Dec 132017
 
Good discussion by two authors on current hard drive products.
File DRIVEBUY.ZIP from The Programmer’s Corner in
Category HD Utilities
Good discussion by two authors on current hard drive products.
File Name File Size Zip Size Zip Type
DRIVEBUY.TXT 64253 22668 deflated
TPCREAD.ME 199 165 deflated

Download File DRIVEBUY.ZIP Here

Contents of the DRIVEBUY.TXT file


Buying a Hard Disk

written for the CPCUG build 1/1/91

Selecting a hard disk is an important question in buying a new
computer and a continuing question in upgrading existing IBM-PC
compatible computers. This document provides background data so
that users can assess tradeoffs between price, performance, and
future upgrade costs, and make better decisions.
This document consists of two parts. The first part was
written by Mike Focke and covers general principles underlying hard
disk function, describes some guidelines which may not apply to a
few PC compatible computers but do apply to the vast majority, and
contains some specific recommendations. The second part of this
document was written by Doug Wagner, and contains further
explanation of some of the tradeoffs and options, along with tables
describing speed, size, and cost of a larger number of currently
widely available hard disks.
It is our hope that this description will be useful to both
beginners and to advanced users.


Buying a Hard Disk

ST506, MFM, RLL, ERRL, ARRL, ESDI, SCSI, and AT/IDE

What do they mean? Which should I buy?

written for the CPCUG build 1/1/91 by Mike Focke

Lets start by defining what are the important questions to ask when
you go to buy a hard disk subsystem.

1. How much data can it store?

2. How quickly can it move the mechanical components and begin
transferring data?

3. How fast can it transfer data once it has begun?

4. How reliable will it be?

5. How much does it cost?

6. What comes with it?

7. What if it breaks?

If you have the answers to these questions, you can make an
informed choice based on your projected needs.

So this paper will discuss items you should know as background when
asking these questions and will make some recommendations of
several subsystems known to work well together and to be easily
installed.

For the purposes of this discussion, I am going to assume that you
are buying a new hard disk subsystem to go in the new 386SX or DX
machine you are going to build.

If you are going to use parts from another machine, then your most
important issues are simple things like do you have a slot that
will fit the controller and will the drive mount in your case and
are the cables long enough. I would suggest you read on ahead
because you may find out things that persuade you to buy a new
subsystem and sell your old one. Over the last three years, amazing
things have been happening in hard disks technology and the prices
have been dropping while the quality has improved.

How much data the drive will store is something you can read in the
advertisement. Make sure the size of the drive is advertised in
formatted bytes. Sometimes raw drives are listed with their
unformatted sizes. The formatting uses about 10% of the drive's
capacity.

Future graphics environments will use lots of disk space to store
the larger programs that will be written to use the larger memories
we now have. Windows and its files and fonts takes more than 10
megabytes on my machine. I think a minimum of 40 megabytes will be
needed on a 386 machine and you may find you need to add still more
storage space once you start using more and bigger programs.

PRINCIPLE # 1: You can never have too big a drive.

How quickly the drive's heads can be moved is something you can
read in the advertisement. Average access time or average seek
time is what is usually quoted. A slow drive is 85 milliseconds,
an average drive is 28 milliseconds and a fast drive is in the mid
to low teens. Buy the quickest drive you can afford. You will
need a fast drive to move all the data that future environments
will require.

PRINCIPLE # 2: You can never have too quick a drive.

MFM, RLL, ARRL and ERRL are ways of storing data onto a hard disk's
surface. MFM usually results in a drive that has 17 512-byte
sectors per track. Since most drives spin at 3600 revolutions per
minute, the maximum theoretical transfer rate that you could ever
hope to get is 17 sectors times 512 bytes per sector divided by the
time it takes to spin one revolution (1/60th of a second). The RLL
variants allow more sectors per track to be stored and thus
increase theoretical throughput (typically 27 but up to 83 sectors
per track).

Maybe I should note here that ESDI and IDE drives often use RLL as
a way of encoding data. So you can buy a ESDI RLL drive with
"ESDI" defining how the drive and controller talk to each other and
"RLL" defining how the data pulses on the disk will be interpreted.
But, all things being equal, the more sectors per track, the faster
the data can be transferred.

PRINCIPLE # 3: More Sectors Per Track Good

Likewise, the more heads that there are, the more data that can be
transferred before you have to mechanically move those heads to the
next track.

PRINCIPLE # 4: More Heads Good

But what if the controller isn't fast enough to keep up with data
moving under its heads at that rate? Then interleaves are
inserted. Interleaving at 2 to 1 will reduce your throughput by
50%. 3 to 1 by 66%. Since 1 to 1 interleave controllers are
available for very little more than slower controllers, there is no
reason today to buy anything but a controller capable of a 1 to 1
interleave.

Even if you are going to use your old ST506 MFM or RLL drive in the
new machine, consider getting a 1 to 1 controller. Without the 1 to
1, you will have a machine that seems slow just because it is
always waiting for data.

GUIDELINE # 1: Interleave at 1 to 1 only

Once the data is read, the controller must transfer data to the
computer's bus before it can get the data into memory. The wider
data path it can use (the more bits it can send at one time) the
faster it will be. There are some 8 bit SCSI interface cards and
many 8 bit MFM controller cards but they all are slow. Get nothing
but a 16 bit controller.

GUIDELINE # 2: 16 bit interface at least

Most of the disk reading we do is sequential. Once the reading
begins, it would be nice to be able to store all of the sectors
from that track into the controller's memory so that the next time
your program asks for a sector, the controller will not have to do
a physical read but can just transfer the data in a single
millisecond. A ST506 MFM or RLL controller does this by having
enough memory to allow full track buffering.

Almost all drives of the ESDI, SCSI and IDE variety have full track
buffering within the drive electronics. Makes no difference where
it is, controller or drive, just so long as it is there somewhere.

PRINCIPLE # 5: Full Track Buffering is Good

Some controllers and or drive electronics are able to low level
format the drive so that not all the tracks start in the same
place. They offset the position of the first sector of each track.
It takes time for the drive electronics to terminate the read of
the last sector on the previous track, switch heads and begin
reading data from the next track. Head skewing allows enough offset
in the positioning of the sectors so that the sectors will be in a
position to be read just as soon as the drive electronics can get
ready to read the data. There is a similar skewing possible to
account for the amount of time it will take the heads to be moved
from cylinder to cylinder. This ability is called cylinder
skewing. Together, these skewing techniques can increase
throughput by 5 to 10%.

PRINCIPLE # 6: Skew is Nice

Some controllers and drives read more of the data we are liable to
need than just a track worth. They have buffers bigger than just
a full track buffer. And they do the reading while we are
processing the data we originally asked for. The seeking and
reading that goes on is totally transparent to the using program.
This is a technique used only in high end controllers and drives.

PRINCIPLE # 7: Read Ahead Logic is Good

How fast you can read and write is limited to how fast the sectors
of the drive can rotate the area to be accessed under the
read/write head. While 95% of all hard disks rotate at 3600rpm,
some very high performance ones rotate faster (3707 and 4000 are
speeds I have seen).

PRINCIPLE # 8: Fastest Spin is Better

Manufacturers have so many different designs that you must be
careful not to judge the reputation of a drive on the basis of a
report on the reliability of a different product line the same
manufacturer builds. Reliability is a hard thing to determine.
Just because your friend has one and his works, should you buy one?
Is that a valid statistical sample? I guess the best place to see
the results from a large sample is the article on page 329 of the
9/25/90 issue of PC Magazine on component reliability and the
survey they had done of their readers. It isn't perfect but it is
liable to be better than a sample of one. Make sure you read the
paragraph on the appropriate series of drives.

GUIDELINE # 3: Know the Reputation

Caching is a technique for storing recently accessed data in memory
somewhere so that if you need it again soon, you can get it from
memory at memory speeds rather than having to read it from disk.
Caching can be done on the controller, in motherboard memory (DOS's
640k, extended or expanded) or both places. Should you buy a
caching controller? Yes and no.

There are two types of caching controllers, one type with a small
amount (usually 32 or 64 k) of cache onboard the controller and one
type that allows several megs of memory on the controller.

With the first type, you can also use a secondary cache located in
motherboard memory. With the second type, you generally can't.

The first type of caching controller costs little more than a
"normal" controller and the caching provides little more than full
track buffering. So buy that kind. With motherboard memory so
cheap, it makes little sense to me to buy an expensive ($500 to
over $2000) caching controller when you can buy 4 more megs of
motherboard memory for $200 and a program for $40 that does the
same thing. There is an excellent review of caching controllers in
the January 1991 issue of Byte which comes to this same conclusion
and should be read if you are looking for maximum file transfer
speeds.

If you must buy a true caching controller with large amounts of
memory on the controller, consider if you want "write- through" or
"write-later" caching. Write-through systems write the data to
cache and also immediately to the physical drive. So, in case of
a power failure, your data has been written. But "write-through"
caches are slower than "write- later" caches in write operations.
"Write-later" caches write the data to cache and then later when
the cache is full or when the heads are passing over where the data
should be written, the data is written out. Some "write-later"
caches have battery backup so even if your machine goes down, the
data will be written out to disk the next time you boot your
system. Only you can determine the amount of speed you are willing
to give up in return for increased safety. You know how reliable
the power is in your home or office. I have used a battery-backup
"write-later" controller for several years with never a problem.

GUIDELINE # 4: Cache in motherboard memory, it's cheaper!

When you buy your subsystem, make sure that you receive a drive and
its manual, a controller or interface card and its manual, cables
for all the floppies and hard drives and tape drives the controller
is ever going to handle, mounting rails for the drive if your case
needs them and if you don't already have them, 5 and 1/4 adaptor
cages for any 3 and 1/2 inch drives you will be mounting in 5 and
1/4 inch mounting holes, colored faceplates of the appropriate
color, mounting screws and any software (diagnostic, partitioning,
device drivers etc.) you will need to install the subsystem.

PRINCIPLE # 9: Get all the pieces

If you buy from one source, you can demand that the controller and
drive will work together and, if there are problems, you will have
only one place to go back to.

GUIDELINE # 5: Get it all at one place

Most controllers will support two hard drives and two floppies.
Some only the two hard drives. There is no real advantage to
separate controllers and two of them will cost more, take up more
space and create more heat. So buy the 2 floppy/2 hard version.

GUIDELINE # 6: Buy an Integrated Controller

Think about doing backups when you configure your system. Using a
floppy backup program, you will want to use two alternating drives
(one writing the data while you change the disk in the other one).
Alternating drives must be of the same drive type (360kb, 720kb,
1.2meg or 1.44meg) which means that you can't have a system have
both 5 and a quarter and 3 and a half inch drive types so you can
read all types of floppies without paying a big penalty in backup
speed.

Get a tape backup if you are getting a bigger drive. Who wants to
load 100 1.44 floppies to do a backup? With a tape drive, I start
a backup and go to lunch. Under these conditions, I do backups.
If I had to fumble with hundreds of floppies, I would seldom (oh,
be honest Mike, the word is "NEVER" !) back-up. There are tape
drives that connect to a three connector floppy cable, that fit in
a 3 and a half inch mounting and that will back up about
120megs/tape. I Use the CMS Jumbo. And they only cost about $300
for the tape drive and $25 for each tape. If you get one of these,
you will need a different type of floppy cable so ask about it.

GUIDELINE #7: How ya gonna backup

Manufacturer's guarantees sometimes begin when the manufacturer
ships the drive to the distributor who then ships it to the dealer
who then ships it to you. Make sure you know who is guaranteeing
the subsystem and for how long. Get it in writing.

GUIDELINE # 8: Know the Guarantee

The WD1003 controller IBM used in the early ATs set the standard
for what commands would be accepted by the controller and what the
controller would do when it received a command. Your best bet is
to buy only a subsystem that is compatible with the WD1003
standard. That way, you'll have no chance for software
incompatibilities. This is particularly important in the OS/2 and
UNIX world but matters with some drive utilities in the DOS world
too. But don't buy the WD1003 itself. It would work but it is
really too slow for a 386 (It fails rules 5 and 7).

PRINCIPLE # 10: WD1003 compatibility is a must

You need your computer's memory for things other than device
drivers. Some disk subsystems need device drivers to provide
support for large partitions (> 32 meg) or drives with greater than
1024 cylinders (DOS has limits of 1024 cylinders, 16 heads and 63
sectors. This can be "overridden several ways. "Drive Splitting" is
a hardware feature of some top-end drives that allows a single
physical drive to appear to DOS as two physical drives.
"Translation" is a hardware feature of a controller or drive that
transforms the physical geometry of the drive to something
acceptable to DOS. Say you have a 2048 cylinder drive with 8
heads. Translation could make DOS think the drive had 1024
cylinders by 16 heads. The third way to overcome limits is through
device drivers. Will they be available for OS/2 or UNIX for your
hardware if you want to go that route someday? See if you can
avoid device drivers or at least buy drives and controllers that
already have device drivers available for the software you might be
interested in someday.

GUIDELINE # 9: Device Drivers are Bad (but sometimes
necessary)

Drives use as much as four times the power they take when running
normally when they are powered up. Some controllers have a feature
called "power-sequencing" that allows them to power up multiple
drives with a bit of time in between each power up. This allows
the power supply to be smaller and to not receive as large a surge
from multiple drives powering up at the same time. If you expect
to have more than two hard drives (via SCSI or secondary
controllers), this is a nice feature to have. But you don't need
"power-sequencing" in a normal machine and we are urging adequate
power supplies of the 200 watt or bigger variety for our build
machines.

GUIDELINE # 10: Power Sequencing Nice at the top end.

ST506, SCSI, ESDI and IDE are standards for passing signals
between the drive and the controller/interface-card. SCSI, ESDI
and IDE put more of the intelligence on the drive itself and less
on the controller than did the ST506 interface originally used on
the XT. (There is a good article on page 427 of the 10/30/90 issue
of PC Magazine describing the techniques these three use to achieve
throughput beyond that of the original XT's ST506 interface.)

SCSI is a standard that allows multiple (7 or more) drives and
other peripherals to all connect through a single adaptor. SCSI
hard drives, CD ROMs, and tape drives can all share the same
adaptor. SCSI is not a new standard but it is one that seems to be
subject to a lot of configuration difficulties. Not that there is
anything wrong with SCSI, it is just that the standard is ambiguous
and different manufacturers have interpreted the standard in
different ways. So you need to be particularly careful when using
SCSI components that they will plug and play together. I see lots
of BBS users reporting compatibility problems. Despite this
warning, there are millions of happy SCSI users. If I were doing
SCSI, I would buy all the parts from one dealer and I would get a
subsystem guarantee that all the components were compatible in
writing. And I would not use any of the cheap 8-bit SCSI adaptors.
Get a good 16-bit adaptor.

The choice of which of these other two (ESDI and SCSI) to use is
less important than in insuring that you follow the other rules we
have mentioned above. I have used ESDI and IDE and can't tell the
difference. Both are available in versions that will COREtest at
better than 900 k/second. Both are available in versions that will
do half of that. If you follow the rules we have talked about, you
will find yourself getting one of the faster versions.

The newer drives almost all use IDE, ESDI or SCSI interfaces so you
will get higher performance and higher value with one of these
three. ST506/MFM and ST506/RLL work fine too, they are just slower
than the newer ones.

You know better than I what you can afford and what you need. I
hope we have helped you understand a bit about hard drives and what
is available and what questions to ask. If you need additional
help, there will be experts at the pre-build session to help answer
your questions and the MIX is always available for questions (use
the HW conference). Good Shopping!

Some Personal Favorite Controller/Adaptors Drive combinations

-------------------------------------------------------------------
-----------
Interface Recording Manufacturer Model Price
-------------------------------------------------------------------
--------------
ST506 MFM Western Digital WD1006V-MM2 $120 (1 to 1
MFM controller, full track buffering, 17 sectors/track, supports
two floppies and two hard drives)

ST506 MFM Mitsubishi MR535 $300 (half
height, 42.5 megs, 5 heads)

ST506 MFM Seagate ST4096 $540 (full
height, 80 megs, 16 heads)

The WD1006/ST4096 combination COREtests at 450k/sec. But MFM is
slower and more expensive per megabyte than any of the other
connection types so buy MFM components only if you are adding an
additional drive to an existing subsystem (And even then consider
selling your old disk subsystem and getting a new fast one. I
would hate to see your brand new 386 slowed to a crawl by a slow
drive interface!). The RLL version of the WD1006 1 to 1 controller
and the MR535 drive cost the same as the MFM versions and give you
26/17ths more storage and 26/17ths more speed. Stay away from
Seagate ST2xx series drives. While any manufacturer can have a run
of bad drives, this series (especially the ST-238/251/277 drives)
seems to suffer from a pair of design defects that cause motor
lockup and striction (heads sticking to the platter).

----------------------------------
------------------------------------------------
Interface Recording Manufacturer Model Price
---------------------------------------------
-------------------------------------

ST506 RLL Western Digital WD1006V-MR2 $120 (1 to 1
RLL controller, full height buffering, 26 sectors/track, supports
two floppies and two hard drives)
ST506 RLL Mitsubishi MR535 $300 (half
height, 65 megs, 5 heads)

The WD1006/MR535 combination COREtests at 720k/sec



-----------------------------------------------------------------
Interface Recording Manufacturer Model Price
-----------------------------------------------------------------

ESDI Ultrastor 12f $177
ESDI adaptor (get the 32k buffer version if you can
-supports
two floppies and two hard drives)

ESDI Maxtor XT4170E $895
(full height, 157Megs
7 heads, 36
sectors/track)

This combination COREtests at 930k/sec.

-----------------------------------------------------------------

IDE MiniScribe adaptor (no model number - for
use with Maxtor drives - supports two floppies
and two hard drives)

IDE RLL Maxtor LXT200A $926 (3.5
inch drive so you may need a 5.25 inch mounting adaptor (TEAC
CS-235), 201 megs, from 33 to 53 sectors/track (!), 7 heads (price
includes IDE adaptor and cables))

This combination COREtests at 1050k/sec.

-----------------------------------------------------------------

SCSI Always IL2000

Though not based on personal experience, people I trust have
reported excellent results with this controller. It seems to be
faster than the competition (above 1 meg with a normal machine and
up to 1.4meg reported with a bus speed of 14mhz) and creates no
problems with installation and compatibility.

------------------------------------------------------------------

All prices are current mail order and may or may not include
cables, mounting brackets, adaptor kits, rails, s/w, taxes,
shipping, installation documentation, a help line, etc. They are
listed to give you an estimate of what various levels of
performance will cost. COREtest is a widely used drive performance
measurement program. COREtest figures will always be lower than
theoretical figures for a drive due to DOS overhead.

-----------------------------------------------------------------

About the Author: Mike Focke works for HFSI, a systems
integration company that sells exclusively to the US Government.
His published articles include reviews of hard disk controllers for
the XT and AT class machines, an article on disk optimization
techniques, reviews of Disk Optimization S/W, etc. He is a
frequent participant on several nationally echoed Hard Disk
Conferences. He is currently designing and testing a hard disk
diagnostic program (DTST300.ZIP when it appears on the bbs).



Additional issues in Hard Disk Selection:
Size, Speed, Vendors, and Prices as of January, 1991.

By Doug Wagner

Written for the Capital PC Users Group. January, 1991.

I agree with virtually everything that Mike Focke suggests in
DRIVEBUY.ZIP, but I have a few more things to add and a few further
explanations. These comments reflect my experience buying hard
disks with severe budget constraints over the last 5 years. Except
for the disk speed tables, the comments contained here are somewhat
less hard fact and more interpretation of the facts than in Mike
Focke's "Principles" and Guidelines. My opinions are not always
correct. Questions or corrections can be directed to me in the HW
conference of the PC User's Group Mix. (301-738-9060)

One important principle to follow to avoid getting burned is
to avoid the leading edge of hot hardware. I tend to wait until
something has been out for a while, in the hopes that the early
bugs will have been cured, and the gadget has established
sufficient market volume so that it will be a long term success and
not a dead end. Owning orphaned hard disks is like driving without
a seat belt in the Dodgem cars. It may not matter, but bumps are
coming. For example, I was unwilling to buy IDE drives 6 months
ago, but now will buy nothing else in the 80 to 200 megabyte range.
The IDE interface is now clearly going to be the way of the future.
If I were richer, or working in a for-profit environment, I would
take more chances with new stuff from the large vendors, because
PCs are so fundamentally productive.

Another example of an important technology which is not yet
ready for prime time is the new hardware compression cards.
EXPANZ! has been out for 4 months and the STACKER is advertised but
not quite yet shipping. Both of these manufacturers provide a card
with a CPU dedicated to executing a disk compression routine for
about $200. The adds claim "double your disk space". This
technology promises to provide a cheap way to increase the capacity
of your current disks by 50% to 75%, probably, within a year or so.
But there are clearly compatibility problems with EXPANZ!, and
Stacker is not shipping in bulk yet, if shipping at all. The
feasibility of using these things reliably without spending undue
amounts of time becoming an expert in new levels of complexity is
not yet established.

Also, remember that computer hardware and software are getting
more powerful at about 20% to 30% per year. Which means what makes
sense this year, might require some changing around next year
because of the coming technological improvements in PC's.

DISK SIZE AND UPGRADE STRATEGIES

My rule of thumb is to try to get at least TWICE the disk
capacity that will be immediately occupied. For serious data
processing machines that rule of thumb is too low. Human time is
too valuable to have someone spending 1/3 of their day trying to
shrink their files to make space to do something else. Also, it
seems to take 2 hours of time to put in a new hard drive and get it
formatted and usable, no matter how many times I do it. On my
first time with a different disk type it takes 3 or 4 hours. So I
don't want to do this very often.

Application software packages are growing in the disk space
they occupy. This also means your software is more dependent on
disk reading, swapping in different parts of the program as you do
things as well as reading your own data files. For example, Word
Perfect 4.2 occupies about 1.1 megabytes of disk space, but WP 5.1
occupies about 3.8 megabytes. Likewise, Quattro consumes 1
megabyte but Quattro Pro consumes about 4 megabytes. In order to
have pop-down menus and more options, fonts, etc, commercial
software is consuming more and more disk space. Therefore, disk
size and speed are becoming even more important.

I WOULD RECOMMEND THAT AT THE MINIMUM YOU GET A 66 MEG RLL OR
AN 80 MEG IDE DRIVE IN ANY 386. I favor 1/2 height drives, and
only one initially, so that I have the option of getting a second
drive of equal or bigger capacity next year. Remember, the
standard AT CMOS Setup allows only 2 hard disks in a single
computer. The only standard way around this is with SCSI, which
allows up to 6 or 7 SCSI drives hanging on the single SCSI
controller (if they are all compatible?) but only 1 non-SCSI hard
disk on a second hard disk controller. There may be some other
special ways around the 2 drive limit, but they are not widely
used, and therefore I would be scared to use them because of
compatibility problems.

I suggest the 66 megabytes as a minimum, because you can get
a 66 meg RLL disk with an RLL controller for only about 60$ more
than the standard MFM 42 megabyte disk drive. The RLL 66 meg drive
is 20 to 50% faster than the MFM and has 50% more capacity.
However, it is not yet clear whether or how one can install an IDE
drive in a computer that already has an existing MFM, RLL, or ESDI
drive. So, the choice to get an RLL drive, may imply getting rid
of that disk drive and controller if you subsequently need more
than about 130 megabytes of disk space. SCSI drives can usually be
installed to co-reside in a computer along with any single hard
disk of a different type. So, if you anticipate going to more than
130 megabytes eventually, I would start off with an IDE 80 or 125,
and plan on adding a second IDE drive later. If you anticipate
getting to more than 400 megabytes of disk space, you might be
better off starting with a 330 or 660 megabyte ESDI drive. But if
you were buying 660 megabytes now, a single full height 660 would
cost about 1900$, while two 330 megabyte ESDI drives with the same
total space would cost 2800$ or so.

IDE versus ESDI versus SCSI.
First, each protocol requires its own type of disk controller,
and most disks will work with only one of the protocols. Second,
there are two bottlenecks in disk speed--the controller protocol
and the hardware speed in the drive. The MFM protocol is slowest
because its controller speed limit is about 500K characters per
second, and its disks contain only 17 sectors per track, limiting
the amount of data that can be read in a single revolution of the
disk to 8.5K. RLL raises the controller speed limit to about 800K
per second, and has 26 sectors per track, meaning that in the same
revolution of the hard disk it can read 13.5K characters. The
SCSI, ESDI and IDE protocols all allow many more sectors per track
(34 to 78), and have a current controller speed limit in the
vicinity of 1 Meg/sec. Some of the 330 and 660 megabyte ESDI
drives can go to 1.5 megabytes per minute or 2 megabytes per minute
with the correct controller.

For drives between 80 and 200 I will be buying exclusively IDE
protocol drives. The IDE drives have really pushed the price/
performance barrier in the past 6 months to new levels. The new
Seagate 71 meg IDE drive is about 350$, and it is 2.5 times as
fast as the 330$ Seagate 296N SCSI 84 meg when the latter is run on
the cheap Seagate 8 bit SCSI controller. I have seen the 200 meg
Maxtor and Seagate IDE drives advertised for 775$, and they only
need a 20$ to 50$ card to interface to your computer--while
comparable performance SCSI or ESDI controller cards cost from 160
to 300$, and the ESDI/SCSI hard disks tend to be somewhat higher
priced also. The IDE drives are usually 1/2 height, 3 1/2 inch,
low power (1 to 10 watt) drives and they are recent engineering
designs. In contrast, many of the 150 meg ESDI drives are full
height, 3 or 4 year old designs that consume 25 watts of power,
generating more heat and more vibration. There are also 1/2 height
ESDI's in the 150-300 meg range, which tend to be newer designs
than full height drives.

As of December, 1990, you must use either SCSI or ESDI above
200 megs with the exception of 2 drives from Seagate. Above 200
megs, most drives are either of size 330 or 660, and most are
available as either ESDI or SCSI. There appear to be fewer
compatibility problems with ESDI, but the SCSI protocol is going to
be pushed in lots of larger mini-computer and mainframe
applications (Business Week, Dec 25, 1990), and performance
breakthroughs in big machines in the SCSI format may move down to
PC's. Vendors are now selling the same 5 1/4 inch SCSI hard disks
for use in mainframes, minicomputers, or PC's. A new SCSI protocol
(SCSI-2) is coming that increases the disk I/O speed to 4 megabytes
per second, but it is not quite here yet, and may still be a year
or two off.

A second new SCSI protocol that is much more expensive, but
promises vast performance improvements is the Drive ARRAY/Disk
Striping approach. This is what will soon be eating into the
mainframe and minicomputer disk market and is available for Compaq
SystemPros now. It runs 2 to 10 different physical disks as one
large disk from the operating system's point of view, by stripping
each character into the 8 bits and writing part of each character
to a different disk. This has radical speed advantages, and with
10 different disks, you can achieve new levels of redundancy
checking. A 10 disk array of 900$ 200 meg Conners 16 ms drives
would appear to the computer as an 8x200 = 1.6 gigabyte disk with
an average access time of 2.0 milliseconds. 2 of the disks would be
devoted to error checking and data redundancy. Unfortunately, the
controllers for these gadgets now cost 5,000 to 20,000$.

ESDI has a slight compatibility edge in PC's now because
1)there is greater incompatibility between SCSI controllers and
specific SCSI devices, and 2)the SCSI drives usually require use of
a separate address space in High RAM between 640K and 1024K. This
can lead to conflicts with other device drivers, network cards, ROM
Shadowing, etc. This should be solvable, but it is one more
constraint on the use of addresses in the upper range and one more
conflict to dodge in setting up 386 memory managers and multi-
tasking.

The one advantage that SCSI has over ESDI and IDE drives in
compatibility is in non-standard ROM BIOS machines or in older ROM
BIOS machines. That is because the SCSI drives are not even
entered into the setup but ESDI and IDE drives are entered in the
setup. Therefore, you have to have a drive type in the list of 14
to 48 physical drive types that matches the drive you buy, or you
have to have a "customizable" drive type in the ROM BIOS, as the
newer AMI, Phoenix, and AWARD ROM BIOS's do have. If you don't
have a customizable drive type (where you can type in the number of
tracks per head, number of heads, and number of sectors per track),
then an IDE drive can still be used without a device driver as long
as the disk is not larger than the largest total amount of disk
space available in the largest fixed drive type available. The IDE
drives have sector remapping built into the drive to translate from
the CMOS table to the drive's actual physical configureation. You
might loose some disk space here. Another alternative is to use a
hard disk device driver such as On-track's generic DiskManager
version 4.2, which handles translation from the CMOS drive types
into a layout that the drive needs. With an older ROM BIOS the
translation from a device driver such as DiskManager is more likely
to be necessary with an ESDI drive than with an IDE drive.

DISK SPEED

The faster the better. But how fast and how do you judge?
First, using an 8 bit MFM or RLL controller in any AT or 386 is
disastrously slow. They require an even larger interleave than the
3 or 4 to 1 they usually require in an XT. So a standard AT 2/1
hard-disk controller will provide a radical improvement in speed,
and can probably be bought used for 30 or 40$.

However, for 100$ you can buy a new 1/1 interleave controller
which does provide substantially faster disk I/O than the 2/1 for
2 reasons: first is the 32K cache on the controller and second is
1/1 versus 2/1 interleaving, which means that the computer can read
a whole track in one revolution of the disk instead of 2. One way
to read interleave numbers is the top number is the number of disk
revolutions required to read all of the data on a single track.

If you are doing data base processing on files large enough to
occupy more than one track on your hard disk, then skewing, track
to track access time, and average access time become more
significant. The average access time doesn't matter as much as the
track to track--IF YOU KEEP YOUR DISK DEFRAGMENTED--and the
opposite is true if your disk is fragmented. There can be
discrepancies between track-to-track and the average seek time to
find a random other track. Unfortunately, track to track times are
not advertised as much as the average access time, and it is
difficult to find out which controllers support skewing. Thus, no
matter what kind of disk you have, if you are doing substantial
data processing on large files you will need to use the disk de-
fragmenting program available in such widely sold commmercial
software packages as Norton Utilities, PcTools, PCKwik Power Pack,
etc.

There are a number of speed testing programs, but none are
perfect. There are "rifle shot" testing programs that test the
hardware's ability to read a specific sector and develop hardware
speed measures, and there are other testing programs that read and
write a 500K to 1meg test file and measure the elapsed time. Both
are useful pieces of information though neither is perfect measures
of how well your own application software will work with different
disks.

Performance measurement is like looking at the world through
multiple layers of a veil. You are rarely sure whether what you
observe is an object on the distant horizon or a fly that landed on
the veil. The best speed tester available seems to be CORETEST,
which is available on bulletin boards. PC MAGAZINE also has
produced a standard performance testing program that is on bulletin
boards. NORTON, SPINRITE, and CHECKIT are commercial software
packages that contain disk speed tests, as part of their package.
Beware, however, that SPINRITE and CHECKIT's disk speed measures
are pure measures of how dense the data is recorded on a single
track, and not composite measures of how fast the disk will be able
to access data files under DOS. In other words the speed measures
in CHECKIT and SPINRITE ignore track-to-track and average seek
times. NORTON and CORETEST seem to reflect the head movement
speed as well as data density.

The following tables summarize some disk speed comparison
tests using the Coretest v 2.92 speed testing software. Readers
should be cautioned that no speed testing software is perfect.
Coretest is probably the best available but it may be biased
against or for particular drive/controller combinations. Coretest
results are not sensitive to the difference between AT's and 386's,
or to clock speeds of the CPU. There is some variability, however,
in individual drive performance within model types. In fact there
is likely to be much more variation in the performance of 2
different disk drives of the same model number than in most other
parts of the computer. Finally, how a disk is formatted may
influence disk speed, and 386 memory managers such as Windows or
QEMM do slow down disk I/O. Many of the disks listed below belong
to other people, and I do not know for sure how their machines were
set up. I do know that they were not running disk cache's, other
than the 16K to 32K RAM track buffer that is built into many
disk/controller systems.

The last column of numbers in Table 1 is recent mail order
prices. These are prices for naked drives, excluding controllers.
Some vendors include cables, 3 1/2 to 5 1/4 adaptor kit and Ontrack
Software, and others don't.

Table 1. Hard disk size, speed and prices

coretest results
drive megs protocol controller k/sec index access times Price
trk-trk average $
40 meg drives

Seag 251 42 MFM DTC 7280 450 4.8 13.4 26.1 230$
Seag 251 42 MFM wd1006v.MM2 450 4.4 13.7 28.5 230
WD 93044a 42 IDE ? 620 5.6 1.8 28 230
WD 93044A 42 IDE DT-3767 669 5.9 x 28 230
Seag 157n 42 IDE ? ? 230
Mits 535 41 MFM DTC 7280 441 4.5 5.3 29.2 260
Seag 157r 49 RLL DTC 7287 713 6.5 7.7 24.5 250
Seag 157r 49 RLL WD1006v.mr2 689 5.9 7.7 31.0 250

60-80 meg drives
coretest results
Mits 535 64 RLL Wd1006v.mr2 689 6.5 5.6 22.6 260$
Tosh 134 66 RLL DTC7287 716 6.6 5.4 23.9 310
Seag 280A 71 IDE ? 657 5.8 4.5 28.4 335
Seag 280A 71 IDE ? 720 6.6 4.3 24.5 335
Seag 4096 84 MFM ? 486 4.7 7.3 30.5 500
Seag 296n 84 SCSI Seag ST01 312 3.5 10.3 32 330
WD 280 81 IDE DT3767 981 9.1 4.9 17 470
Maxt 7080 81 IDE ? 805 7.7 5.6 19.1 435
ConP 3184 84 IDE ? ? 490
Quantum 85 ESDI Zenith? 455 5.3 4.9 21.3 ?
Seag 1102 89 IDE ? 845 8.0 1.0 18.7 420
Hardcard 80 SCSI Built in 439 4.7 3.6 26.1 ?

90-160 meg drives Kb/sec Index Access times Price

coretest results
Tosh Mk234 105 IDE DT3767B 820 7.3 1.0 23.2 480$
Seag 1126a 110 IDE ? 774 6.8 5.6 24.9 600
Seag 1144a 125 IDE ? 970 8.8 1.0 18.3 499
ConP 3104 100 IDE ? 560 5.6 5.6 24.3 520
Miniscribe 150 ESDI Zenith? 994 9.6 4.7 16.0 ?
Maxt 157 ESDI Ultrastor12f 930 ? ? ? 880
200 meg drives
coretest results
Seag 1239a 211 IDE ? 1014 9.8 1.0 14.5 775
Maxt LXT200a 200 IDE Miniscribe 1000 9.6 3.0 15.2 749
ConP 3200 200 IDE ? 1126 10.6 1.9 14.2 840
WD Piranha 212 IDE ? 1730 14.3 3.2 13.7 800
Seag 2383e 330 ESDI Zenith* 1400 11.8 3.3 15.0 1400
Maxt 4766E 660 ESDI WD1007v.SE2 775 7.8 3.4 17.1 2000

For comparison purposes, I have also included some benchmark
numbers using coretest on some older Disk drives in AT machines and
in XT machines in Table 2. Thus, if you are wondering how much
faster some hot new drive will be than your existing drive, you can
look up in table 2 a drive similar to your existing drive, and
compare the speed numbers with those you are considering in Table
1 above. Or if you know the Norton SI speed rating on your disk,
you can find a drive in Table 2 with the same Norton SI, and read
its Coretest Index number off the same line. Then Compare Coretest
numbers with those in Table 1.

I tend to examine both characters per second and the index
number. It is the index number that is probably the best summary
comparison number. If your current disk drive has a Coretest Index
of 3.0, and the disk you are thinking of buying gets a 6.0 on the
same measure in the table above, then that disk is approximately
twice as fast as your current disk.

Table 2 disk speed in older MFM/RLL drives in XT and AT computers.

coretest results
drive megs Inter- controller k/sec index trk-trk average Norton
leave (ms) (ms) SI
In XT computers

Seag225 21 4/1 WD1002WX2 123 1.45 20.3 76.9 1.4
Seag238 32
Seag4038 32 2/1 DTC 5150 2.1
Seag251 42 4/1 WD1002WX1 122 2.9 12.9 25.5 1.8
Seag251 42 3/1 WD1002WX1 167 3.2 16.0 25.4 2.2
Kalok 32 4/1 RLL DTC 5160 156 2.4 11.1 38.5 1.3

In AT Computers with 16 bit controllers

Seag225 21 2/1 WD1003WAH 241 2.2 20.2 70.5 2.2
Seag4053 44 2/1 WD1003WAH 237 3.2 6.5 30.2 3.3
Seag251 42 3/1 WD1003MM2 161 2.8 15.6 29.7 2.4
Seag251 42 2/1 WD1003WAH 222 3.2 13.0 29.8 3.4
Seag251 42 2/1 WD1003WAH 222 2.7 15.8 38.7 2.6

RECOMMENDATIONS

What do Tables 1 and 2 mean?
FIRST, they illustrate the substantially higher disk speeds
available with the IDE protocol and the RLL protocol than the best
available from 1/1 interleave on the MFM protocol. That is why I
would now avoid the MFM protocol. I would avoid any MFM disk
drive.
SECOND, most of the 80 to 120 meg IDE drives are not that
much faster than 66 meg RLL drives with the notable exception of
the Western Digital 81 meg and the Seagate 1144A 125 meg drive. So
an RLL disk drive protocol is acceptable, if the total disk space
you need is not larger than 132 megabytes.

THIRD, The cheap Seagate 296N with the accompanying Seagate
SCSI controller is the slowest disk drive tested in all of Table 1.
It should be avoided, unless you need SCSI and cannot afford a
faster controller.

FOURTH, all of the 200 meg drives are substantially faster
than most of the rest of the disks, and they are cheaper per
megabyte. So, if I knew that I needed 80 megabytes of storage now,
I would be inclined to get the 200 rather than a 125 meg drive.

FIFTH, the ESDI results do reveal that there is some
variability in ESDI drive speed, but it is possible to get up to
1400 characters per second with 330 meg ESDI drives and the
appropriate $200 controller. So for maximum size and speed ESDI
(and SCSI) do offer some advantages over IDE.

Finally, it appears that the WD drives are a bit faster than
most of their competitors on this test. It is possible that this
apparent advantage would be reduced in real data analysis with
larger files, because these disks have larger (64K) caches built
into the drive than the 32K cache built into many of the other
drives. On the other hand the advantages may be real. A much
larger effort would be required to test that hypothesis.

For new machines, the IDE protocol is probably the best
choice, unless you 1) have an existing disk controller and disk
that you want to use, perhaps with a second new disk, 2) you
anticipate growing to more than 300 to 400 megabytes of data within
12 to 18 months, or 3) this is a network server. In latter 2
cases, you are probably better off with a full height 660 megabyte
disk drive from the beginning. But barring those three exceptions,
the IDE protocol now delivers more disk capacity and speed per
dollar, and it has a lot of growth potential in half height disk
drives, up to 330 megabytes on a single disk. (Conner Peripherals
has just announced a 510 megabyte IDE drive).

The choice of disk drive protocol, RLL versus IDE, versus ESDI
or SCSI is important because it constrains your future upgrade
choices and costs. It is difficult if not impossible to have 2
hard disks of different protocols in the same computer among the
MFM, RLL, IDE, and ESDI protocols. A computer with one of those
disks can usually also have a SCSI disk, but a computer with one of
the MFM, RLL, IDE, or ESDI usually cannot also have a different one
of those 4 types also. Thus, if you get an RLL now, and need more
space next year, you will have to get either an RLL or a SCSI.
One caution about IDE disks is that most IDE disks above 70
megabytes will probably require that your CMOS have a "customizable
drive type" in the setup. Older ROM BIOS chips may be missing
that, or may not have a customizable drive type that works. Thus
if your BIOS is older than early 1990, you may need a new ROM BIOS
chip. One specific brand, Conner Peripherals, was reported to be
incompatible with the AMI BIOS dated before April 1990. In the
absence of a customizable drive type, On-Track's generic Disk
Manager version 4.2 might solve the problem, but that might also
cause compatibility problems with Windows 3.0.

BRANDS-DISKS

I tend to stick with the market leaders in volume for several
reasons. First, there are enough of them around that I can get
good prices and will be able to get them repaired cheaply, if
necessary. Second, why are they the market leaders? Their
hardware must be relatively good. Third, everything works with the
high volume disks. New software and hardware developers design for
compatibility with the largest number of machines they can achieve.
They design for compatibility with the highest volume producers
first.

For an RLL 66 meg half height drive I also have heard very
good things about the Mitsubishi MR 535. The Toshiba is $20 to $40
more expensive but it has more heads and is a 3 1/2 inch drive.
The Seagate 277r is of the group of Seagate drives which have had
many reports of "stiction", where the drive will no longer start
up, but once started will continue to spin. One disadvantage of
the Mitsubishi MR535 is that you often have to use a "customizable
Drive type" for it. Even though there are 3 different drive types
in my AMI BIOS with 977 tracks per head and 5 heads, none of them
would work.

Right now the market leaders in the IDE protocol are Conner
Peripherals, Seagate, and Maxtor, with Western Digital probably
4th. Each makes a 40, 80, 100, and 200, and Seagate also makes a
125, 160, 240, and a 330. The 200's have faster seek times than
the 80's and 100's, and usually have denser recording of data, with
more 512 character sectors per track, so they should produce faster
throughput.

Seagate also has 2 overlapping lines in the 90 to 200 megabyte
range, with the more expensive ones having faster 15 Ms seek times
and the slower ones 20 millisecond seek times. This has more
serious implications for the do-it-yourselfer than might be
thought. The 15ms access time disk drives are the line of disk
drives developed by Control Data Corporation, whose disk drive
subsidiary Seagate bought 15 months ago. The drives were designed
to be installed in many mini-computers and are generally accorded
high quality ratings. However, they were not designed to be
installed by a do-it-yourselfer in PC and they are somewhat more
difficult to install physically.

At least the 200 meg Seagate 1239A that I bought last week was
more difficult. It is the first disk drive I have ever seen
without a clear mark on the data connection end of the disk drive
as to which end was line 1 and which end was line 40. Also, it was
the first disk I have seen in 50 whose screw holes did not line up
with the standard locations for physical attachment. To compound
the insult, Seagate apparently does not routinely include a manual
with the disk drive. In contrast, a 20 millisecond Seagate 1144a,
one of the Seagate line as opposed to the Seagate owned
CDC/Imprimis/Swift line, was the easiest disk I have ever
installed. The absence of a manual can be solved by downloading
Seagate manuals from their bulletin board. But the different
height for the screw holes can cause this half height drive to
occupy 2 half height bays, or might make it difficult to firmly
attach in a 3 1/2 inch bay.

For disks over 300 megabytes there seem to be 5 or 6 large
manufacturers, with the three leaders being Maxtor,
Seagate/CDC/Imprimis, and Micropolis. Hewlett-Packard, Fujitsu,
and Microscience also compete here, as do a few others, including
Core. The first 6 manufacturers produce 660 meg drives with
essentially identical advertised performance and similar prices.
Core drives are more expensive, but I think they provide a 5 year
replacement warrantee, while the others come with a 1 year or 2
year (Maxtor) warrantee.

BRANDS -CONTROLLERS

First, for IDE drives there is not much to the controllers.
Some new computers have the IDE interface built into the
motherboard and others require a separate card, but in the latter
case there really is not much intelligence on the card. The works
are on the disk drive. So controller performance probably only
matters for SCSI, ESDI, RLL, and MFM.

If you are concerned about Novell, Os/2, or Unix, Mike Focke's
discussion about WD1003 compatibility is extraordinarily important.
It may be important for other things too, but it may not. WD is
Western Digital, and most of the other controller manufacturers
mentioned here make controllers which achieve WD1003 compatibility.
Often the other brand controllers also have a non-compatible mode
that is useful in some circumstances, or might provide higher
throughput.

WD makes both SCSI and ESDI controllers as well as MFM and RLL
controllers. However, in among SCSI vendors Always, Future Domain,
and Adaptec seem to be the market leaders. But it is crucial to
obtain a controller that works with your disk. For SCSI it is
important to buy controller and disk from the same vendor. In the
ESDI arena the most widely sold brands appear to be the Ultrastor
12F, WD 1007v.SE2, and Adaptec. WD has just come out with a WD1009
ESDI controller which is rumored to be quicker. Another
manufacturer that has long produced a good line of MFM and RLL
controllers is DTC, which is a subsidiary of Qume. Note, many of
these manufacturers have a lower priced model which does not
support 1/1 interleave or does not have a full track buffer on the
controller. It is penny wise and pound foolish not to get the
controller with the 32K or 64K buffer/cache built into the
controller.

I stay away if at all possible from no-name controllers or
off-brand controllers. A few dollars here can make a world of
difference to performance and compatibility in the future. There
are often new brand names with some "hot" new board, but in disk
controllers I tend to be most conservative.

VENDORS.

There are a wide variety of vendors of these disks. Many mail
order places advertise in PC MAGAZINE, the PC SHOPPER, PC WEEK or
INFO WORLD, with PC SHOPPER probably having the best disk prices.
If you are buying a big disk, you ought to check out the prices in
the adds, and in Softwarehouse, the local superstore in Tysons
Corner, before negotiating with other retailers. Many other
retailers will probably come close to or beat the mail order price,
and they may help you format the disk and provide valuable
advice/handholding. If they don't, or you don't need that
assistance, why pay more?

Among the larger mail order vendors are
1)Hard Drives International (AZ) 1-800-736-DISK
2)PCPros/Touche (Darien IL) 1-708-810-1010
3)Treasure Chest (Louisiana) 1-800-245-3040
4)Computer Products Corp (Col.) 1-800-338-4273
5)Quick Electronics (Florida) 1-800-338-4072
6)Megahaus (Houston, TX) 1-800-426-0560
7)Warehouse 54 ( Where??) 1-800-735-0054
8)Dirt Cheap Drives (Texas) 1-800-872-6007

PC SHOPPER has a vendor index, so you can look up what page the
adds are on. I have bought from 4 or 5 of these vendors, as well
as local suppliers, but have not bought from all of them, and I
have bought from other vendors not listed here as well as local
retailers.

I suggest buying on Mastercard, and stopping payment if they
don't include the manual for the disk and the controller. Two
years from now, when you want to reformat to use this hard disk on
your new 586 and quad speed disk protocol, you will probably need
to change a jumper setting on the controller or the disk. Some
vendors include On-track formatting software with all big disks,
and others don't. So ask what is included. Also, some include
cables, rails, and some include 5 1/4-3 1/2 mounting kits, while
others charge extra. Also one needs to ask about warrantee, and
time allowed for returning hardware that is dead on arrival.

Regarding latest prices, in looking through PC SHOPPER to get
the vendor phone numbers above, I discovered the following best
prices (with vendor number from list above in parentheses): Seagate
251-1 229$(5), Mitsubishi 535 260$(5), Seagate 277r 257$(3), Maxtor
Lx200 749$(4), Conners 200 meg 839$(8), Seagate 1144A 495$(3),
Seagate 1239A 775$(3).


FORMATTING, PARTITIONING AND DOS CHOICES.

Compatibility with the WD 1003 standard can be a problem, because
the limits in that standard were 1) 16 heads (or surfaces) per
disk, 63 sectors per cylinder, and 1024 cylinders per head. When
the standard was set unfortunately they did not anticipate the
great increase in the material sciences and stepper motor control
of head location, allowing much denser data recording on magnetic
surfaces. Most disks over 100 megabytes now use more than 1024
cylinders per head, and some of the bigger disks are using up to 78
sectors per cylinder. In order to achieve DOS compatibility, low-
level formatting software has to do some sector-remapping, which
can, depending on how it is done, cause problems with WD1003
compatibility.

The IDE drives usually do not require a low level format--just
partitioning and high level formatting. Apparently they have the
sector remapping built into the disk in ways that allow the
computer to think it is a 1003 compatible device.

The ESDI's and SCSI's usually do require a low-level format,
and how you do it can be difficult, primarily because of the issues
related to mapping sectors above 1024. Another issue regarding big
disk drive partitions is how do you maintain a logical disk
partition over 132 megabytes. There have been a number of reports
on the mix of difficulty doing defragmentation with either PCTools
v 6.0 or Norton v.5.0 on logical partitions greater than 132
megabytes. With a 660 megabyte ESDI drive low-level formatted and
partitioned with Disk Manager Generic version 4.2, and high level
formatted with DOS 4.01, however, I have successfully defragmented
a 200 megabyte logical partition containing 190 megabytes of data
with PCTools version 6.0.

At this time, MS DOS 3.3, MS DOS 4.01 and DR-DOS 5.0 are the
options. Microsoft's MS DOS 5.0 is due in April 1991, and is
likely to dominate earlier versions of MS DOS and offer the same
advantages currently in DR-DOS 5.0. DR-DOS is believed to be 100%
compatible but it was done by a different company, and only a tiny
minority of PC users have tried it yet. The manufacturer of DR-DOS
5.0 is, however, a highly regarded operating system programming
firm. They developed and distributed CP/M.

The principle advantages of DOS 4.01 are that it allows larger
than 32 Megabyte hard disk partitions-both in the boot drive and in
upper logical drives. Secondly, it is compatible with Windows 3.0.
The 3 disadvantages are 1)it is limited to 1024 cylinders and to
512 Megs of storage per disk without using 3rd party software to
partition, 2)DOS 4.01 uses more of the 640 K of DOS RAM for systems
programs than does DOS 3.3 3)DOS 4.01 is a bit harder to install
because of its 5 floppy disk shuffle installation.

DOS 3.3 forces the boot drive to be less than 33 megabytes,
and all other partitions are also less than 33 megs, unless you use
Speedstor or OnTrack's Diskmanager to partition your disk. In that
case, the boot drive is still restricted to 33 megs, but the upper
partitions can be larger. Disk Manager may achieve better
compatibility with the WD1003 standard and may achieve worse
compatibility, depending on the particular disk drive and
controller. There appear to be some potential conflicts between
the way some IDE drives handle their sector remapping and how On-
track's DiskManager tries to set up the sector remapping. By
sector remapping here, I mean how the disk drive physical layout
with greater than 1024 cylinders is mapped so that the computer
thinks it has a 1024 cylinder disk with more heads or more sectors
per head than actually physically occur on the disk.

Recently, I have had some serious difficulties with an IDE
drive formatted with Disk Manager generic version 4.2 and DOS 3.3.
Diskmanager handled the CMOS drive type differently than I was
expecting, and I believe that the Diskmanager BIOS overlay that
handles sector translation may be incompatible with the sector
translation built into this particular IDE drive, or at least the
way that I have done it is not compatible. Therefore, I WOULD
RECOMMEND THAT WITH THE LARGER IDE DRIVES, ONE USE DOS 4.01 AND USE
THE DOS FDISK PROGRAM TO PARTITION THE DISK if you want disk drive
partitions greater than 33 megabytes. If you do not need partitions
greater than 33 megabytes then either DOS 3.3 or Dos 4.01 will do,
but one should probably use the DOS FDISK rather than Diskmanager.
This may merely a problem with an old ROM BIOS, correctable by an
upgrade, but the IDE drive is not functioning reliably using the
generic DISK Manager.

Summary.

Microcomputer hardware is getting better at 20 to 30% per
year. Maybe what is available now will be good enough forever, but
maybe not. Current hardware selection that is cheapest and best to
do the job now may imply higher future upgrade costs. Hard disk
size and speed will be ever more important in running a
microcomputer to do new applications, because many new applications
are now being designed to run with a hard disk, without much
concern for usability with floppy disks alone. Think carefully
about the trade-off between best performance per dollar now, and
best performance per dollar next year. Your hard disk needs will
grow.

For a 386SX or larger computer I would now buy, at a minimum,
a 66 meg RLL drive or a 70/80 meg IDE. Because of the rapid pace
of conversion to the IDE protocol, using an IDE protocol intially,
even with only a 40 megabyte drive, will maintain greater options
for low-cost future upgrades, as long as your total disk space
needs do not exceed 400 megabytes. The IDE protocol seems
attractive now, up to 200/330 megs on a single disk. The 200's are
not only cheaper per megabyte of storage than smaller disks, but
also faster than smaller drives. If maximum disk size and speed is
needed ESDI or SCSI on 330 meg or 660 meg disk drives can
apparently outperform the 200 meg IDE drives, though at
considerably higher total cost but still lower cost per megabyte.
Also, for networking, there may be substantial advantages to
ESDI/SCSI interfaces.



----------------------------------------------------------------
About the author: Doug Wagner works at George Washington U. Medical
School as a statistician/health services researcher in a long term
research project to develop a better measure of severity of illness
for acutely ill, hospitalized patients. The research project has
transitioned from being the largest paying customer of University
mainframe computing to using a VAX 780, to based entirely on PC's
for extensive statistical computation and data base management.
Doug Wagner can be reached in the Capital PC User's group MIX
bulletin board (301) 738-9060 in the HW (hardware) or TRAIN
conferences.



 December 13, 2017  Add comments

Leave a Reply