mlawless@ncrwic.Wichita.NCR.COM (Mike Lawless) (03/24/89)
I have recently completed a major reconfiguration on the AT-compatible
system I use in a home-based word-processing/resume business (we also do a
certain amount of disk format conversions). This involved addition of
various drives and controllers, as well as the software to control them.
Since there are frequent questions on the net about what can or cannot be
done, I thought the net in general might benefit from my experience.
I started with a PC Designs model GV286-12 (a 12 Mhz AT-compatible with 8
expansion slots, 1 megabyte of memory (640K base, 384K extended), a 2 MB
Intel AboveBoard-PS configured as LIM 4.0 expanded memory, one parallel and
two serial ports, two Toshiba drives (A=360K, B=1.2M) controlled by a
Western Digital WD1002-FOX flex controller, and an old-style IOmega
Bernoulli Box controlled by their PC0 controller (booted from floppy). The
display was (and still is) Hercules monochrome graphics. This machine also
has a proprietary 32K static RAM cache, allowing it to run with zero wait
states using 120ns DRAMs. It has jumpering allowing you to "cripple" the
I/O bus, adding wait states to make the bus no faster than a 6 Mhz AT (while
the processor continues to run at a full 12 Mhz), allowing you to use older
interface cards that can't hack it in a 12 Mhz, zero-wait-state machine
running the bus full speed. I was having to use this feature because the
IOmega PC0 board would not run at 12 Mhz.
The changes I made were:
Upgraded to the newer PC2 Bernoulli Box host adapter, which will work at
up to 16 Mhz;
Added a DTC 7180 1:1 interleave hard disk controller and a Miniscribe
model M3085 hard disk (1/2 height, rotary voice coil actuator with auto-
park, 72.4 MB formatted, 7 heads, 1170 cylinders, 18ms average seek,
5ms track-to-track seek);
Replaced the Western Digital flex controller with a MicroSolutions
Compaticard I, and added a third, 1.44 MB flex drive (Teac);
Added a MicroSolutions MatchPoint board (short card allowing the 360K flex
drive to read/write/format Apple II disks using either Apple II format);
Added a Central Point Deluxe Option Board (short card allowing the 3.5"
flex drive to read/write/format Macintosh disks, as well as supporting
duplication of copy-protected disks, which is what it was originally
designed for, but not my purpose for installing it);
Upgraded my BIOS (AMI) to the latest version, providing full OS/2 support,
as well as expanded hard-disk and floppy support;
Added an 80287-10 math coprocessor;
Upgraded DOS from 3.10 to 3.30.
The IOmega PC2 adapter accomplished my desired goal of allowing me to run
the I/O bus at full speed, but I had to used a /f switch in the device
driver line in config.sys in order to work (this is documented only in a
read.me file on the distribution disk). This also involved switching the
board from DMA to programmed I/O mode for best performance on a fast 286
machine. It seems that a 286 string move instruction is *much* faster than
the brain-damaged Intel 8237 DMA; this apparently extends to the Chips and
Technologies knockoffs of the Intel architecture. There was a curious side-
effect to this change; even though the controller was running at a faster
bus speed than its predecessor (still using an 8-bit port, however), disks
that I had been using at interleave 2 became extremely slow, and the
solution was to change back to interleave 4; apparently the new adapters
cannot handle interleave factors that the older ones could (progress!).
However, since the Bernoulli is being relegated to a backup device, I am
not terribly concerned about this; the performance at interleave 4 is still
quite acceptable.
The Compaticard seems to be every bit as good as people say it is. However,
there are some limitations that are not widely known. First of all, it is
not a direct replacement for an AT flex controller, only a PC controller;
this means that even if an AT-style machine has BIOS support for a 1.44 MB
drive, it will still need the CCDRIVER.SYS driver to control such a drive
through a Compaticard. With DOS 3.3, CCDRIVER.SYS automatically invokes
DOS's DRIVER.SYS (assuming it is present in the root directory of the boot
disk), allowing the DOS format, diskcopy, etc. programs to work on the
affected drives. Otherwise, you must use the provided CCFORMAT program to
format, and there is no equivalent for diskcopy. With these limitations,
other versions of DOS > 2.0 will work. I also discovered that their cheaper
Compaticard II, which only supports 2 internal drives, has a couple of
other limitations. First of all, the Compaticard I has both edge and header
connectors for an internal cable for the first two drives; the Compaticard
II has only the header (common in AT-style machines), but not the edge
connector (common in PC/XT-style machines). Since Compaticards do not
normally come with cables, you must order a new cable if you want to replace
a primary controller in a PC/XT with a Compaticard II. Also, the Compati-
card II does not support single-speed (300 rpm) 1.2MB drives, which happen
to be the more common flavor; it only supports the dual speed (300/360)
variety; Compaticard I supports either. Finally, the Compaticard I supports
single-density 5.25" media, while the Compaticard II does not. This is an
issue if you want to use MicroSolutions' Uniform driver, which reprograms
the controller to read/write/format CPM etc. diskettes, to handle single-
density CPM formats. Note that with both boards, a 360K drive must be the
drive A if the Compaticard is the primary controller, because other types
of drive must have the driver loaded in order to work (unless you know you
will NEVER want to boot off of flex). The driver does allow you to specify
two drive letters for one drive, to support same-drive copying.
The MatchPoint and Deluxe Option boards seem to work pretty much as adver-
tised. Note that both of these boards require the affected drive to be
either A or B, so that constrained me to assignment of A=360K and B=1.44M.
I had to tell SETUP that B was a 720K for the sake of the Deluxe Option
Board, but it cannot be accessed as such by DOS, because the Compaticard
device driver assigns letters which follow any hard disk that may be
present; in my case, drive G=1.44M and H=1.2M.
As you may have gathered from the list of specs above, my hard disk setup
really hums (or rather, buzzes; the Miniscribe head actuator is kind of
loud, and at 5ms track-to-track seeking continuously, that is pretty much
what it sounds like). My main problem here was finding out the hard way
about DOS's 1024 cylinder limit. In the first place, I ordered the drive
thinking it was an 8-head, 1024 cylinder drive, like the the M6085. Wrong.
Then I discovered that the closest I could come to an appropriate BIOS
drive type was type 19 (7 heads, 1024 cylinders), yielding a capacity of
about 62 MB, wasting 10 MB of drive capacity. I then started playing with
OnTrack Drive Manager, which was included with the drive for support of such
disks under pre-3.3 versions of DOS, and found that it could be used to
partition and control the drive, via a device driver, to get the full
capacity out of it. There was just one catch; the resulting combination
might corrupt your disk if you use certain, unspecified, programs which
access the disk in too direct a manner. Specifically, they told me not to
try to use any type of cache program on the market with Drive Manager-
controlled drive if I valued my data. Also, versions of Windows prior to
the current one could cause problems. Plus of course, the required device
driver takes up precious base RAM. So, I partitioned with DOS 3.3 fdisk,
set up drive type 19, and am currently running two partitions totaling
62 MB for the time being. I am seriously considering getting Golden Bow's
vFeature, which will reportedly allow me to format the drive as a single
72 MB partition, while keeping the cluster size at 2K (this is supposedly
done by treating the sector size as 2K rather than 512 bytes, with 1 rather
than 4 sectors per cluster). Note that this requires the more expensive
vFeature Deluxe, rather than the cheaper standard edition. This driver
reportedly is compatible with most caches, including the one I use (Multi-
soft's Super PC-Kwik), as well as with Norton Utilities, etc. (this infor-
mation comes from both Golden Bow and Norton/Multisoft, so it sounds like
it should be accurate). Golden Bow has a 30-day unconditional money-back
offer, so I will feel more comfortable about trying it out if I decide to
go that route. Also, Multisoft is reportedly working on a fix to Super
PC-Kwik to correct compatibility problems with OnTrack and SpeedStor, to
be ready shortly. Also, with respect to Norton, don't try the Norton Disk
Doctor from version 4.5 Advance Edition on any drive controlled by any
third-party partitioning software without getting the fixed version from
Norton (all registered users should have received notification from Norton
by now). It WILL trash your disk, which could ruin your whole day.
One further note related to the hard disk. I have the Mace hFormat package,
which allows testing for back tracks, and for optimum interleave, as well
as track level formatting and sector copying. They recommend that the DOS
partition start at head 1 (not 0) of cylinder 0, but DOS 3.3 FDISK won't
allow you to do this. The advantage is that the partition table occupies
the first track by itself. If that track is trashed, hFormat can be used
to format only that track, and then copy a saved copy of that track back
to the disk, allowing recovery from an otherwise very disastrous event.
If you use this package, or another one with similar capabilities, this is
another possible reason to go with a third-party disk management program
rather than DOS 3.3 (I don't know about 4.01). By the way, their perfor-
mance analysis routines tell me I am getting a peak transfer rate of about
506 Kb/sec at 1:1 interleave. Outstanding.
In summary, I now have a wide variety of drives supported on my system,
which are working wonderfully well, although it did take some doing to get
there, and a significant amount of time spend on the phone. There are some
booby-traps you will run into if you are not careful, and you may even
stumble into one even if you are careful. But by and large, the products
I have discussed above are very good, fairly well documented, and the
manufacturers provide good, competent technical support. By the way,
almost all of this stuff was purchased by mail order, from a variety of
places, and I was treated very well by almost all of them, most of which I
hadn't done business with before, or even heard of. Maybe I just got lucky.
There was one exception among the mail-order companies: Blue Dolphin in
Minnesota. I ordered 20 high-density 3.5" diskettes from them for $2.90
each, from an ad in Computer Shopper (100% satisfaction guaranteed, free
shipping no less). When I got them, it was painfully obvious that I had
been had. What I got was 1MB diskettes that had the media sense hole
drilled in them; the hole was round, not square like on a real high-density
disk. Also, on closer inspection, I could see plastic shards around the
perimeter of the hole, between the front and back of the shell. About half
of these disks formatted with anywhere between 1 and 50 bad sectors. I
did not have a warm, fuzzy feeling about using these for their intended
purpose, long term archiving of customer files. On the plus side, when I
called to complain about this, they told me to return them and my money
would be cheerfully refunded. I do plan to protect myself by not paying
that portion of my MasterCard bill until I get either a check, or a credit
posted. By the way, I recently received a catalog from Elek-Tek, a
reputable company, advertising their house-label diskettes (made by Dysan,
100% surface tested, incuding the area *between* tracks!). Their price
for high-density 3.5" diskettes from this line is about the same as the
above. I am planning to try some out. In general, it is a bad idea to
risk important data on low density media formatted as high-density. It is
more than a matter of defect density; the high-density media has more
sensitive magnetic properties in general, by design, and use of low density
media in high-density mode is asking for trouble in the long run. Its sort
of like running a 16 Mhz processor at 20 Mhz because it seems to work, and
then wondering why the system crashes so much when it gets warm. Besides,
if I had been willing to risk pushing cheap diskettes like that, I would
have bought them cheap; $2.90 is too much to pay for 1MB disks when I can
buy name-brand (but unlabeled) disks for less than a dollar and drill the
hole yourself.
I was especially impressed with Hard Drives International (Tempe, AZ, I
think); they are very helpful technically, their prices can't be beaten
($649 for a M3085), they have a 30-day unconditional money-back guarantee,
and they provide detailed installation and technical documentation. This
is definitely a first-class operation, and I wouldn't hesitate to recom-
mend them to anyone. They advertise in Computer Shopper (4 or so ads per
issue). And, they have toll free phone numbers.
I hope this treatise has been helpful.--
Mike Lawless, NCR E&M Wichita, Box 20 (316) 636-8666 (NCR: 654-8666)
3718 N. Rock Road, Wichita, KS 67226 Mike.Lawless@Wichita.NCR.COM
{ece-csc,hubcap,gould,rtech}!ncrcae!ncrwic!Mike.Lawless
{sdcsvax,cbatt,dcdwest,nosc.ARPA}!ncr-sd!ncrwic!Mike.Lawless