA bracket ($9), external SATA enclosure/power ($35)

Bryan J. Smith b.j.smith at ieee.org
Wed Jan 25 18:23:13 EST 2006


Jason Boxman <jasonb at edseek.com> wrote:
> I'd heard ATA isn't as bad as it once was for slaving.
> http://slug.archives.nks.net/List/slug.archive.0601/0196.html
> http://slug.archives.nks.net/List/slug.archive.0601/0203.html

Umm, _no_.  You have to look at the _context_ of that discussion.

What the person was saying is that putting 2 ATA (hard drive) devices
on the same ATA channel is _better_ than an ATA (hard drive) and
ATAPI (optical drive).  Most ATAPI devices are only UltraDMA mode 2
(8MHz DDR @ 16-bit = 33MBps), a few UltraDMA mode 4 (16MHz DDR @
16-bit = 66MBps) whereas most ATA devices are UltraDMA mode 5 (?MHz
DDR/QDR? = 100MBps) or mode 6 (?MHz DDR/QDR? = 133MBps).  You either
have to "slow down" the channel, or do constant bus resets to handle
both signals.

Furthermore, only *1* device is "on the channel" at a time -- be it
the master or slave.  When you do device-to-device transfers and they
are on the same channel, you copy to memory, change control, then
back from memory.  ATA is _not_ SCSI, devices do _not_ talk directly.

It's _much_better_ if you have the devices on their own channels, so
they simultaneously write to memory while the other reads from it. 
It really _does_ make a _lot_ of difference.  Let alone you'll
_never_ have inter-vendor/signal issues if the devices differ.

The master/slave setup is from the PIO days of IDE, _not_ the DMA
days of ATA.  Even the original IDE in the ESDI specification _never_
allowed for master/slave, it was introduced as a vendor hack by
Western Digital "EIDE" and other vendor equivalents.  One device per
channel -- the master/slave setup is only designed to work for legacy
PIO and can cause issues with DMA.

Since SATA _only_ implements DMA ATA, the master/slave setup isn't
supported at all.  They also call it a "port" instead of "channel" or
"bus."

> I haven't tried it myself.  I still prefer one ATA disk per
> bus to ensure a disk poping doesn't effect the other disk on
> the bus.

It's actually _not_ a "bus" in the traditional sense -- it's an
end-device to the _other_ bus arbitrator (e.g., PCI).  That's why the
term "channel" is used -- there's _no_way_ to actually "target" the
master or slave (long story).  There is _no_ "addressing" on a "bus"
in ESDI or ATA.

Only one device can be on the channel at a time.  The master/slave
setup is a PIO select to enable one drive or another.



-- 
Bryan J. Smith     Professional, Technical Annoyance
b.j.smith at ieee.org      http://thebs413.blogspot.com
----------------------------------------------------
*** Speed doesn't kill, difference in speed does ***



More information about the Pc_support mailing list