Media Share 9/MEDIASHARE_09.ISO/private/mpc93mar.zip/IDE.DAT

Text File | 1993-02-14 | 28KB | 488 lines

                               The March of IDEs

     By: Jan L. Fagerholm, The Clubhouse News

     As an enthusiastic purveyor of hardware, it has not escaped my
     attention that in Diskburg, there is a new kid in town. In addition to
     MFM (Modified Frequency Modulation), RLL (Run Length Limited
     encoding), SCSI (Small Computer Serial Interface), and ESDI (Enhanced
     Small Device Interface), we now have IDE, or Integrated Drive
     Electronics, so called for the fact that the brains of the hard disk
     have been moved from the controller card and put on the drive itself.

     "Oh, great," says you. "Just what I need; another blizzard of initials
     to figure out, as if there weren't enough standards for hard disk
     interfaces already. (Aren't standards great, especially when there is
     such a variety to choose from?) If I RLL the MFM, the ESDI gets SCSI
     as it is. What do I need with IDE?"

     "Well," says I, "this time they have actually done us a favor. These
     drives are as fast as the fastest interface on the market. And, just
     in case you don't particularly care about arcane performance
     standards, consider this: they are about half the price of other disk
     drive standards of the same size."

     Despite my enthusiasm for new technologies, I am cautious about
     throwing money at the newest stuff on the shelf. While the engineer in
     me is infatuated with ambitious new concepts, the technician in me
     spoils it all by saying, "Yeah, that's pretty neat, but I'm tired of
     fixing it." I prefer to wait until the first ones blow up and all
     those bright people who invented it figure out why they blow up before
     putting any money on them. So, despite the fact that IDE drives have
     been common for several months no I have resisted acquiring any until
     recently. Having taken the plunge, though, I thought that my
     experience in getting them up and running may be of value if you are
     considering doing the same.

                     The Academy of Awkward Acronyms

     But first, some remedies for the pain of Buzzword Blues. Take one for
     each buzzword swallowed.

     MFM (Modified Frequency Modulation) ; A method for writing data pulses
     on disks. The distinction for our purposes is that this standard
     applies to the drive itself, rather than the controller.

     RLL (Run Length Limited encoding) ; A newer method for writing data
     pulses on disks. The data is encoded by the controller with a scheme
     that results in about 50% fewer pulses needed to represent that data.
     The result is that you get 50% more data on a given disk, as well as
     read and write it 50% faster. The distinction here is the same as for
     MFM: the standard applies to the drive itself, though the actual
     encoding is done by the controller card.

     ST-506 ; The original Seagate specifications for MFM interface of
     small hard disks to computers. It is important to realize that ST-506
     specified two standards: one to read/write to the disk's surface, and
     another to pass the data to/from the computer. Thus, the
     disk/controller card pair that we find for present PC hard disks.

     SCSI (Small Computer Serial Interface) ; General interface standard
     developed to make peripherals talk to small computers. The standard
     encompasses nearly all peripherals, not just disk drives. SCSI stands
     between the computer and the devices that the computer talks to,
     translating and coordinating between them. SCSI peripherals require
     that the computer have a SCSI adapter (controller) built in or plugged
     in, in order to hook up the SCSI peripherals. Advantages are speed and
     compatibility. The disadvanta is cost, mostly due to the extra silicon
     required to make SCSI smart enough to talk to many different types of
     devices.

     ESDI (Enhanced Small Device Interface) ; A subset of SCSI, optimized
     for magnetic storage devices (i.e., disk drives and the like). ESDI
     doesn't handle the range of peripherals that SCSI does, but it has
     special talents for very large disk drives. ESDI drives work only with
     ESDI controllers because the drives themselves have different
     electronics from ST-506 or SCSI drives. The advantage is speed; the
     disadvantage is cost, for the same reason as SCSI.

     IDE (Integrated Drive Electronics) ; Read the rest of this article,
     silly.

                         Whose Idee Was This, Anyway?

     The origins of the IDE specification go back several years to a
     concept created by Compaq for its (long discontinued) Portable II, as
     a method to save a slot. Compaq created the specs, and Western Digital
     built it. The actual term "Integrated Drive Electronics" first
     appeared in 1986 when Compaq, Western Digital, and Control Data Corp.
     (now part of Seagate) jammed a WD controller onto a CDC 40M
     half-height drive. This drive was first used in the Compaq Deskpro
     386, and was in large measure responsible for initiating Compaq's
     reputation for speedy peripherals.

     While this was going on, Compaq was also engaged in a joint project
     with Conner Peripherals to develop a silicon gate array that emulated
     the new IDE, and that would actually be part of the disk drive's
     electronics (and enjoy an accompanying speed increase). This was the
     first real IDE. The system connector was not through the AT bus slots,
     but through a connector on the motherboard (of the Compaq Portable
     III), as are all true IDE implementations now.

     Lots of other people thought that all this was a good idea and began
     implementing the idea themselves. Of course, they all experienced NIH
     (Not Invented Here) Syndrome and added all sorts of "improvements" to
     make it their own. Consequently, many early IDE drives suffer
     compatibility problems when introduced to current IDE drives.

     Fortunately, a group of drive manufacturers decided to save a good
     idea before everybody re-invented it, and submitted a proposed ANSI
     standard for IDE in November 1990. By the time you see this, it will
     probably be a real ANSI standard. Ever since the proposal, virtually
     everybody who makes IDE drives has been conforming to the draft IDE
     standard.

                      Buzzwords, Our Best Byproduct

     It is useful to realize that the term "IDE" is generic, not specific
     to a certain type of PC hard disk. Strictly speaking, SCSI drives and
     ESDI drives are IDE drives. It just happens that SCSI and ESDI contain
     specifications for certain other devices in addition to disk drives.
     SCSI is probably the best illustration, as the drive of choice (the
     only choice) for Apple Macs. (The only way you can attach anything to
     the original Mac is through a SCSI interface, because Apple wouldn't
     tell anybody anything about its internal bus.)

     The point here is that "IDE" is in reality an umbrella term for any
     hard drive with electronics smart enough to communicate with a host
     (your computer) without the host knowing anything about how the drive
     works. So, to pick picayune PC nits, the "IDE" hard drives that most
     IBM-compatible PC users are concerned with is properly known as the ATA
     (AT Attachment) interface. (It's probably not an "ISA Attachment"
     because "ISA" is not an ANSI term.) Indeed, the ANSI specification for
     IDE has standards for the bus and Microchannel, right alongside the AT
     standard.

           How Come IBM Doesn't Call AT "Advanced Technology" Anymore?

     Why would seemingly intelligent drive manufacturers make an interface
     dedicated to a bus that everyone says is obsolete? Well, as ancient as
     the AT bus is (as technology goes), it never has been very well tapped
     for its real speed capabilities. Most of the speed limits have been
     imposed by things like DOS and ST-506, so we have never seen even half
     the speed of which the AT bus is actually capable.

     Hard drives are the fastest devices plugged into the AT bus, yet the
     fastest of these do less than half the speed that an 8-MHz AT bus is
     capable of, let alone the 12-MHz bus clocks now common on PC clones,
     or even the EISA bus. Even the current standards for SCSI-2 and ESDI
     deliver ideal rates of only 10 MHz. (I say "only" because these rates
     are for serial transfer at the read/write heads, and do not account
     for the serial-to-parallel and all other conversion overhead before
     the data is given to your PC bus.)

     Without advancing a discussion of bus wars any further here, I'll just
     state that no one has come anywhere near the speed limit of the AT
     bus. Considering that the AT bus will continue to exceed any
     reasonable requirements for single-user PCs for the foreseeable
     future, it makes perfect sense to give in and let the hard drive
     electronics talk directly to the AT bus.

     I'm going to pick on SCSI for a minute to illustrate the major
     advantage of the IDE concept. The down side of SCSI is that there are
     at least two interpretations of data when going through this
     interface; one to translate the stuff from the host's format to
     SCSI format, and another on the other end to translate from SCSI
     format to the peripheral's format. Of course, if the device in
     question is a hard drive or anything else with two-way communication,
     this must be done going each way.

     This is like two people who speak different languages talking to each
     other through an interpreter: "C" talks to the interpreter "S," who
     turns around and translates to "D." Communication is streamlined
     considerably if "D" (disk drive) learns the language of "C" (computer)
     and disemploys "S." Faster, too.

     Any electron pusher already knows that you can make anything work
     faster by making it communicate directly with whatever it talks to,
     without passing through an intermediary. And any bean counter knows
     that you probably have to sell more than three of these things to make
     any money, so you don't make a device-specific interface unless there
     is a large market for it.

     There are about a gazillion AT-bus machines out there now, so let's
     just make the hard disk speak one language, AT Bus, and cut out all
     the middlepersons (SCSI, ESDI, etc.) All we have to do to take better
     advantage of the performance possible from the latest technological
     refinements is to put in enough intelligence to fool MS-DOS, which
     isn't much because MS-DOS isn't very bright.

                              Improved Products

     Hard-drive hardware has improved lots in recent times, primarily due
     to better media, increasing the number of bits that can be recorded on
     a given surface area. Performance improvements have come from
     technological improvements in corresponding areas, namely:

     * Finer grained media allow the heads to fly closer to the media.
     Thus, the heads can read and write smaller pulses (bits), which also
     allow the bits to be stored closer to each other. More bits per inch
     mean that you can read and write more data for each revolution of the
     platter.

     * Improved coercivity. This is a physical characteristic of the stuff
     on the platters where the magnetic pulses are stored. Coercivity is a
     value used to describe the ability of the media to store magnetic bits
     close to each other without them demagnetizing each other. Again,
     denser bits mean more data per revolution.

     * Improved head-tracking techniques, allowing cylinders to be closer
     to each other on the platter. This allows more cylinders on a given
     size platter and reduces stepping times between the cylinders.

     * Smaller, lighter read/write heads. These can be stepped from track
     to track faster than heavier heads, thus improving average access
     times by not waiting as long for the heads to move between cylinders.
     To get a handle on this, realize that each millisecond of step time
     costs you about 1 kilobyte (real world) of data transfer while the
     head is stepping. Thus, there is a large difference in average
     transfer rates between drives with 65 msec average access vs. those
     with 28 msec average access times , or between 28 msec and 16 msec.

     * General improvements in IC and VLSI design and fabrication. In
     English, this means that the newer silicon circuits just plain run
     faster on less power.

     IDE drives have a variety of creative enhancements in addition to
     these straightforward physical improvements. Among them are:

     * Higher spindle speeds. Where 3600 rpm was commonly imposed because
     of the ST-506 standard, IDE drives turn at 4500 rpm and higher. (The
     faster the platters turn, the faster the transfer rate.)

     * Translation. The ability to take cylinder-sector-head counts from
     DOS and rearrange them to the physical reality of the drive removes
     the physical limitations previously imposed by the maximum figures
     built into DOS.

     * Increased cylinder counts, well beyond the 1024-cylinder limit of
     DOS.

     * Unusual sector counts, and different sector counts on different
     parts of the platters (Zone Bit Recording). Because the outer
     cylinders of the platter are longer than the inner cylinders, you can
     pack more bits on the outer cylinders. (On a 3" diameter platter, the
     outer cylinder is 9.4" long, while the cylinder at the 2" diameter is
     only 6.3" long. The smaller the platter, the more acute this problem
     becomes.)

     * Advanced encoding techniques, usually RLL 1,7 or RLL 2,7, are
     employed in the drive-level electronics, although this is not stated
     as a "feature" of IDE drives.

                              Will IDE Work for You?

     Life with DOS machines has grown complex, and; contrary to
     expectations; IDE drives do nothing to simplify things.
     Unfortunately, you (usually) cannot simply purchase an IDE drive, plug
     it in, and expect it to work. There are several areas of compatibility
     that need to be assured first. Otherwise you may find that your
     inexpensive hard drive is relegated to the role of costly door stop.
     (They are too small to be good boat anchors.)

     The first thing to consider is the brand of machine that you have.
     There are some big-name machines that have "enhanced" AT busses that
     IDE will choke on. AT&T, AST, Tandy, and, yes, some Compaqs, to name a
     few. IBM PS/2s have their own version of IDE for the Microchannel bus;
     make sure you don't wind up with an ATA drive for these machines.

     In general, the more generic your clone, the better chance you have of
     an IDE drive working properly. If you have one of those 286 or 386
     motherboards that you can't even identify, it will probably work
     painlessly with an IDE ATA.

     If you have a newer motherboard (made after 11/90, not to be confused
     with the BIOS date), check to see if it has a 40-pin ATA connector on
     it somewhere. If so, this is the method of choice for plugging in ATA
     drives. Even if you do have this, you are going to need a plug-in card
     to run the floppy disks. If you don't have the motherboard connector,
     you will need an ATA adapter card to plug into the AT bus. In addition
     to accessing the necessary data and control lines from the AT bus,
     these cards provide t necessary buffering between the bus and whatever
     is plugged into it. Most of these adapter cards now have the necessary
     electronics to run the floppy disks, also.

     Hardware overcome, the next thing is firmware. IDE does not work with
     many older versions of BIOSs. ATA drives are notoriously balky because
     the older BIOSs don't always have a hard drive configuration in their
     setup that ATA will translate properly. In some cases, the low-level
     hard disk routines are just plain incompatible. Generally, if your
     BIOS provides for a custom configuration of hard disk parameters
     (cylinder-sector-head count), it will probably work with IDE. Most IDE
     drives are provided with a suggested configuration to use with your
     BlOS's setup routine.

     You need to determine the manufacturer, revision number, and date of
     your computer's BIOS. These are usually displayed on the screen by
     most BIOSs at boot time. If you have one of those motherboards that
     runs in Warp Numbers instead of measly megahertz and the BIOS date is
     only on the screen for a few nanoseconds, you will need to use DEBUG
     or some other memory snooper to check the contents of address FFFF0
     (or FFFF:0000 in Intel's segmented parlance), which is where most
     BIOSs put their revision date. (Alternatively, boot from a floppy disk
     that has no CONFIG.SYS or AUTOEXEC.BAT files. Without all that stuff
     scrolling up the screen, the BIOS message may remain on screen after
     booting.)

     The following table, while not exhaustive, will cover about 75% of the
     machines out there. Your BIOS should be the same as or newer than
     version shown in the table.

                         BIOS Mfr.        Version

                        AMI              04/09/90 or later
                        Award            3.04 or higher
                        Phoenix 286      3.10 or higher
                        Phoenix 386      1.10 or higher

     If you have an older BIOS, upgrade paths are various. For Phoenix,
     just go to Fry's Electronics or someplace similar and buy it off the
     shelf. For AMI, you will need to deal with them directly about
     upgrades, or possibly with the vendor where you got your computer. For
     manufacturer-specific BIOSs, you can find out only by asking them,
     although a good clue is whether the BIOS was issued after the IDE
     draft ANSI standard came out (November 1990).

     If you are a victim of an older BIOS, though, you should not
     necessarily be discouraged from getting an ATA drive. The cost of a
     new BIOS added to the cost of the ATA drive will still be far less
     than a similar sized drive of any other standard.

     With your sparkling new ATA drive in your eager little hands, you will
     now need to educate your machine about it. If your BIOS is one of the
     recent AMI BIOSs, life is idyllic. Simply take the numbers that were
     provided with the drive for cylinder-sector-head counts and plug them
     into the Custom Drive part of the BIOS setup routine, and specify that
     drive as the drive type.

     If your BIOS does not have a custom drive type available, you will
     need to peruse your BIOS's drive table to find a drive type nearest in
     size to your new ATA drive and select that as your drive type. You must
     be careful here to select a BIOS drive type that is equal to or smaller
     in size than the physical drive. Otherwise, the BIOS will attempt to
     write data to nonexistent places, which may cause you to invent
     imaginary words about imaginary places for your now imaginary data. In
     all cases, do not exceed cylinder count of 1024, because DOS will
     choke. If the real number of sectors is known on the IDE drive (i.e.,
     translation is not used), do not specify a number of sectors that is
     more than the drive actually has.

                          DEBUG Thinks IDE Means "Hide"

     ATA drives do not contain low-level format routines in their beady
     little minds, so the familiar DEBUG command "G=C800:5" (and similar
     others) will leave your machine staring blankly back at you as if to
     ask what you are trying to do. ATA drives come low-level formatted
     from the factory. This is not merely a courtesy, as you will find that
     most ATA drives may not be low-level formatted by you. ATA drives that
     generally may not be low-level formatted are those that have built-in
     RAM buffers between you and he disk surface. (You will need the
     technical data on the drive to know this.)

     In all cases, ATA drives require external software and/or external
     hardware to accomplish a low-level format. Personally, I regard this
     as a major inconvenience because low-level formatting is hardly
     permanent. Most of the problems that I encounter with hard drives more
     than two years old are due to lost sectors from deteriorating data
     address markers. These drives need low-level formatting to refresh the
     address markers.)

     While some of the smaller ATA drives (those that do not employ
     translation) may actually be formatted with external software, all
     drive makers warn against doing so. ATA drives all have their bad
     sector maps written to the disk instead of on that piece of paper that
     always gets lost before you get the drive. Low-level formatting will
     destroy the bad sector table, so DOS will not lock out these sectors
     when you do the DOS format on the drive, and you may get data loss
     from trying to use bad or marginal sectors. Personally, I prefer to
     perform my own surface-analysis tests before DOS formatting the drive
     anyway. Hopefully, Gibson and Norton and Mace will include routines to
     tinker with the low-level format on these drives before this gets to
     be a problem.

     While low-level formatting is supposedly a thing of the past with ATA
     drives, you are still faced with the onerous task of touching FDISK
     long enough to partition the drive, then proceeding with a DOS Format
     so that DOS knows how to find things on the drive. There is an upside
     to this confusion, though: As there is no hard disk BIOS, you gain the
     space in the memory map where it usually goes. Users of DOS 5.0's
     EMM386, Quarterdeck's QEMM, or similar memory managers will find that
     the addresses normally occupied by the hard disk ROMs are now
     available for remapping. Depending on the type of controller card
     being removed, this will free up an additional 16;32K, which may be
     used for TSRs, device drivers, and the like.

                              The Envelope, Please...

     Though my exposure to ATA drives is not exhaustive, I have seen enough
     different ones to form some likes and dislikes.

                                    Picks

     * Seagate 1102A (85M) and 1144A (125M). These drives both have a 3.5"
     format, the best prices in their class, and the fastest transfer rates
     in their class regardless of price. They boast real-world transfer
     rates of 975K bytes per second (as fast as the fastest SCSI drives).
     Benchmark testing returns average access times of 18.5 msec<197>better
     than their rated 19 msec. They are quiet, out-silenced only by the
     high-quality Conner Peripherals drives. They have garnered several
     editorial awards, and are commonly available for the lowest street
     prices in their class. (Seagate is second only to IBM in volume sales
     of disk drives.)

     * Most of the Conner Peripherals IDE drives. They are quality, but you
     will pay a premium price to go with them. ("Premium price" for IDE
     drives is still cheaper than other standards.) They are so quiet that
     you will need to watch the drive access light to see if they are
     working. They will make you doubly irritated with the noise emitted by
     your PC cooling fan.

                                     Nits

     * The entry-level Western Digital WD280 (80M). While one of the lowest
     priced drives on the market, I think that it is unacceptable to make a
     hard drive in this day without automatic head parking. These guys must
     be the last to realize that we do not run our PCs like mainframes, but
     that we actually turn them off when we are through with them. It also
     has the slowest average access times of any of the 80M drives (23
     msec).

     * The Conner Peripherals 3204 (210M). (Nobody's perfect.) This is the
     slowest ATA drive of its size. While its average access figures look
     good, its DOS file transfer rates lag. (The real world performance
     doesn't live up to the benchmark figures.) Though speedy compared to
     other drive standards this size, the 3204 still sells at a premium
     price while turning in DOS performance that averages 30% slower than
     other IDE drives of the same capacity.

                      De Termination of Dis Diskussion

     You do not need to be considering a move to GUIs to consider upgrading
     your hard disk these days. If you have a common 40M MFM hard drive,
     you can double your disk speed with a 40M ATA drive for about $150. If
     you are running a 40M drive of any type with current character-based
     software and any amount of your own data files, you are probably
     feeling cramped for space. You can double your capacity and double
     your disk speed with an 80M ATA drive for what you would have paid for
     a second 40M MFM drive a year ago. If you wake up your computer some
     day and it greets you with "Good Morning ; 1701 error ; Would you like
     to play Hangman with a real rope?" you can double your hard disk
     capacity and double your disk speed for the same price that it will
     cost you to fix Old Plod.

     If you run Windows with any of its major applications, getting a
     larger hard drive is a virtual necessity. The 6M of disk space that
     Windows itself takes is relatively trivial, as you find out when you
     install Ami Pro 2.0 (8M) or Excel 3.0 (8M) or Word for Windows 2.0 (a
     whopping 15M for the full version). Only three years ago, you could
     run a major word processor, spreadsheet and database on a 32M
     partition and have 20M left over. Now you need a full 32M partition
     just to run Windows, a word processor, or a spreadsheet, with a couple
     of measly megabytes left over for data. This is the price you pay for
     GUIs and graphical software. And, of course, it is going to get worse
     before it ever begins to look better. (I see the light at the end of
     the tunnel, but it looks a lot like a train.)

     Now is a particularly good time to consider an upgrade to your hard
     disk. There is a price/performance jump here that hasn't been seen
     since turbo clones entered the market against true blue IBM machines.
     A year ago, a 100M hard disk with controller card cost about $600, you
     can now get that same 100M on a smaller, quieter, faster IDE hard
     drive for half that price.

     The current state of the economy is only a small part of the reason
     why these drives are so cheap. IDE is both faster and much cheaper to
     make than other standards. It is virtually certain that your next
     computer will come with an IDE drive, but why wait? Putting an IDE
     drive in your present bit-twiddler may spiff up its performance enough
     to beat back 486 envy a while longer. At least you may no longer feel
     the need to find your PC's drain plug and change the electrons because
     the old ones seem kind of sluggish.

  ************************************************************************

  1992, Jan L. Fagerholm. Reprinted with minor edits from The Clubhouse
  News (PC Clubhouse, Hayward, CA), April 1992. Mr. Fagerholm can be
  contacted via the PC Clubhouse BBS (510/581-8529)