Difference between IDE and SCSI ??
Lamar Owen
lowen at pari.edu
Mon Feb 4 16:50:43 UTC 2008
On Saturday 02 February 2008, William Case wrote:
> Hi;
>
> Can someone briefly explain to me the difference between an IDE (ATA)
> and a SCSI device.
You've gotten quite a bit of information here already, but I'll add a few
things.
In a nutshell, IDE/ATA and SCSI are two different approaches to do the same
thing. The 'thing' they do is get the critical timing of the data separator
and clock recovery circuits of the drive read/write data pipe on the drive
itself, and get a lengthy ribbon cable out of the way of the unseparated and
timing/interference sensitive data pipe. ESDI was an intermediate step
between the old ST506/ST412 (the difference between 506 and 412 is kindof
interesting, and something that bit me on my old Xenix box, a Tandy 6000 with
two 72MB ST412 interface drives) and the newer designs of ATA/IDE and SCSI.
The original Integrated Drive Electronics concept was pioneered by the
companies that made 'hard cards' (hard drives on an expansion card). Western
Digital was one of these pioneers, introducing drives in the mid 1980's. The
1986 timeframe seems about right.
The concept was simple: make the drive have all the controller electronics on
board, and just have a small 'paddleboard' card that did some control and
address interfacing to the bus. In particular, the XT variety of IDE (they
did exist, and were used in the Tandy 1000 among other XT semi-clones) was
essentially a WD1002 XT hard disk controller with either an MFM or RLL
ST506/ST412 interface internally. The 40 pins were a slimmed down XT 8 bit
bus with some address decoding done by the interposer card.
The REASON this was done is less obvious, but involved the difficulty in doing
clock recovery/data separation at the end of a ribbon cable for the data
(anybody else remember the dual bus design, with the control being a daisy
chain and the data being point to point drive to controller?). The Enhanced
Small Disk Interface (ESDI) was brought out primarily by Maxtor in the mid
80's to put the data sep/clock recovery function on the drive and off the
controller; this allowed an increase in the data rate up to 20Mb/s (from
MFM's 5Mb/s and RLL(2,7)'s 7.5Mb/s). I had a 690MB ESDI drive in my first
486; it was the fastest thing in town at that time! But ESDI still had
substantial issues; although, most of the high end IDE and SCSI drives of
that day were ESDI internally, on the drive's own controller card.
The AT variety of IDE was the same deal as the XT type, but it's a stripped
down ISA (AT 16 bit) bus instead, with address decode for the I/O port done
in the 'controller' on the motherboard or interposer card. The standard has
been extended a number of times, but the old WD1002 AT controller commands in
PIO mode still work on the latest drives. Yes, even that newest 'no ISA
anywhere' PCI Express motherboard still has vestiges of ISA if it has a
parallel ATA (PATA) 40 pin port, since PATA _IS_ ISA. Beefed up, speeded up,
DMA'd ISA, but ISA nonetheless.
SCSI, on the other hand, was derived in a completely different manner, and
predates the ATA by several years. ATA, after all, being IBM PC/AT derived
has to post date the IBM PC/AT, which was introduced in 1984.
In 1981, the disk drive company Shugart and Associates, a manufacturer of both
floppy and hard drives, needed a system-independent bus (this predates the
IBM PC's dominance; the dominant bus at the time was S-100, with the Apple ][
bus a close second) for its hard drives, and invented the 'Shugart Associates
System Interface' (SASI). Today's Small Computer Systems Interface (SCSI) is
a direct descendant; in fact, SCSI-1 and SASI are almost completely
interchangeable. We don't call it SASI anymore simply because ANSI had
issues with having a companies name as part of the standard. But this bus
does not owe its heritage to the PC/XT or PC/AT system buses as IDE/ATA does.
The convergence between SCSI, which has always had a packetized command
interface, and ATA which was originally just the ISA bus mechanics accessing
the same old AT-bus WD1002 disk controller with its registers, is one of the
more interesting stories. The two standards have borrowed from each other
quite heavily; ATA was 16 bits before wide SCSI at 16 bits was done; SCSI had
the packetized command set that the AT Attachment Packet Interface (ATAPI)
(that made ATA/IDE CD-ROMs more easily possible) emulated, and so on. The
latest cross-pollination is the use of a souped-up Serial ATA (SATA) physical
interface (PHY) for Serial Attached SCSI (SAS).
SCSI has, however, cross-pollinated other technologies; the Gigabit Ethernet
fiber PHY is just a souped-up Fibre Channel PHY, which is essentially Serial
SCSI over fiber, for instance.
IDE/ATA (I use these together, since IDE is such a broad term and compasses
IDE/XT and IDE/MCA as well as the ATA standard; and in all technicalities
SCSI is a type of Integrated Drive Electronics, but not derived from the
system bus like the buses we call IDE) and SCSI are two different ways of
doing hard drive control; they do have similarities in many ways, but they
are different beasts and have different strengths and weaknesses. Again,
with SCSI-originated features like command queuing making there way into the
ATA standards, the line is blurring in areas; but they are still different
interfaces.
Much like Token Ring versus Ethernet.
--
Lamar Owen
www.pari.edu
More information about the fedora-list
mailing list