 VOICE Home Page: http://www.os2voice.org |
[Previous Page] [Next Page] [Features Index] |
|
VOICE Home Page: http://www.os2voice.org |
[Previous Page] [Next Page] [Features Index] |
By Andrew
Stephenson ©December 2000Adaptec Ultra160 SCSI Family: http://www.adaptec.com/products/guide/ultra160scsi.html |
My old HDD configuration was simple enough. About five years ago I switched to
SCSI and had progressed to using an Adaptec 2940UW adapter card whose 68-pin port
drove a lone Quantum Viking 2 HDD. This gave what I regarded as a decent but unmeasured
performance with my Cyrus 6x86 133 MHz CPU. Yet I had already begun to think of
upgrading mother board, RAM, display etc. As OS/2 users often do, I improve my hardware
by rolling replacement, spreading costs and multiplying the fun. Speed fever was
in me.
Hence the HDD upgrade had to suit the envisaged enhanced machine. This induced
me to investigate the Ultra 160 standard, or U160+, whose transfer rate is intended
to hover around 160 MHz.
Over the years, depending as I do on my computer in my business of fiction writer
and occasional programmer, I have become wary of false economy. Ten pounds saved
now can cost hundreds later. SCSI and quality do not come cheap, but I bit the bullet
and went for an IBM Ultra 160 9.1 Gb HDD.
The IBM HDD arrived. Helpfully, IBM had marked it with their HDD data web site
so I went over and looted whatever seemed relevant. Scanning the specs, I saw that
Smart Alec had removed the cable termination option. Hmm... Phoned IBM Greenock,
Scotland, and netted the news that Alec was not guilty: Low Voltage Differential
signalling standard for 68-pin SCSI cables means termination is missing from all
U160+ drives and must be provided by a unit on the end of the cable. Oh, brave new
world.
Okay, need a terminator. An active one. Brain now off-line, I tried a UK mail
order supplier, who declared a willingness and ability to supply one, for around
GBP 15. Ordering, I sat back to wait for arrival.
Most of a week passed. Then my brain rebooted and I realized it was probably
not much use to buy a U160+ HDD and waste its speed by mixing it with a bodged mix
of 2940UW and Brand X terminator.
Not a problem. My VISA credit card would cringe but I could change my so far
unfulfilled order for the terminator to one for a suitable SCSI adapter. And here
I made my second mistake.
Adaptec UK was very helpful. Oh yes, the Nice Lady assured me, they do use a
range of adaptors. One, she declared, the 19160, was the same card as the 29160N.
Either would be suitable. But the 19160 was much cheaper. Aimed at desktop machines.
Briefly she mumbled something about Windows, which I really should have paid more
attention to.
Let's fast-forward... Alas, the Nice Lady was so wrong-o. When the card arrived,
I fitted it then ran the modified OS/2 install disk set, now loaded with a U160
Family driver obtained from the Adaptec web site (http://www.adaptec.com
-- search for "29160N", follow the link to drivers, then the one for the
29160N, then go down to the OS/2 section).
It is amazing how confused a computer can become, if a mere SCSI driver is missing.
The AICU160.ADD comprehensively snubbed the 19160 card: "Not there, not found,
not no how mate." Everything fell over, OS/2's installer begging plaintively
and repeatedly for a CD-ROM. Unable to take this pathetic display, I pulled the
plug. And pondered.
After a few phone calls, three-cornered between me, Adaptec UK and the mail order
supplier (who had actually done a sterling job in delivering the wrong SCSI card
so speedily and accurately), it was agreed that Adaptec would take back for a refund
the card I was misled into buying and we would all pretend nothing sad had happened.
An order was then put in for a 29160N.
Time out, Gentle Reader, for the First Dire Warning: if you are running OS/2,
DO NOT buy an Adaptec 19160 card. This seems to be Adaptec's take on the "WinThing"
idea. Cheaper it may be; but it is of the M$ world, body and soul. Adaptec's OS/2
driver does NOT make it work. I doubt even exorcism would.
And, whilst we are sitting around, chugging tea and noshing tiny cucumber sandwiches,
let us extend the time out with the Second Dire Warning: DO NOT confuse the "29160N"
and the "29160". Both are U160+ adapters but the "*N" is the
one for vanilla machines. The one without the "N" is a more complex beast
in the same card family, aimed at macho server builders, with an additional cable
socket. And a higher price.
One delivery deadline after another slipped by. It seems I have excellent taste
in SCSI adapters. Every computer in West Europe needed the same model. Or something.
The 29160N is
perhaps mildly more ambitious than most users would need, but only to the extent
that its 68-pin cable can cope with four devices, rather than the commonplace two.
(I do not include the SCSI card end of the cable in those figures.)
Obviously what I say here applies to the PC version; there is one for Macs too.
Physically the 29160N resembles other modern SCSI cards in size and shape and mounts
as usual.
A 50-pin socket for "legacy devices" (Adaptec's terminology) sits on
the outside of the endplate, for external cabling. This links to an internal 50-pin
connector, for which a familiar grey cable is supplied. An oddity of this cable
is that, the way it plugs in, it is directed down towards the PCI slot, so it is
necessary to double the ribbon hard back on itself if the card is crowded against
its neighbour.
A second connector is for internal 68-pin cabling. Here, there are obvious differences.
The newer, high frequency, signalling at low voltages with differential drive of
paired wires (LVD), places severe demands on mass production techniques. To reduce
cross-talk and interference, whilst keeping things simple, the new cabling is a
weird mix of normal bonded ribbon and twisted pairs, in gorgeous shades of white
and salmon pink. On the end remote from the SCSI card, a credit-card-sized terminator
board hangs free, although in practice this is supported well by the nearby plug
if this is used to connect to a device on the chain.
There is also the usual connector for the "HDD activity" light.
One nice improvement over the 2940UW is that there is no talk in the manuals
of limits on how one employs connectors. As before, the card can be told to decide
whether to provide termination, so it appears one plugs in, or not, at will.
After a week, the card arrived. Ignoring all the WinJunk disks and such, but
reading the paperwork, and afflicted by a sense of deja vu, I repeated the
installation process. And was amazed when it worked -- after I had disconnected
my Yamaha 4416S CD-RW unit, to which the Adaptec card took exception. (The mystery
as yet remains unsolved. The mail order supplier's technician said he had heard
of parity conflicts. And I have heard of fairies. in other words, it could be anything,
including damage by me during swapping in and out. I am not suspicious of the SCSI
card yet.)
At start up, one gets the usual SCSI BIOS messages. Hitting ^A
(Cntrl + A) as suggested leads into the familiar
set of Adaptec BIOS screens. Continuing the boot-up shows a useful improvement,
in that now the sign-on messages from the AICU160.ADD driver lists the actual
speeds the card was able to negotiate with the individual devices.
Another time out: Adaptec wisely repeats, again and again, not to mix slower
devices with fast ones on a cable, if possible. Doing so drags all devices on that
cable down to the speed of the slowest. No developmental magic has yet fixed that.
You should also not mix LVD and non-LVD devices on a U160 cable. It won't break
anything but performance will be hit.
In the situation where one's machine lies in pieces, when one is upgrading the
very system which gives access to those CDs, paper versions have so much to recommend
them. Why should we be forced to print out our manuals in advance, when the manufacturer
can do it better and cheaper?
The 19160 package, by contrast, had all the right bits in sensible quantities.
Maybe Windows users need special coddling?
And yet another strangeness...
Adaptec has very kindly gone to the trouble of providing decent OS/2 drivers
on their Website, yet do not promote their existence in the general lists of systems
supported. All right, this could be due to the recent date of the OS/2 drivers (June
2000) and is forgivable. However, failure to add them to the supplied floppy disk
set is a puzzler indeed, for our hypothetical, potentially computerless, user then
must contrive to visit Adaptec's
Website -- an occasion when the right friends can be a blessing.
It is not always possible to know in advance what one will need. Through world-weary,
OS/2-user, cynicism, I had anticipated this shortage and downloaded the drivers
in advance. That should not be necessary.
If U160+ is indeed the coming standard, the future looks bright. The technology
should become sufficiently common and cheap to be applied to all SCSI devices, so
that one may yet have a need for those four 68-pin LVD sockets. Thus far, I am pleased.
Compared with my previous SCSI-2 HDD configuration, U160+ scorches along.