Usenet 1993 July/InfoMagic USENET CD-ROM July 1993.ISO/answers/pc-unix/hardware

Internet Message Format | 1993-06-11 | 97.4 KB

From: [e-mail] (Eric S. Raymond)
Path: senator-bedfellow.mit.edu!enterpoop.mit.edu!gatech!swrinde!elroy.jpl.nasa.gov!usc!sol.ctr.columbia.edu!news.kei.com!ub!dsinc!gvls1!snark!esr
Newsgroups: comp.unix.sys5.r4,comp.unix.pc-clone.32bit,comp.sys.intel,news.answers
Subject: PC-Clone UNIX Hardware Buyer's Guide
Message-ID: <1lkjJL#188rWs4PZ2q41pM6OB9m30Hf=[e-mail]>
Date: 10 Jun 93 16:07:16 GMT
Expires: 9 Aug 93 23:00:00 GMT
Sender: [e-mail] (Eric S. Raymond)
Followup-To: comp.unix.pc-clone.32bit
Distribution: world
Summary: Tips on how and where to buy hardware for your UNIX.
Approved: [e-mail]
Lines: 1876
Xref: senator-bedfellow.mit.edu comp.unix.sys5.r4:3536 comp.unix.pc-clone.32bit:4918 comp.sys.intel:7557 news.answers:9263

Archive-name: pc-unix/hardware
Last-update: Thu Jun 10 11:55:22 1993
Version: 15.0

Many FAQs, including this one, are available via FTP on the archive site
rtfm.mit.edu (alias rtfm.mit.edu or 18.172.1.27) in the directory
pub/usenet/news.answers.  The name under which this FAQ is archived appears in
the Archive-name line above.  This FAQ is updated monthly; if you want the
latest version, please query the archive rather than emailing the overworked
maintainer.

What's new in this issue:
   * Some new product gotchas.
   * More about fans and quelling fan noise.

Gentle Reader: if you end up buying something based on information from this
Guide, please do yourself and the net a favor; make a point of telling the
vendor "Eric's FAQ sent me" or some equivalent.  The idea isn't to hype me
personally, I've already got all the notoriety I need from doing things like
_The_New_Hacker's_Dictionary_ --- but if we can show vendors that the Guide
influences a lot of purchasing decisions, I can be a more powerful advocate for
the net's interests, and for you.

0. Contents
I.    Introduction
II.   Overview of the Market
III.  Buying the basics
    A. Getting Down to Cases
    B. Power Supplies and Fans
    C. The Heart Of The Machine
    D. Motherboards and BIOSes
    E. Memory
    F. Peripherals
    G. Monitors and Video Cards
    H. Keyboards
    I. Power Protection
    J. Radio Frequency Interference
IV.   Performance tuning
    A. How To Pick Your Processor
    B. Of Memory In...
    C. Cache Flow
    D. Bus Wars
    E. IDE vs. SCSI
    F. Other Disk Decisions
    G. Souping Up X Performance
V.    Tape Drive Follies
VI.   Of Mice and Machines
VII.  Special considerations when buying laptops
VIII. When, Where and How to Buy
IX.   Questions You Should Always Ask Your Vendor
    A. Minimum Warranty Provisions
    B. Documentation
    C. A System Quality Checklist
X.    Things to Check when Buying Mail-Order
    A. Tricks and Traps in Mail-Order Warranties
    B. Special Questions to Ask Mail-Order Vendors Before Buying
    C. Payment Method
XI.   Which Clone Vendors to Talk To

I. Introduction

The purpose of this posting is to give you the background information you need
to be a savvy buyer of 386/486 hardware for running UNIX.  It is aimed
especially at hackers and others with the technical skills and confidence to go
to the mail-order channel, but contains plenty of useful advice for people
buying store-front retail.  It was formerly part of 386-buyers-faq issues 1.0
through 4.0, and is still best read in conjunction with the pc-unix/software
FAQ descended from that posting.

This document is maintained and periodically updated as a service to the net by
Eric S.  Raymond <[e-mail]>, who began it for the very best
self-interested reason that he was in the market and didn't believe in plonking
down several grand without doing his homework first (no, I don't get paid for
this, though I have had a bunch of free software and hardware dumped on me as a
result of it!).  Corrections, updates, and all pertinent information are
welcomed at that address.  The editorial `we' reflects the generous
contributions of many savvy USENETters.

This posting is periodically broadcast to the USENET group comp.unix.sysv386
and to a list of vendor addresses.  If you are a vendor representative, please
check to make sure any information pertaining your company is current and
correct.  If it is not, please email me a correction ASAP.  If you are a
hardware-knowledgeable user, please send me any distillation of your
experience that you think might improve this posting.

II. Overview of the Market

The central fact about 386/486 clone hardware that conditions every aspect of
buying it is this: more than anywhere else in the industry, de-facto hardware
standards have created a commodity market with low entry barriers, lots of
competitive pressure, and volume high enough to amortize a *lot* of development
on the cheap.

The result is that this hardware gives you lots of bang-per-buck, and it's
getting both cheaper and better all the time.  Furthermore, margins are thin
enough that vendors have to be lean, hungry, and *very* responsive to the
market to survive.  You can take advantage of this, but it does mean that much
of the info in the rest of this document will be stale in three months and
completely obsolete in six.

One good general piece of advice is that you should avoid both the highest-end
new-technology systems (those not yet shipping in volume) and the very cheapest
systems put out by vendors competing primarily on price.  The problem with
the high end is that it usually carries a hefty "prestige" price premium, and
may be a bit less reliable on average because the technology hasn't been
through a lot of test/improve cycles.  The problem with the low end is that
price-cutters sometimes settle for marginal components.  UNIX is more
sensitive to hardware flakiness than DOS, so cut-price systems that might
deliver consistently for DOS lemmings can come around and bite you.  Use a
little care, and spend the $200-$300 to stay out of the basement.  It's worth
it.

The last point deserves a little amplification.  In the PC world, there's a lot
of "if it doesn't fail, it's OK".  It is common to ignore normal engineering
tolerances --- the allowances for variations in components, temperature,
voltage margins, and the like --- and to assume that anything which doesn't
fail outright must work.  Watch out!  For example, The ISA bus was originally
designed for 6 MHz.  IBM later updated that to 8 MHz, and that's as much of a
standard as there is, yet there are motherboards that will let you (try to!)
run it at 12 MHz --- 50% over spec.  Some cards are actually designed to work
at that speed with proper tolerances.  Others might work...or they might flake
out when they get warm.  Any systems vendor above the fly-by-night level is
going to shoot for a little more reliability than this, burning in systems and
(often) doing at least a token system test with some kind of UNIX (usually
XENIX).  Pay the few extra bucks it costs to deal with a more careful vendor.

III. Buying the basics

In this section, we cover things to look out for that are more or less
independent of price-performance tradeoffs, part of your minimum system
for running UNIX.

A. Getting Down to Cases

Cases are just bent metal.  It doesn't much matter who makes those, as long as
they're above an easy minimum quality (on some *really* cheap ones, cards fail
to line up nicely with the slots, drive bays don't align with the access
cutouts, or the motherboard is ill-supported and can ground out against the
chassis).  If you're fussy about RFI (Radio-Frequency Interference), it's worth
finding out whether the plastic parts of the case have conductive coating on
the inside; that will cut down emissions significantly, but a few cheap cases
omit it.

Should you buy a desktop or tower case?  Our advice is go with tower unless
you're building a no-expansions personal system and expect to be using the
floppies a lot.  Many vendors charge nothing extra for a tower case and the
absolute *maximum* premium I've seen is $100.  What you get for that is less
desktop clutter, more and bigger bays for expansion, and often (perhaps most
importantly) a beefed-up power-supply and fan.  Putting the box and its fan
under a table is good for maybe 5db off the effective noise level, too.
Airflow is also an issue; if the peripheral bays are less cramped, you get
better cooling.  Be prepared to buy extension cables for your keyboard and
monitor, though; vendors almost never include enough flex.

The airflow thing is a good argument for a full tower rather than the `baby
tower' cases some vendors offer.  However, baby towers are getting more
attractive as boards and devices shrink and more functions migrate onto the
motherboard.  A state of the art system, with all 3" disks, 200W power supply,
half-size motherboard, on-board IDE and 64meg of RAM sockets, and half-sized
expansion cards, will fit into a baby tower with ample room for expansion; and
the whole thing will fit *under* a desk and make less noise than a classic
tower.

For users with really heavy expansibility requirements, rackmount PC cases do
exist (ask prospective vendors).  Typically a rackmount case will have pretty
much the same functionality as an ordinary PC case.  But, you can then buy
drive racks (complete with power supply), etc. to expand into.  Also, you can
buy passive backplanes with up to 20 or so slots. You can either put a CPU card
in one of the slots, or connect it to an ordinary motherboard through one of
the slots.

B. Power Supplies and Fans

Power supplies can matter but quality is cheap; give preference to those with a
Underwriter's Laboratories rating.  There's some controversy over optimum
wattage level; on the one hand, you want enough wattage for expansion.  On the
other, big supplies are noisier, and if you draw *too little* current for the
rating the delivered voltage can become unstable.  And the expected wattage
load from peripherals is dropping steadily.  The big old 300-watt supplies that
were designed for running several full-height 5.25" floppies and hard disks are
overkill in these days of portable-ready lightweight 3.5" drives.  200 watts is
good enough these days, and the new breed of compact 200W supplies is quieter
to boot.

About that annoying fan noise, ask if the power-supply fan on a target system
has a variable speed motor with thermostatic control --- this will cut down on
noise tremendously.  If not: I have seen a rave about, but haven't used, a
thermostatic fan controller called "The Silencer".  This tiny device mounts
inside your power supply and connects to the fan's power leads.  It
automatically varies the fan motor speed to hold a 79 to 82F temperature.  The
basic model is: $49.95 + $4.95 (S&H).  For details, write:

Quiet Technology Inc.
500 Executive Center - Suite 3C
P.O. Box 18216
West Palm Beach, FL 33416
407-683-6200
1-800-SILENCE

Warning: installing this may void your warranty!

Also, be aware that a thermostatic sensor basically measures the temperature
*at the sensor* (typically within the power supply box) and makes sure there is
enough airflow to keep the power supply from overheating.  However, the sensor
does not know a thing about the temperature in certain hot spots likely to
develop in a PC case (CPU, between SIMMs, between drives mounted in vertically
adjacent bays).

This can be a problem, because in garden variety tower cases there often isn't
enough airflow to cool all components effectively even if a single is going at
full speed.  This is especially true if your computer has lots of add-on cards
or hard disks (not much airflow between cards or between drives). Note that the
fan in the power supply was basically designed to cool the power supply, not
the components in the case. Not providing additional fans is a case of cheap
engineering. On PCs with "expensive" engineering (e.g. HP Vectra, Compaq) one
will find one to two extra fans besides the one in the power supply.

So the bottom line is, use thermostatic controls if you can to cut noise.
But if you want high reliability, use two or more fans.

The noise produced by a fan is not just a function of the speed with which it
turns.  It also depends on the nature of the airflow produced by the fan blades
and the bearings of the rotor. If the blades causes lots of turbulent airflow,
the fan produces lots of noise. One brand of fans that, I'm told, has a
reputation for being much more silent than others even if going at full
throttle is the German manufacturer Pabst.  Their fans offered in US computer
magazines.

C. The Heart Of The Machine

Yet another basic decision, of course, is processor speed and type.  Until
recently, the hot sellers in this market were the 386/33DX and AMD 386/40DX,
which I'd say are reasonable minimum-speed engines for UNIX with X.  However,
recent pricing moves by Intel have moved the price of a 486SX25 below the
roughly equivalent 386DX33 chip.  The 386 is therefore effectively dead for new
hardware, and the 486SX/25 defined as the new low end (at least for the next
90 days or so).

At the system level, the current premium for 486 over 386 is about $150 as
many vendors move to phase out their 386 designs.  Unless you're buying a
portable, we definitely recommend going 486.

A 486DX/33 has enough power make a good personal UNIX box.  For UNIX, this is
your floor; how far above them you want to buy depends on your budget and job
mix.  We'll have much more to say about this in the section on performance
tuning.

D. Motherboards and BIOSes

Provided you exercise a little prudence and stay out of the price basement,
motherboards and BIOS chips don't vary much in quality either.  There are only
six or so major brands of motherboard inside all those cases and they're pretty
much interchangeable; brand premiums are low to nonexistent and cost is
strictly tied to maximum speed and bus type.  Unless you're buying from a
"name" outfit like Compaq, Dell, or AST that rolls its own motherboards and
BIOSes, there are only four major brands of BIOS chip (AMI, Phoenix, Mylex,
Award) and not much to choose between 'em but the look of the self-test
screens.  One advantage UNIX buyers have is that UNIXes are built not to rely
on the BIOS code (because it can't be used in protected mode without more pain
than than it's worth).  If your BIOS will boot properly, you're usually going
to be OK.

If the above sounds too rosy, there is a catch; it describes *current*
hardware, not some of the historical botches.  And it's hard to know how old
what you're buying is.  You might actually be buying a motherboard that's been
sitting on the dealer's back shelf for a year, with a BIOS chip in it that was
in the drawer for another year before he ever stuck it in the board.  And some
of those older BIOSes and board designs are to be desperately avoided.  There
have been quite a few bogus cache designs that either don't work at all
(instant panic under UNIX) or that severely degrade performance.  A lot of
earlier designs have bus timing problems that show up in bad interactions with
host adapters and fancy graphics boards.  Bad memory designs were also not
uncommon.

A good, tricky way to keep the vendor from shipping you these fossils is to
specify a motherboard that can take 4 or 16MB SIMMs (as opposed to just the
older 1MB kind).  You want to do this anyhow for functional reasons.

There are a few other potential gotchas to beware of, especially in the cheaper
off-brand boards.  One is "shadow RAM", a trick some boards use for speeding up
DOS by copying the ROM contents into RAM at startup.  It should be possible to
disable this.  Also, on a cacheing motherboard, you need to be able to disable
cacheing in the memory areas used by expansion cards.  Some cheap motherboards
fail to pass bus-mastering tests and so are useless for use with a good SCSI
interface; on others, the bus gets flaky when its turbo (high-speed) mode is
on.  Fortunately, these problems aren't common.

You can avoid both dangerously fossilized hardware and these little gotchas by
sticking with a system or motherboard design that's been tested with UNIX (some
help with that below).

Some other good features to look for in a motherboard include:

* Gold-plated contacts in the expansion slots and RAM sockets.  Base-metal
  contacts tend to grow an oxidation layer which can cause intermittent
  connection faults that look like bad RAM chips or boards.  (This is why, if
  your hardware starts flaking out, one of the first things to do is jiggle
  or remove the boards and reseat them, and press down on the RAM chips
  to reseat them as well -- this may break up the oxidation layer.  If
  this doesn't work, rubbing what contacts you can reach with a soft
  eraser is a good fast way to remove the oxidation film. Beware, some
  hard erasers, including many pencil erasers, can strip off the plating, too!)

* Ability to go to 64MB on the motherboard (that is, without plug-in
  daughterboards).  Most EISA boards seem to have this (the popular Mylex
  MAE486 board is an exception).  Note: many newer EISA boards have only
  8-sockets, but these handle 16MB SIMMs.

* The board should be speed-rated as high as your processor, of course.
  It's good if it's rated higher, so upgrade to a faster processor is
  just a matter of dropping in the chip and a new crystal.

Finally, beware the infamous FP exception bug!  Some motherboards fail to
handle floating point exceptions correctly; instead of generating a SIGFPE they
lock up. The following fragment of C code will reproduce the problem:

    double d;

    d = 0.0;
    d = 1.0 / d;    /* floating divide by zero should yield SIGFPE */

John R. Levine <[e-mail]> explains: "The difficulty stems
from the fact that there are two ways to handle floating exceptions on a 486,
the right way and the PC way.  What the 486 wants to do is to generate an
interrupt 16 when there is a floating point error, all entirely internal to the
CPU.  This has been the native way to handle floating point interrupts since
the 286/287.  The 286/287 and 386/387 each have a dedicated ERROR pin that the
FPU uses to tell the CPU that it's time for an error interrupt.

Unfortunately, the 8086/8087 handled interrupts differently.  The error pin on
the 8087 was wired to the 8259A interrupt controller, the same interrupt
controller that handled keyboard, disk, clock, etc.  interrupts.  The PC/AT
enshrined IRQ 13 as the one for floating interrupts.  (The details of this are
a little hazy to me, since the XT didn't have IRQ 13 tied to an 8259A, so the
AT must have at least changed the interrupt number.)  PC designs have generally
wired the 287 or 387 ERROR pin to the 8259A, not to the ERROR pin on the CPU,
or at best had some poorly documented way to switch between the two interrupt
methods.

In the interest of backward compatibility, the 486 has a mode bit that says not
to handle FP exceptions automatically, but rather to freeze the FPU and send a
signal on the FERR pin, which is usually tied to an 8259A which then feeds the
interrupt back as IRQ 13.  There is some extra complication involved here
because the FPU has to stay frozen until the interrupt is accepted so the CPU
can go back and look at the FPU's state.  Early 386/25 chips had a bug that
would sometimes freeze up on a floating point interrupt and you had to get a
kludge socket with a PAL that fixed the timing glitch that provoked the bug.

So as likely as not, the motherboard hardware that runs FERR out and back isn't
working correctly.  It's not surprising, few DOS users take floating point
seriously enough to notice whether the interrupts are working right."

When you specify a system, make clear to your vendor that the motherboard must
handle float exceptions properly.  Test your motherboard's handling of
divide-by-zero; if it doesn't work, press your vendor to replace it *and
send me email*!  Only by publishing a list of boards known bad can we
protect ourselves and pressure vendors to fix this problem.

Norbert Juffa <[e-mail]> adds: Actually, the IBM PC,PC/XT and
most compatible use the NMI (non-maskable interrupt) to report coprocessor
errors. They don't go through the interrupt controller. Only a few not quite
compatible machines did use the 8259 PIC and one needed special startup code
for Microsoft-C for example to ensure correct handling of coprocessor
interrupts in programs. The PC/AT and compatibles do use the 8259, and the
coprocessor interrupt comes in as INT 75h (IRQs from second [slave] 8259 are
mapped to INT 70h-77h) to the CPU.  On the PC/XT it comes in as INT 2
(NMI). The problem with using the NMI was that NMI is also used for other
purposes (e.g. parity error reporting) and that the service routine has to
figure out what really caused the interrupt. The reason not to use the 8259 on
the PC might have been that not enough IRQs were available. The AT has two
cascaded 8259 chips and therefore has more IRQs available.

The 386 UNIX Buyer's Guide posting (pc-unix/software) includes tables of
motherboards and systems known to run with various UNIX ports.

David E. Wexelblat <[e-mail]> reports that, as of early 1993, Mylex EISA
motherboards have some serious compatibility problems (including with some
of Mylex's own peripheral cards) and should be avoided.

E. Memory

As of January 1993, standard DRAM parts are quite reliable at 60ns cycle speed.
With a decent cache, this is fast enough for any processor speed below 50MHz.
You can skip the rest of this section unless you've got your heart set on a
full 50MHz machine --- but be sure to read the `Cache Flow' section later on.

However, memory sufficiently fast and reliable for 486/50DX systems running
UNIX seems to be a particular problem.  The following war story by one
comp.unix sysv386 regular is typical: "Dell 2.2 ran perfectly on 3 different
AMI 486/50 EISA boards.  That is, after I replaced faulty memory chips which
caused repeated panics.  My conclusion, after consulting with our hardware
suppliers, was that current quality control on top-end memory chips (NEC,
Toshiba) is not good enough for 486/50s running serious Unix.  The memory will
pass every DOS-based test.  One has to plug and play to get a set of simms that
work reliably.  Part of the hazerds of leading edge technology."

F. Peripherals

Peripherals are another matter, especially hard disks.  A good rule of thumb
for balanced configurations is that the hard disk should comprise about half
(or maybe a bit more) of the total system hardware price (exception: if you're
buying a really good monitor, like 16" or over, it's going to be expensive
enough to bust this rule).  Unless you're the exception who has to invoke
warranty due to a system arriving dead, most of what you buy from a dealer or
mail-order house is their ability to surf the Winchester market, make volume
buys, and burn in your disks before shipping.  We'll look at disk choices in
more detail later on.

You'll need a monitor, of course.  The next section goes into monitor
options in detail.

You should have a tape drive for backup.  Unfortunately, the tape drive market
is rather confusing.  Rather than try to give a capsule summary, we give it
its own section below.

We'll have much more to say about price/performance tradeoffs in peripherals
in the next major section, on performance tuning.

G. Monitors and Video Cards.

A quick review of monitor standars for ISA machines:

                                 Horizontal   Vertical
Name         Resolution  Colors  Frequency   Frequency   Notes
-----------------------------------------------------------------------------
MDA           720x350            18.43 KHz     50 Hz

CGA           640x200      2     15.85 KHz     60 Hz     Obsolete
              320x200      4

EGA           640x350     16     21.80 KHz     60 Hz     Obsolete

VGA           640x480     16     31.50 KHz     60 Hz
              320x200    256

VESA VGA      640x480     16     38.86 KHz     72 Hz
              320x200    256

VESA SVGA     800x600     16     48.01 KHz     72 Hz
              640x480    256

8514/A        1024x768    16     35.20 KHz     43.5 Hz   Interlaced

XGA           1024x768   256     ??            ??        IBM proprietary

VESA 1024x768 1024x768   256     56.48 KHz     70 Hz

The Horizontal and Vertical Frequency columns refer to the monitor scan
frequencies.  The vertical frequency is the monitor's flicker rate; 60Hz
is minimal for ergonomic comfort, 72Hz is VESA-recommended, and 80Hz is
cutting-edge.

XGA is included for completeness, but is vanishingly rare in the clone market.

SVGA or `Super VGA' strictly refers only to 800x600 resolution, but is widely
used for 1024x768 and even 1280x1024 resolutions.  Standards above 1024x768
are weak and somewhat confused, largely because VESA's efforts have been going
into the forthcoming VDID standard for auto-configuring intelligent monitors.

These days, most vendors bundle a 14" monitor and super-VGA card with 1024x768
resolution in with their systems.  Details to watch are whether the card comes
loaded with 512K or 1MB of RAM (which will affect how much of that maximum
resolution and how many colors you actually get), whether the memory is
dual-ported VRAM (slightly more expensive but much faster), and whether the
monitor is interlaced or non-interlaced.  The latter is better and should no
longer cost extra; look for the abbreviation NI in the ad or quote and be
suspicious if you don't see it.

You should check ahead of time which Super VGA chipset the vendor normally
ships.  Though DOS/Windows doesn't really care, the UNIX software that uses it
(most notably X servers) will definitely notice the difference.  Most
implementations of X servers for Intel UNIX (especially those supplied by the
UNIX vendors) don't know how to use the SVGA modes of the cheapie Oak and
Trident SVGA chipsets, which are the ones most often bundled with systems. The
ATI VGAWonder chipset, while better supported, isn't usually bundled.  Cards
based on the Tseng 4000 chipsets are often bundled or available as as an
extra-cost opstion (usually less than $50 above Oak or Trident), perform
better, and are supported by the most implementations of Intel UNIX-based X
servers. The Tseng chipset is also the one best supported by XFree86.

Other things to check for:

* Dot pitch of 0.28 or smaller on a 12" or 13" monitor; 0.30 is acceptable on
  larger ones, especially 19" to 21" screens.  Dot pitch is the physical
  resolution of the screen's phosphor mask.  Larger dot pitches mean that small
  fonts and graphic detais will be fuzzy.

* 72Hz or better vertical scan frequency, to cut flicker.

* Non-interlaced display.  Interlacing cuts the required scan frequency for a
  given resolution in half, but makes flicker twice as bad.  As a result,
  interlaced monitors are rapidly disappearing; don't get stuck with one.

* Does it have a tilt-and-swivel base?  Adequate controls, including
  both horizontal and vertical size and horizontal and vertical centering?
  A linearity control, a trapezoidal control, and a color-temperature control
  are all pluses; the last is particularly important if you compose graphics
  on screen for hardcopy from a printer.

* Is it *color*?  Yes, if you don't see it in the ad, ask; some lowball
  outfits will try to palm off so-called "black & white VGA" monitors on you.

For X use, a 14", .28mm dot pit, non-interlaced 72Mhz monitor at 640x480
resolution is the bare minimum for comfortable use, and that resulution leaves
you rather s\queezed for screen real estate.  1024x768 is much better.  If your
budget will stand it at all, a 17" or 20" monitor is a good investment.  A
17" monitor is minimum if you're going to go with 1280x1024 resolution.

H. Keyboards

It's important to get a high quality keyboard with good key feel.  See the
typing-injury FAQ from sci.med.occupational to see what happens if you don't.
Carpal tunnel syndrome is no fun for anyone, but it hits hackers particularly
hard.  Don't be a victim!

Hal Snyder of Mark Williams, Co. <[e-mail]> sent us the following caveat:
"We find that about 10% of cheap no-name keyboards do not work in scan code set
3.  We are interested in scan code set 3 because only there can you reprogram
the keyboard on a per-key basis as to whether keys are make-only, make-break,
or autorepeat.  It is a big win for international support and for X."

He continues: "Keytronic, Cherry, and Honeywell keyboards, as well as a large
number of imports, work fine.  My advice is to either by a respected brand of
keyboard, or deal with a vendor who will allow you to return an incompatible
keyboard without charge."

I. Power Protection

Finally, I strongly recommend that you buy a power conditioner to protect your
hardware.  MOV-filtered power bars make nice fuses (they're cheap to replace),
but they're not enough.  I've been delighted with my TrippLite 1200, which you
can get for $139 or so by mail order.  A fringe benefit of this little beauty
is that if you accidentally pull your plug out of the wall you may find you
actually have time to re-connect it before the machine notices!

The technical info in the remainder of this section is edited from material
supplied by David E. Wexelblat <[e-mail]>.

There are several levels of power protection available to the home computer
user.  I break this down into 4 levels; others may have different ways of
classifying things.  The levels are:

    1. Surge Suppressor
    2. Line Conditioners
    3. Standby Power Supplies
    4. Uninterruptible Power Supplies

and here's what they mean:

1. Surge suppressors

These are basically a fancy fuse between the source and your hardware; they
clamp down spikes, but can't fill in a low voltage level or dropout.

This is a bare minimum level of protection that any piece of expensive
electronics should have.  Note that this applies to more than just AC power;
surge suppressors are available for (and should be used on) phone lines, and
RS-232 and parallel connections (for use on long lines; generally not needed if
the devices is colocated with the computer and all devices are protected from
outside sources).  Note also that *all* devices connected to your computer need
to be protected; if you put a surge suppressor on your computer but not your
printer, then a zap on the printer may take out the computer, too.

An important fact about surge suppressors is that *they need to be replaced if
they absorb a large surge*.  Besides fuses, most suppressors rely on on
components called Metal-Oxide Varistors (or MOVs) for spike suppression, which
degrade when they take a voltage hit.  The problem with cheap suppressors is
that they don't tell you when the MOV is cooked, so you can end up with no
spike protection and a false sense of security --- better ones have an
indicator.

You can buy surge suppressors at any Radio Shack; for better prices, go
mail-order through Computer Shopper or some similar magazine.  All of
these are low-cost devices ($10-50).

2. Line Conditioners

These devices filter noise out of AC lines.  Noise can degrade your power
supply and cause it to fail prematurely.  They also protect against short
voltage dropouts and include surge suppression.

My Tripp-Lite 1200 is typical of the better class of line conditioners --- a
box with a good big soft-iron transformer and a couple of moby capacitors in it
and *no* conductive path between the in and out sides.  With one of these, you
can laugh at brownouts and electrical storms.

Price vary widely, from $40-400, depending on the power rating and capabilities
of the device.  Mail-order from a reputable supply house is your best bet.
Line conditioners typically *don't* need to be replaced after a surge; check
to see if yours includes MOVs.

3. Standby power supplies (SPSs)

These devices are battery-based emergency power supplies that provide power for
your system via an inverter if the power fails. An SPS will generally have all
the capabilities of a line conditioner as well.

Note: these devices do not come on line until after the power fails, and have a
certain amount of delay of some of milliseconds before they come on line.  If
the capacitors in your power supply are not large enough, the SPS may not cut
in in time to prevent your computer from seeing the power failure.

Note also that many SPSs are marketed as Uninterruptable Power Supplies (see
below).  This is incorrect.  Any device with a non-zero cutover time cannot be
a true UPS.  If the ad mentions a cutover time, it's an SPS, and not a UPS.

The price range for these devices (depending largely on size and cutover time)
is $200-2000.  An SPS will *not* need to be replaced after absorbing a large
surge.

4. Uninterruptable power supplies (UPSs)

These devices provide full-time isolation from the incoming AC line through a
transformer of some sort.  These devices are on-line at all times, and if the
AC line fails, the batteries will cut in.  Your devices will see no
interruption of their incoming AC.  UPSs cost more, and provide more features.
They are the ultimate in power protection.  Many UPSs have an intelligent
interface that will notify a connected device of a power failure, allowing it
to shut down cleanly.  UPSs also provide the capabilities of a line
conditioner. The price range (for devices in the size range for a home
computer) are $400-$2500.  An UPS will *not* need to be replaced after
absorbing a large surge.

Now, given this information, how does one decide what to get?  For a system
that runs unattended, like most Unix systems, it is best to have a device that
provides both power holdover and a power failure signal.  Hence, for a Unix
system, a UPS or SPS with UNIX monitoring software id the best choice.  At
least one vendor sells ordered-shutdown software for Unix, and it's fairly
simple to write your own daemon to monitor a serial port, and send init a
SIGPWR signal when it sees a certain signal.  Tripp Lite and APC (the two
vendors with the lion's share of the market) have good UNIX monitoring
software.  The APC "Powerchute" software, for examples, allows you to monitor
the SPS's internal temperature, or perform a UPS self-test, from any UNIX
terminal!

Many UPS/SPS signal ports work by asserting a pin, so that one
could use a modem-control serial port on the PC and wire this pin to
"Carrier Detect" in order to monitor it. Some, like the APC "SmartUPS"
series, actually conduct a "dialog" with the host through a serial line
in order to accomplish the monitor functions.

Our recommendation for a home Unix environment is a configuration like the
following:

    a) An on-line UPS or SPS for the computer system.  An intelligent
       interface is mandatory, along with appropriate software for
       ordered shutdown.
    b) Surge suppression on all phone lines, and also on serial/parallel
       lines that leave the room.
    c) Line conditioners on any devices not connected to the UPS.  If
       you do take a power hit, it's cheaper to replace a $50 line
       conditioner than a $1500 laser printer.

An important question is "How do I know how big a UPS/SPS to get?"  The watt
rating of the UPS/SPS should be at least the sum of the peak ratings off all
equipment connected to it (don't forget the console monitor).  Power-supply
marketroids tend to quote you capacities and formulas like "sum of VA ratings +
20%" which (surprise!) push you towards costler hardware.  Ignore them.  If a
watt rating is not given, watts = 0.75*VAmax.

One other consideration is that you typically shouldn't put a laser printer on
a UPS --- toner heaters draw enough current to overload a UPS and cause a
shutdown within seconds.  The other thing is that you can't even put the laser
printer on the same circuit with a UPS --- the heater kicks on every 20-30
seconds, and most UPSs will see the current draw as a brownout.  So buy a
separate line conditioner for the laser printer.

Finally, read the UPS's installation manual carefully if you're going
to use it with other power-protection devices.  Some UPSs don't like having
surge suppressors between them and the equipment.

David personally recommends surge suppressors and line conditioners from
Tripp-Lite (available both mail-order and retail), and UPSs from Best Power
Technologies (Necedah, WI - 1-(800)-356-5737).  I can enthusiastically second
the TrippLite recommendation, but haven't dealt with Best Power at all.  Evan
Leibovich says "Add American Power Conversion to the list. They have paid good
attention to the UNIX market, and have (by far) the best UNIX UPS monitoring
scheme on the market.  They're also widely available."  There are many other
vendors for all of these devices.

Tripp-Lite has a whole range of products, from a $10 phone-line
surge-suppressor, to line conditioners and SPSs with prces in the hundreds of
dollars.  They have a line of $50-80 line conditioners that are good for most
peripherals (including your home stereo :->).

Best Power Technologies sells two lines of UPSs in the range for home systems.
The older and more expensive FERRUPS line (which is what David has) has a smart
interface, and very good filtering and surge-suppression capabilities.  He says
"I have a 1.15kVA FERRUPS for my home system, which is overkill with my current
hardware (although it rode out a 45 minute power failure with nary a whisper -
no reboot).  In 1990, I paid ~$1600 for this device, and that has since gone
up.  They also sell a newer line of Fortress UPSs.  These are better suited in
price for home systems.  I don't know much about them, as they were not
available when I bought my UPS.  I expect that this is what most people will
want to consider, though.  In addition, Best sells Check-UPS, a software
package (in source form) for monitoring the UPS and shutting it down.  I have
found Best to be a good company to deal with, with competent, knowledgeable
sales people (who will be able to help you pick the right device), and helpful,
courteous, and responsive technical support."

Other things to know:

A UPS should be wired directly to (or plugged directly into) the AC supply
(i.e. a surge suppressor is neither required nor suggested between the wall and
the UPS).  In addition, a surge suppressor between the UPS and the equipment
connected to it is redundant and also unnecessary.

J. Radio Frequency Interference

(Thanks to Robert Corbett <[e-mail]> for contributing
much of this section)

Radio Frequency Interference (RFI) is a growing problem with PC-class machines.
Today's processor speeds (20-50MHz) are souch that the electromagnetic noise
generated by a PC's circuitry in normal operation can degrade or jam radio and
TV reception in the neighborhood.  Such noise is called Radio Frequency
Interference (RFI).  Computers, as transmitting devices, are regulated
by the Federal Communications Commission (FCC).

FCC regulations recognize two classes of computer:

If a PC is to be used in a home or apartment, it must be certified to be FCC
class B.  If it is not, neighbors have a legal right to prevent its use.  FCC
class A equipment is allowed in industrial environments.

Many systems are not FCC class B.  Some manufacturers build boxes that are
class B and then ship them with class A monitors or external disk drives.  Even
the cables can be a source of RFI.

It pays to be cautious.  For example, the Mag MX17F is FCC class B.  There are
less expensive versions of the MX17 that are not.  The Mag MX17 is a great
monitor (I wish I had one).  It would be painful to own one and not be allowed
to use it.

An upgradeable system poses special problems.  A system that is FCC class B
with a 33 MHz CPU might not be when the CPU is upgraded to a 50 or 66 MHz CPU.
Some upgrades require knockouts in the case to be removed.  If a knockout is
larger than whatever replaces it, RFI can leak out through the gap.  Grounded
metal shims can eliminate the leaks.

IV. Performance tuning

Here are the places where you can trade off spending against the performance
level you want to buy and your expected job mix.

A. How to Pick your Processor

The following information appeared in article <[e-mail]>
by [e-mail] (|S| Norbert Juffa).  It gives a good indication
of the relative speeds in Intel's processor line:

      UNIX performance of Intel processors as given in Intel's literature


   Processor       SPECmark SPECint SPECfp Whetstone Dhrystone Linpack   Ref Rm
                                           double p.     2.1   dp MFLOPS

1) Intel 386/387-33   4.3    6.4     3.3     3290     15888      N/A      1  *+
2) Intel 386/387-33   4.1    6.0     N/A     3200     18900      0.4      2  #
3) RapidCAD-33        6.6    7.3     6.1     5300     18275      N/A      1  *+
4) 486DX-25           8.7   13.3     6.6     5640     32000      1.0      2
5) 486DX-33          11.1   17.5     8.2     7200     43000      1.5      3
6) 486DX-33          12.1   18.3     9.2      N/A       N/A      N/A      4
7) 486DX-33          14.5   19.0    12.2    12300     43500      1.6      5  &
8) 486DX-50          18.2   27.9    13.6    10710     64400      2.5      3
9) 486DX2-50         19.2   25.4    15.9    18500     63966      2.3      5  &
10)486DX-50          21.9   28.5    18.3    18500     65400      2.4      5  &
11)486DX2-66         25.6   34.0    21.2    24700     85470      3.1      5  &

Remarks:

* Whetstone/Dhrystone are 32-bit DOS results
+ SPEC ratios recomputed from SPEC timings (computed wrong in report)
& note huge increase in SPEC floating point performance over previous results
  due to new experimental FORTRAN compiler
# machine with AMD 386-40/Cyrix 83D87-40/128 kB cache is estimated by me at:
  7.7 SPECint, 5.0 SPECfp, 6.1 SPECmark,
  5600 double prec. Whetstones, 23000 Dhrystones,
  0.6 Linpack double prec. MFlops
  These estimates based on my own measurements and data from:
  FasMath 83D87 Benchmark Report, Cyrix 1990
  World's Fastest 386 40 MHz Am386(tm)DX Microprocessor Performance Summary,
  AMD 1991

References:

1) Intel RapidCAD(tm) Engineering CoProcessor Performance Brief. 1992
2) i486(tm) Microprocessor Performance Report. 1990.
   Order No. 240734-001
3) 50MHz Intel486(tm) DX Microprocessor Performance Brief. 1991.
   Order No. 241120-001
4) i486(tm) Microprocessor Business Performance Brief. 1990.
   Order No. 281352-002
5) Intel486(tm) DX2 Microprocessor Performance Brief. 1992
   Order No. 241254-001

Configurations:

1) COMPAQ SystemPro 386/33 MHz, 8 MB memory, AT&T UNIX System V/386 Release 4.0
   Version 2.0
2) 64 kB write back cache,
   AT&T UNIX System V Release 3.2CC, MetaWare High C R2.2c,
   SVS FORTRAN V2.8
3) COMPAQ SystemPro 386/33 MHz, 8 MB memory, AT&T UNIX System V/386 Release 4.0
   Version 2.0
4) 128 kB write-back cache, 12 MB RAM,
   AT&T UNIX System V Release 3.2CC, MetaWare High C R2.2c,
   SVS FORTRAN V2.8
5) No 2nd level cache, 16 MB RAM,
   AT&T UNIX System V/386 R3.2, MetaWare High C R2.3p
   SVS FORTRAN V2.8
6) ALR PowerCache 33/4e, 128 kB cache, 16 MB RAM
   SCO UNIX System V R3.2.2, MetaWare High C R2.2c/R2.3k,
   SVS FORTRAN V 2.8
7) Intel Modular Platform, 256 kB write-back cache, 32 MB RAM,
   AT&T UNIX System V R4.0.4, Metaware High C R2.4b,
   Intel Scheduling FORTRAN 77 Compiler V0.2
8) 256 kB write-back cache (82495DX/82490DX), 16 MB RAM,
   AT&T UNIX System V/386 R3.2, MetaWare High C R2.3p
   SVS FORTRAN V2.8
9) Intel Modular Platform, 256 kB write-back cache, 32 MB RAM,
   AT&T UNIX System V R4.0.4, Metaware High C R2.4b,
   Intel Scheduling FORTRAN 77 Compiler V0.2
10)Intel Modular Platform, 256 kB write-back cache, 32 MB RAM,
   AT&T UNIX System V R4.0.4, Metaware High C R2.4b,
   Intel Scheduling FORTRAN 77 Compiler V0.2
11)Intel Modular Platform, 256 kB write-back cache, 32 MB RAM,
   AT&T UNIX System V R4.0.4, Metaware High C R2.4b,
   Intel Scheduling FORTRAN 77 Compiler V0.2

One of Intel's most recent wrinkles is the "clock-doubler" chips.  The 50DX2
runs at 25MHz externally but computes at 50MHz.  A 66DX2 (bus speed 33MHz) is
also shipping, and there are persistent rumors of a clock-doubled 50 in the
works that would compute at a blistering 100MHz!  Intel likes to claim a 70%
speedup for the doublers over their undoubled brethren.  I've expressed
skepticism about this in previous issues, but the SPECmarks above suggest that
just this once the marketroids may not be lying -- much.  Under UNIX, a 50DX2
is in fact nearly as fast as a true 50DX.  Still, beware of anyone whose
literature passes off the DX2 qualification in the fine print; they may be
scamming about other things, too.

Right now you'll still pay a premium for a 486/50, as that's relatively new
technology and demands extra-fast memory to run full-out.  Also, these
processors run really hot (one correspondent described the 50 as a "toaster on
a chip").  If you go this route, be sure your configuration has an
extra-heavy-duty cooling fan.  Or two.  And, for preference, a hefty heat
sink.  On current trends, a 66DX2 is probably the better way to go.

B. Of Memory In...

Buy lots of RAM, it's the cheapest way to improve real performance on any
virtual-memory system.  At $30-$50 maximum per megabyte it's just plain silly
to stick with the 2-4mb now standard on most clone configurations.  Go to 8,
you won't regret it; 16 if you're going to use X.

William Davidsen <[e-mail]> writes: "There are two places where
memory addition will show an improvement [under sar(1)], in %wio and in avwait
(sar -d) on individual devices.  Note that you may have to tune kernel params,
some systems have a limit of 600k on i/o buffers.  Also, you can hurt
performance on V.3 systems with way too many buffers (like 4+MB) if you have a
slow CPU."  Add memory until avwait stops dropping like a rock.

Above 16 is iffy on ISA boxes because the stock USL 4.0.3 kernel may try to do
DMA from a location the bus can't deal with.  Most UNIX vendors have fixed this
by adding code that forces DMAs to take place from low memory; make absolutely
sure that includes yours before you load up beyond 16MB.  The pc-unix/software
FAQ posting includes information on which vendors are known to have fixed this
problem.

Some motherboards have 16 sockets for SIMM memory modules.  Some only 8.  Some
take only 1MB mdules, some handle 4MB.  These constraints interact in funny
ways.

You should make sure if you are buying an entry level 2 or 6 MB system with a
16-socket motherboard that you will not have to ditch the SIMMs that are
already installed in order to go to your maximum (if 16 MB is your maximum).
Some systems only allow you to mix 1M and 4M SIMMs in certain combinations.
Try not to get any 1M SIMMs in your initial configuration, because you'll
probably end up turfing them later.  That is, buy a 4MB, 8MB, 12 MB or 16MB
system to start.

Newer ISA designs have a 32 MB upper limit with only 8 sockets, since they can
take 4Mx9s...however, this means different interleaving (only 2 banks), which
limits the possible configurations.  You don't want to start off with an 8 MB
configuration, because that's 8 ea 1Mx9's, filling up all the sockets...the
next upgrade requires replacing 1Mx9 with 4Mx9.  You can't even set up 12
MB!...the first reasonable config (that won't require tossing hardware) is 16
MB, since that's one bank full of 4Mx9.

Most new EISA motherboards have 64MB capacity, either as 16 4MB-capable sockets
or as 8 16MB-capable sockets.

C. Cache Flow

The most obscure of the important factors in the performance of a UNIX 486
system is the motherboard's memory cache size and design.  The two questions
performance-minded buyers have to deal with are: (1) does the cache design
of a given motherboard work with UNIX, and (2) how much cache SRAM should
my system have?

Before normal clock speeds hit two digits in MHz, cache design wasn't a big
issue.  But DRAM's memory-cycle times just aren't fast enough to keep up with
today's processors.  Thus, your machine's memory controller caches memory
references in faster static RAM (SRAM), reading from main memory in chunks that
the board designer hopes will be large enough to keep the CPU continuously fed
under a typical job load.  If the cache system fails to work, the processor
will be slowed down to less than the memory's real access speed --- which,
given January 1993's typical 70ns DRAM parts, is about 7MHz.

The 486 includes an 8K cache right on the processor chip.  If memory accesses
were reliably sequential and well-localized, this would be fine.
Unfortunately, one side-effect of what's today considered "good programming
practice", with high-level languages using a lot of subroutine calls, is that
the program counter of a typical process hops around like crazy; locality is
really poor.  This gives the cacheing system a workout.  (UNIX makes the
problem worse, because clock interrupts and other effects of multitasking
design degrade locality still further).

Thus, the 486's 8K internal primary cache is typically supplemented with an
external caching system using SRAM to reduce the cost of an internal cache
miss; in January 1993, 20ns SRAM is typical.  The size and design of your
motherboard cache is one of the most critical factors in your system's real
performance.

Unfortunately, cache design is a complicated black art, and cache performance
isn't easy to predict or measure, especially under the rapidly variable
system loads characteristic of UNIX.  Thus, the best advice your humble editor
can give is a collection of rules of thumb.  Your mileage may vary...

Rule 1: Buy only motherboards that have been tested with UNIX
   One of DOS's many sins is that it licenses poor hardware design; it's too
brain-dead to stretch the cache system much.  Thus, bad cache designs that
will run DOS can completely hose UNIX, slowing the machine to a crawl or even
(in extreme cases) causing frequent random panics.  Make sure your motherboard
or system has been tested with some UNIX variant.

Rule 2: Be sure you get enough cache.
   If your motherboard offers multiple cache sizes, make sure you how much is
required to service the DRAM you plan to install.
   Bela Lubkin writes: "Excess RAM [over what your cache can support] is a very
bad idea: most designs prevent memory outside the external cache's cachable
range from being cached by the 486 internal cache either.  Code running from
this memory runs up to 11 times slower than code running out of fully cached
memory."

Rule 3: "Enough cache" is at least 64K per 16MB of DRAM
   Hardware caches are usually designed to achieve effective 0 wait state
status, rather than perform any significant buffering of data.  As a general
rule, 64Kb cache handles up to 16Mb memory; more is redundant.

Rule 4: If possible, max out the board's cache -- it wlll save hassles later
   Bela continues: "Get the largest cache size your motherboard supports, even
if you're not fully populating it with RAM.  The motherboard manufacturer buys
cache chips in quantity, knows how to install them correctly, and you won't end
up throwing out the small chips later when you upgrade your main RAM."

A lot of fast chips are held back by poor cache systems and slow memory.  The
50DX has a particular problem this way, because its cycle spead is as fast as
that of a 20ns cache SRAM.  To avoid trouble, cloners often insert wait states
at the cache, slowing down the 50DX to the effective speed of a 50DX/2.

Worse than this, a lot of cloners have taken the 50DX/2 and 66DX/2 as
invitations to reuse old 25- and 33MHz board designs without change.  The
trouble is that these chips take a double hit for each wait state, because
the wait states are timed by *external* cycles.  And there can be lots of
them; a look at the CMOS setup screen of most 33Mhz and 50MHz system will
usually reveal many wait states.

[The intro to cache design has been temporarily deleted while I straighten
out technical errors with several correspondents.]

D. Bus wars

There are three bus standards in the clone market; ISA (the original 16-bit
PC/AT bus), EISA (a 32-bit bus upward-compatible from ISA), and MicroChannel, a
proprietary IBM bus used in IBM's PS/2 and its few clones.  Your first
intelligent decision, however, is to forget MicroChannel's existence.
MicroChannel is technically sweet, but PS/2 clones are rare, expensive, and
doomed to stay that way by IBM's licensing terms.  The action is all in the
commoditized ISA and EISA market.

Should you buy ISA or EISA?  You'll pay up to a $300 premium for the latter.
What you get in return is the ability to use things like fast 32-bit SCSI
controllers and a smoother upward-migration path.  On the other hand, EISA
cards are significantly more expensive.  And so far, there isn't much support
for EISA-specific hardware --- a couple of vendors will drive EISA SCSI and
network controllers and that's about it.  All ISA cards will still work.

Actually, there doesn't *need* to be wide EISA support for EISA to be a win.
Disk and network controllers are the big bandwidth-eaters in most systems;
faster disk access, by itself, can easily be enough to justify the EISA
premium.  And this need not involve any software support for EISA's extensions
at all; it has been widely reported, for example, that the Adaptec 1740 EISA
disk controller is faster in 1542 emulation mode than a 1542 in an ISA box.
Many EISA board makers supply their own drivers, so this may not be apparent
when scanning vendors' compatibility guides.

There is also some convenience gain from doing software-based configuration of
cards rather than jumpers.  Unfortunately, you have to boot DOS to do that!

Of course, most of what you get from EISA is a performance boost.  There are
two different theories about why EISA is better; both have their adherents.

Theory A: Bandwidth matters

UNIX has always been an I/O-intensive operating system.  According to this
theory, increasing processor speed on clones can leave it spending all its time
waiting on the limited I/O capacity of the poor old 5.3MB/sec ISA bus.  The
vendors all seem to think this starts at around 33MHz and that if you're buying
50MHz it definitely pays to go EISA.

Theory B: Cache is what matters

According to this theory, UNIX never comes even close to saturating the ISA-bus
bandwidth.  EISA boards are faster because the premium vendors can charge for
them allows the motherboard designer more freedom and a richer parts budget.
The most important performance effect of this is that EISA boards have larger
and better-designed caches, increasing the effective memory-access speed.

There's probably some truth to both analyses.  If your machine is going to
spend most of its processor time running X displays and doing other classically
compute-bound tasks, cache size matters most.  On the other hand, benchmarks
show that the combination of TCP/IP and multi-user disk activity *can* saturate
ISA, and one can sometimes *see* a fast-processor machine slow down during disk
accesses...

If you're contemplating any kind of heavy-duty networking, EISA network
adapters will become rather important.  A correspondent tells me he's seen
benchmarks showing what percentage of bus bandwidth is consumed by various
cards when flooding an ethernet (i.e. consuming the entire 10Mbit bandwidth of
a quiet net, as you might be when doing an FTP transfer, for instance).  8-bit
ISA cards consume 40-60% of bus bandwidth; 16-bit cards, 20-40%.  32-bit EISA
cards consume only about 5-10%.  This would be particularly important in a
machine being used as a bridge, where you might be handling a large portion of
the traffic on two or more separate nets.  The advantage of EISA cards may be
due to their shorter-cycle bus mastering DMA.  At time of writing, only
SCO supports these, but other UNIX vendors are known to have their own drivers
in the pipeline.

E. IDE vs. SCSI (vs. ESDI!)

Another basic decision is IDE vs. SCSI.  Either kind of disk costs about the
same, but the premium for a SCSI card varies all over the lot, partly because
of price differences between ISA and EISA SCSI cards and especially because
many motherboard vendors bundle an IDE chip right on the system board.  SCSI
gives you better speed and throughput and loads the processor less, a win for
larger disks and an especially significant consideration in a multi-user
environment; also it's more expandable.

Another important win for SCSI is that it handles multiple devices much more
efficiently.  If you have two IDE (or ST506 or ESDI) drives, only one can
transfer between memory and disk at once.  In fact, you have to program them at
such a low level, one drive might actually be blocked from *seeking* while
you're talking to the other drive.  SCSI drives are mostly autonomous and can
do everything at once; and current SCSI drives are not quite fast enough to
flood more than 1/2 the SCSI bus bandwidth, so you can have at least two drives
on a single bus pumping full speed without using it up.  In reality, you don't
keep drives running full speed all the time, so you should be able to have 3-4
drives on a bus before you really start feeling bandwidth crunch.

All this having been said, don't write off IDE too quickly.  Sure, it's
compatible with the nasty old ST506 interface, but it's *much* faster.  It
remains the cost-effective choice for smaller drives (up to 500MB) on systems
that won't be hitting the disk constantly.  Unless you're running a heavily
used network or database server, don't assume SCSI will make any noticeable
difference.

Also, of course, IDE is cheaper.  Many motherboards have IDE right on board
now; if not, you'll pay maybe $15 for an IDE adapter board, as opposed to $200+
for the leading SCSI controller.  Also, there are reports that the cheap SCSI
cabling most vendors ship can be flaky.  It's alleged that you have to use
expensive high-class cables for consistently good results.  If anyone out there
has hard data on this, don't be shy --- I'd like to be able to confirm or
deny it in a future Guide!

One savvy netter observes "Don't discount ESDI, which is making a comeback.
At least with ESDI the system knows what the tracks and sectors are -- the OS
should know this to do good seek optimization."  He goes on to observe that
some ESDI drives are actually faster than SCSI.  ESDI hardware is cheaper, too.
Our editorial opinion is that this is probably a good idea if you're sure
you're *never* going to want a tape drive --- the SCSI/ESDI price difference
will get eaten if you have to buy a separate tape controller.

(If you can do your own installation, I hear that used 150/250MB SCSI drives
are getting quite common and cheap on the net.  All 150MB QIC type drives can
do 250MB on extended-length tapes, though some manufacturers discourage you
from doing this to avoid excessive heade wear.  But back to disks...)

The following, by Ashok Singhal <[e-mail]> of Sun
Microsystems, is a valiant attempt to demystify SCSI terminology.

The terms "SCSI" and "SCSI-2" refer to two different specifications.
Each specification has a number of options. Many of these options are
*independent* of each other.  I like to think of the main options (there are
others that I'll skip over because I don't know enough about them to talk
about them on the net) by classifying them into five categories:

1. Logical
    This refers to the commands that the controllers understand.
    SCSI-2 defined a common command set that is pretty much a
    superset of the SCSI command set.

2. Data Width
    8 bits (+ 1 parity) -> "normal"
    16-bits (+ 2 parity) -> "wide"
    32-bits (+ 4 parity) -> I don't know, "extra-wide??"

    All three options are available in SCSI-2 (yes,
    the draft spec I have even shows 32-bits!), although
    8-bit wide is still by far the most common.  Not sure, but I believe
    SCSI defined only 8-bit wide data path.

3. Electrical Interface
    single-ended (max cable length 6 meters)
    differential (max cable length 25 meters)

    Both options are available for SCSI-2 (I'm not sure about SCSI,
    but I think both options were available also)
    and this option is independent of options 2, 4, 5.  Differential
    is less common but allows better noise immunity and longer
    cables.

4. Handshake
    Synchronous (requests and acks alternate)
    Asynchronous (multiple requests can be outstanding)

    Both options are available for SCSI-2 (Not sure about SCSI,
    but I think both were available also).  This is negotiated
    between each target and initiator; asynchronous and synchronous
    transfers can occur on the same bus.  This is independent of
    2, 3 (Not sure about 1).

5. Synchronous Speed (does not apply for asynchronous option)
    "Normal" is up to 5 Mtransfers/sec ( = 5MB/s for 8-bit wide, more
                        for wider)
    "Fast" is up to 10 Mtransfers/s ( = 10 MB/s for 8-bit wide, more
                        for wider)

    The fast option is defined only in SCSI-2.
    This options basically defines shorter timing parameters
    such as the assertion period and hold time.
    The parameters of the synchronous transfer are negotiated
    between each target and initiator so different speed transfers
    can occur over the same bus.

F. Other Disk Decisions

Look at seek times and transfer rates for your disk; under UNIX disk speed and
throughput are so important that a 1-millisecond difference in average seek
time can be noticeable.

Previous issues said "Disk cacheing is good, but there can be too much of a
good thing.  Excessively large caches will slow the system because the overhead
for cache fills swamps the real accesses (this is especially a trap for
databases and other applications that do non-sequential I/O).  More than 100K
of cache is probably a bad idea for a general-purpose UNIX box; watch out for
manufacturers who inflate cache size because memory is cheap and they think
customers will be impressed by big numbers."  This may no longer be true on
current hardware; in particular, most controllers will interrupt a cache-fill
to fulfill a `real' read request.

In any case, having a large cached hard drive (particularly in the IDEs) often
does not translate to better performance.  For example, Quantum makes a 210Mb
IDE drive which comes with 256Kb cache.  Conner and Maxtor also have 210Mb
drives, but only with 64Kb caches.  The transfer rate on the drives, however,
show that the Quantum comes in at 890Kb/sec, while the Maxtor and Conner fly
away at 1200Kb/sec.  Clearly, the Conner and Maxtor make much better use of
their smaller caches.

Many retailers seem to enjoy advertising the "9ms" Quantum 52/80/120/200Mb
drives.  This speed, of course, is bogus.  All the quantum drives are at least
16ms is average access.  The 9ms already includes the cacheing speedup.

However, it may be that *any* hardware disk cacheing is a lose for UNIX!  Scott
Bennett <[e-mail]> reports a discussion on comp.unix.wizards:
"nobody found the hardware disk caches to be as effective in terms of
performance as the file system buffer cache...In many cases, disabling the
hardware cache improved system performance substantially.  The interpretation
of these results was that the cacheing algorithm in the kernel was superior to,
or at least better tuned to UNIX accesses than, the hardware cacheing
algorithms."

On the other hand, Stuart Lynne <[e-mail]> writes:

Ok. What I did was to use the iozone program.

What this showed was that on my root disk in single user mode I could get
about 500kb for writing and 1000kb for reading a 10MB file. With the disk
cache disabled I was able to get the same for writing but only about 500kb
for reading.  I.e. it appears the cache is a win for reading, at least if you
have nothing else happening.

Next I used a script which started up iozone in parallel on all four disks,
two to each of the big disks (three) and one on the smaller disk. A total of
seven iozone's competing with each other.

This showed several interesting results. First it was apparant that higher
numbered drives *did* get priority on the SCSI bus. They consistantly got
better throughput when competing against lower numbered drives. Specifically
drive 1 got better results than drive 0 on controller 0. Drive 4 got better
results than drive 3 on controller 1. All of the drives are high end Seagate
and have similiar characteristics.

In general with cache enabled the results where better for reading than
writing. When the cache was disabled the write speed in some cases went up a
bit and the read speed dropped. It would seem that the readahead in some
cases can compete with the writes and slow them down.

My conclusions are that we'll see better performance with the cache. First
the tendency is to do more reading than writing in your average UNIX system
so we probably want to optimize that. Second if we assume an adequate system
cache slow writes shouldn't affect an individual process much. When we write
we are filling the cache and we don't usually care how long it takes to get
flushed. Of course we would notice it when writing very large files.

Thus, I can only recommend experiment.  Try disabling the cache.  Your
throughput may go up!

G. Souping Up X Performance

One good way to boost your X performance is to invest in a graphics card with a
dedicated blitter and a high-speed local-bus connection, like the ATI 8514/A
series or the S3-based Quantum, Wind/X and Orchid Fahrenheit 1280.  A number of
clone vendors offer these accelerator options relatively cheap and can make
your X go like a banshee; however, stock X doesn't support them yet.

These cards speed up X in two ways.  First, they offload some common screen-
painting operations from the main processor onto specialized processors on the
card itself.  Secondly, by using a local bus, they make it possible to send
commands to the card faster than the ISA bus could allow.  The combined effect
can be eye-poppingly fast screen updates even at super-VGA resolutions.

In general, the ATI approach (normal bus, dedicated blitter and optimization
for special functions like character drawing) will speed up text display, text
scrolling and window resize/move operations a lot, but line-drawing and
graphics only a little.  S3, on the other hand, speeds up high-bandwidth
graphics drawing a lot but doesn't have as big an advantage for ordinary
text operations.  You pays your money and takes your choice.  Benchmarks
indicate that most non-CAD users are better served by the ATI approach.

However, I am now using SGCS X on an S3 with a 17" monitor on a 486/50DX2 and
can report that it is quite fast enough to make X pleasant to use, thank you.
Opaque windows can be dragged like paper.  This is *fun*!

The X servers on SCO, Dell and Esix support the ATI Ultra and Fahrenheit 1280,
and third-party servers for SVr4 are available from MetroLink (email
[e-mail]) or SGCS ([e-mail]).  There is said to be a third
vendor in this market, "Pittsburgh Powercomputing", but the name is all the
info I have on them.

Here is a current price list from MetroLink:

                         Description                        Price
      ---------------------------------------------         ------
      Runtime (all servers, standard and contrib clients)   299.00
      Development (full X11 and Motif 1.1.4 libraries)      299.00
      Xv - Real-Time Video in an X window (true server       99.00
           extension)
      Xie - X Imaging Extension                             199.00

And here is the corresponding info from SGCS:

                         Description                        Price
      ---------------------------------------------         ------
      Full X11R5 binaries licensed for a single CPU         295.00 **
      Enhanced X11R5 source code                            195.00 **
      MIT source code of contributed clients                 50.00
      Motif binaries for a single CPU                      245.00 **
      X11R5 Documentation Set                              150.00 **
      PHIGS Documentation Set                               75.00

    ** DISCOUNTS:
    If your choose more than one selection from any of the (**) items above
    you will receive the following discounts:  $50 off on 2 selections,
    $75 off on 3 selections, $100 off on 4 selections

If you're feeling *really* flush, plump for a 15", 17" or even 20" monitor.
The larger size can make a major difference in viewing comfort.  Also you'll be
set for VESA 1280x1024 when everybody gets to supporting that.  In the mean
time, the bigger screen will allow you to use fonts in smaller pixel sizes so
that your text windows can be larger, giving you a substantial part of the
benefit you'd get from higher pixel resolutions.

If you can, buy your monitor from someplace that will let you see the same
monitor (the very unit you will walk out the door with, not a different or
`demo' unit of the same model) that will be on your system.  There's a *lot* of
quality variation even in "premium" monitor brands.

The VESA (Video Electronics Standards Association) standard for local bus video
connectors is now out.  When you buy local-bus motherboards, insist that they
be VESA-conforming.  Be very clear about this and get a commitment from your
vendor; some unscrupulous operations may still be attempting to unload pre-VESA
motherboards on unsuspecting customers.

However, beware of an associated problem.  The VESA standard only recommends
local bus connector speeds up to 40 mb/sec (this allegedly has to do with
either holding down RFI emissions or clock skew problems; depends on who you
listen to).  For unbuffered designs (which most vendors prefer for performance)
VESA recommends at most two expansion slots on 33MHz machines, at most one on
40MHz, and none at all sbove that speed (that is, all devices should be
directly on the motherbord).  They recommend a limit of 2 VESA devices at all
speeds.

Thus, true 50DX or higher processors (but *not* 50DX2s) may actually have to be
*slowed down* to work with VESA hardware on expansion boards.  The long-term
solution is either hard-disk & video controllers right on the motherboard, a
revised "mezzanine" VESA that decouples the local-bus signals from the CPU,
or something like Intel's proposed PCI standard.

In the meantime, beware of vendors purporting to sell 50MHz "VESA"
mptherboards.  They're not.

V. Tape Drive Follies

You should have a tape drive for backup, and because most UNIX vendors like to
distribute their OS on tape.  Ideally, your tape backup should be able to image
your entire disk.  Unfortunately, this can get quite expensive for large disks,
as we'll see below.

There are two major technologies in today's desktop tape drive market; QIC
(Quarter Inch Cartridge) at the low end and midrange, and DAT (Digital Audio
Tape) at the high end.  The dividing line is about 1GB capacity.

DAT is a new technology; it's not far down its price curve yet, but clearly
where the future is.  DAT drive capacities are quoted in *gigabytes* (that is,
thousands of megabytes).

Most conventional QIC drives have capacities up to 525 megabytes (a little more
than half a gig).  A few high-end units have 1.35GB capacity.  QIC is a mature
technology, but one plagued by hardware incompatibilities and driver bugs.
Part of the problem is that, until recently, hard disks were small enough
relative to a floppy's capacity that demand for high-volume backup technology
was low in the PC world; QIC vendors tended to be small, insular,
technology-driven firms relatively uninterested in standardization.

As a result, understanding tape drive specifications is far from trivial.
Tape drive standards are developed by Quarter Inch Cartridge Drive Standards,
Inc. (805-963-3853), a consortium of drive and media vendors.  They develop
standards for controllers, transports, heads, and media.  Some of these
become ANSI standards.  We'll discuss the most important ones here.

Common Tape Drive Interfaces:

    QIC-02     --- intelligent hardware tape interface
    QIC-36     --- simple hardware tape interface
    QIC-104/11 --- SCSI-1 tape interface
    QIC-121    --- SCSI-2 tape interface

These standards describe the drive controller.  QIC-02 is presently by far the
most common, and QIC-36 nearly obsolete (it was designed at a time when
on-board intelligence for controllers was much more expensive than now).  The
SCSI standards are only rarely cited by number; usually, QIC-104 and QIC-121
devices are referred to simply as "SCSI tapes".

Common Recording Formats:

    QIC-24     --- 9-track  60-Mbyte tape format
    QIC-120    --- 15-track 125-Mbyte tape format
    QIC-150    --- 18-track 150-Mbyte tape format
    QIC-525    --- 26-track 525-Mbyte tape format

These standards describe the drive itself.

Now, in theory, these standards are upward compatible; that is, a QIC-120 drive
can read a QIC-24 tape, a QIC-150 drive can read both QIC-120s and QIC-24s, and
so on.  There's a potential gotcha here, though, called "media
incompatibility".  Thus, we also need to consider:

Common media:

    DC600A     --- for QIC-24 and QIC-120 drives
    DC6150     --- for QIC-150 drives
    DC6525     --- for QIC-525 drives

The Wangtek 5150ES (and possibly some other 525-megabyte drives) will,
according to its documentation, decode QIC-24 --- but it won't read a DC600A
medium formatted to QIC-150!  This is also reported of the Tandberg 3640
(QIC-120) drive.

So, make sure your tape drive can read the media your OS vendor is going to
ship on.  QIC-24 on DC600As and QIC-150 on DC6150s are very widely used as a
software distribution format in the UNIX world, and you probably want to make
sure your drive can read them.

60/120MB QIC drives are fairly cheap now but larger sizes (typically 150, 250,
525 QIC tapes and 1.3gig DAT) are not.  DAT drives, in particular, cost more
than a grand each (however, if you have large drives the up-front cost
difference can quickly get eaten up by media costs).

One interesting point is that if you've gone SCSI, a 150MB QIC (comparable to
the drives now popular on Suns) may well be cheaper than older 60MB technology;
the win is in the controller prices, which have plummeted since QIC-24 was the
cutting edge.

Tape drives are easy to find and pretty safe to buy through mail order.  It's
also possible to buy reconditioned but warrantied used drives substantially
cheaper than new.  One correspondent recommended Super Technologies of Chino,
CA (800 322 3999); they'll sell you a rebuilt Wangtek 150 with a 7-month
warranty and a controller card for $300 and change, or a DAT drive for $800.

One warning: a lot of DOS-box vendors push Colorado Memory Systems "mini-QIC"
drives with jumperless cards configured at runtime by the CMS backup software.
Make sure you do *not* get one of these.  They're cheap, and work for DOS, but
UNIX doesn't know that it has to poke controller registers to make the tape
transport accessible.  Besides, they *look* cheap, like they're put together
out of baling wire and spit --- I wouldn't trust their long-term reliability.

Another warning: The Wangtek 5150ES is incompatible w/ the Adaptec 1742 or 1740
in the EISA "enhanced" or 32 bit mode.  Running the Adaptec EISA card in
"standard" mode (16 bit ISA mode) is the only solution if you get stuck with a
5150ES.

Your humble editor has a few battle scars from tape drive integration at this
point (the rants about Wangtek and CMS drives are from personal experience).
We recommend the Archive ST525, a fine fast drive that works nicely with the
Adaptec 1542B, *can* read DC600A/QIC-24, and handles highest-capacity QIC-525
tapes.  Note however that some versions of its documentation have a critical
typo in the section on setting SCSI drive IDs; they give the ID jumpers as
JP3/JP2/JP1 when they are actually JP8/JP7/JP6.  If you are in any doubt about
your drive or manual, call Archive tech support and check.  Also, it does *not*
seem to be able to read QIC-120 tapes as claimed; at least, 125MB backup tapes
from my old AT&T 6386WGS are unreadable.

VI. Of Mice and Machines

Mice and trackballs used to be simple; now, thanks to Microsoft, they're
complicated.  In the beginning, there was only the Mouse Systems 3-button
serial mouse; this reported status to a serial port 30 times a second using a
5-byte serial packet encoding now called "C" protocol.  The Logitech Series 7
and 9 mice were Mouse Systems-compatible.  All UNIXes that have any mouse
support at all understand C-protocol serial mice.

Then Microsoft got into the act.  They designed a two-button serial mouse which
reports only deltas in a three-byte packet; that is, it sends changes in button
status and motion reports only when the mouse is actually moving.  This is
called `M' protocol.  Microsoft sold a lot of mice, so Logitech switched from
`C' to `M' --- but they added a third button, state changes for which show up
in an optional fourth byte.  Thus, `M+' protocol, upward-compatible with
Microsoft's `M'.  Most UNIX vendors add support for M+ mice, but it's wise to
check.

Bus mice are divided into 8255 and InPort types.  These report info
continuously at 30 or 60 Hz (though InPort mice have an option for reporting
deltas only), and you get interrupts on events and then have to poll hardware
ports for details.

In addition to serial mice and bus mice, there are "keyboard mice".  On PS/2s
there are two identical-looking keyboard ports, labeled (with icons) "mouse" &
"keyboard".  Both are 8 or 9 pin mini-DIN's that look like the regular PC
keyboard port only smaller.  I don't know what logical protocol the keyboard
mouse speaks.  Physically, the connector is eventually connected to the
keyboard processor (often an 8042).  The same keyboard processor that decodes
the keyboard decodes the mouse.  PS/2s have this port, many newer ISA/EISA
motherboards do as well.

All things considered, UNIX users are probably best off going with a serial
mouse (most current clone motherboards give you two serial ports, so you can
dedicate one to this and still have one for the all-important modem).  Not only
are the compatibility issues less daunting, but a serial mouse loads the
multitasking system less due to interrupt frequency.  Beware that most clone
vendors, being DOS oriented, bundle M-type mice for which UNIX support is
presently spotty, and they may not work with your X.  Ignore the adspeak about
dpi and pick a mouse/trackball that feels good to your hand.

VII.  Special considerations when buying laptops

Right now (March 1993) the laptop market is completely crazy.  The technology
is in a state of violent flux, with "standards" phasing in and out and prices
dropping like rocks.  We do not recommend buying a laptop until things have
settled out a bit.

However, if you have an immediate need for such a creature, there are a few
basic things to know that will help.

First: despite what you may believe, the most important aspect of any laptop is
*not* the cpu, or the disk, or the memory, or the screen, or the battery
capacity.  It's the keyboard feel, since unlike in a PC, you cannot throw the
keyboard away and replace it with another one unless you replace the whole
computer.  NEVER BUY ANY LAPTOP THAT YOU HAVE NOT TYPED ON FOR A COUPLE HOURS.
Trying a keyboard for a few minutes is not enough.  Keyboards have very subtle
properties that can still affect whether they mess up your wrists.

A standard desktop keyboard has keycaps 19mm across with 7.55mm between them.
If you plot frequency of typing errors against keycap size, it turns out
there's a sharp knee in the curve at 17.8 millimeters.  Beware of "kneetop" and
"palmtop" machines, which squeeze the keycaps a lot tighter and typically don't
have enough oomph for UNIX anyway; you're best off with the "notebook" class
machines that have full-sized keys.

Second: be careful that your laptop meets the minimum core and disk
requirements for the UNIX you want to run.  This is generally not a problem
with desktop machines, which can be upgraded cheaply and easily, but laptops
often have more stringent constraints.  Reject outright any machine that can't
carry 8MB RAM and a 120MB fast disk.

Third: with present flatscreens, 640x480 VGA color is the best you're going to
do.  If you want more than that (for X, for example) you have to either fall
back to a desktop or make sure there's an external-monitor port on the laptop
(and many laptops won't support higher resolution than the flatscreen's).

Fourth: look for Nickel-Metal-Hydride (NiMH) batteries, as opposed to the older
(Nickel-Cadmium) NiCad type.  NiMH batteries are great because they have
considerably higher energy capacity per pound that NiCads.  They need special
circuitry to charge them fast, so don't try to throw out your NiCads and
replace them with NiMH cells if you use a fast charger intended for NiCads.
Both kinds of cells can be damaged by overcharging at rates faster than 10
hours.

Fifth: Most laptop electronics are still 5-volt CMOS.  The coming thing is
3.3-volt CMOS with power-management features on the processor.  Buy this,
if you can, to nearly double your use time between recharges.

Sixth: about those vendor-supplied time-between-recharge figures; DON'T BELIEVE
THEM.  They collect those from a totally quiescent machine, sometimes with the
screen or hard disk turned off.  Under DOS, you'd be lucky to get half the
endurance they quote; under UNIX, which hits the disk more often, it may be
less yet.  Figures from magazine reviews are more reliable.

One final note.  Initial load of your UNIX can be a serious hurdle with
laptops, as they don't tend to have on-board QICtape drives :-).  The best
solution is to spring for an Ethernet card on the portable and use the
network-load facility supported by Dell or ESIX.  Otherwise you're going
to be shuffling a *lot* of 3.5" floppies.

VIII. When, Where and How to Buy

If you're a serious UNIX hacker for either fun or profit, you're probably in
the market for what the mail-order vendors think of as a high-end or even
`server' configuration, and you're going to pay a bit more than the DOS
lemmings.  On the other hand, prices keep dropping, so there's a temptation to
wait forever to buy.  A tactic that makes a lot of sense in this market, if you
have the leisure, is to fix in your mind a configuration and a trigger price
that's just a little sweeter than the market now offers and buy when that's
reached.

Direct-mail buying makes a lot of sense today for anyone with more technical
savvy than J. Random Luser in a suit.  Even from no-name mail-order houses,
parts and system quality tend to be high and consistent, so conventional
dealerships don't really have much more to offer than a warm fuzzy feeling.
Furthermore, competition has become so intense that even mail-order vendors
today have to offer not just lower prices than ever before but warranty and
support policies of a depth that would have seemed incredible a few years back.
For example, many bundle a year of on-site hardware support with their medium-
and high-end "business" configurations for a very low premium over the bare
hardware.

Note, however, that assembling a system yourself out of mail-order parts is
*not* likely to save you money over dealing with the mail-order systems
houses.  You can't buy parts at the volume they do; the discounts they command
are bigger than the premiums reflected in their prices.  The lack of any
system warranty or support can also be a problem even if you're expert enough
to do the integration yourself --- because you also assume all the risk of
defective parts and integration problems.

Cruise through "Computer Shopper" and similar monthly ad compendia.  Even if
you decide to go with a conventional dealer, this will tell you what *their*
premiums look like.

You may want to subscribe to ClariNet Communications's "Street Price Report",
a digest of lowest current quoted prices and sources (send inquiries to
[e-mail]).  It's $29.95 per year, so using it just once is likely
to save you more money than the subscription.

The Street Price Report is issued every other Thursday; you can have it
emailed to you, or get it from an FTP site and decrypt it using an emailed
key.  It covers a wide variety of hardware and software.  Quotes are collected
from the ad sections of major magazines including "Computer Shopper" and "PC
Magazine".  Once you've cruised the magazines, you know what you want and are
after the lowest price, you can nail it without fail with the Street Price
Report.

Another alternative to conventional dealerships (with their designer "looks",
stone-ignorant sales staff, and high overheads that *you* pay for) is to go
with one of the thousands of the hole-in-the-wall stores run by immigrants from
the other side of the International Date Line.  They're usually less ignorant
and have much lower overheads; they do for you locally what a mail-order house
would, that is assemble and test parts they get for you from another tier of
suppliers.  You won't get plush carpeting or a firm handshake from a white guy
with too many teeth and an expensive watch, but then you didn't really want to
pay for those anyway, right?

A lot of vendors bundle DOS 5.0 and variable amounts of DOS apps with their
hardware.  You can tell them to lose all this cruft and they'll shave $50 or
$100 off the system price.  However, David Wexelblat observes "there are at
least two situations in which the Unix user will need DOS available: 1) most,
if not all, EISA configuration utilities run under DOS, and 2) SCSICNTL.EXE by
Roy Neese is a godsend for dealing with SCSI devices on Adaptec boards."

Don't forget that (most places) you can avoid sales tax by buying from an
out-of-state mail-order outfit, and save yourself 6-8% depending on where you
live.  If you live near a state line, buying from a local outfit you can often
win, quite legally, by having the stuff shipped to a friend or relative just
over it.  Best of all is a buddy with a state-registered dealer number; these
aren't very hard to get and confer not just exemption from sales tax but
(often) whopping discounts from the vendors.  Hand him a dollar afterwards to
make it legal.

(Note: I have been advised that you shouldn't try the latter tactic in
Florida -- they are notoriously tough on "resale license" holders).

(Note II: The Supreme Court recently ruled that states may not tax out-of-state
businesses under existing law, but left the way open for Congress to pass
enabling legislation.  Let's hope the mail-order industry has good lobbyists.)

You can often get out of paying tax just by paying cash, especially at computer
shows.  You can always say you're going to ship the equipment out of the
state.

On the other hand, one good argument for buying locally is that you may have to
pay return postage if you ship the system back.  On a big, heavy system, this
can make up the difference from the savings on sales tax.

IX. Questions You Should Always Ask Your Vendor

A. Minimum Warranty Provisions

The weakest guarantee you should settle for in the mail-order market should
include:

* 72-hour burn-in to avoid that sudden infant death syndrome.  (Also,
  try to find out if they do a power-cycling test and how many repeats
  they do; this stresses the hardware much more than steady burn-in.)

* 30 day money-back guarantee.  Watch out for fine print that weakens this
  with a restocking fee or limits it with exclusions.

* 1 year parts and labor guarantee (some vendors give 2 years).

* 1 year of 800 number tech support (many vendors give lifetime support).

Additionally, many vendors offer a year of on-site service free.  You should
find out who they contract the service to.  Also be sure the free service
coverage area includes your site; some unscrupulous vendors weasel their way
out with "some locations pay extra", which translates roughly to "through the
nose if you're further away than our parking lot".

If you're buying store-front, find out what they'll guarantee beyond the
above.  If the answer is "nothing", go somewhere else.

B. Documentation

Ask your potential suppliers what kind and volume of documentation they supply
with your hardware.  You should get, at minimum, operations manuals for the
motherboard and each card or peripheral; also an IRQ list, and a bad-block
listing if your Winchester is ESDI rather than IDE or SCSI (the latter two
types of drive do their own bad-block mapping internally).  Skimpiness in this
area is a valuable clue that they may be using no-name parts from Upper
Baluchistan, which is not necessarily a red flag in itself but should prompt
you to ask more questions.

C. A System Quality Checklist

There are various cost-cutting tactics a vendor can use which bring down the
system's overall quality.  Here are some good questions to ask:

* Is the memory zero-wait-state?  One or more wait states allows the vendor to
  use slower and cheaper memory but will slow down your actual memory subsystem
  throughput.  This is a particularly important question for the *cache*
  memory!

* If you're buying a factory-configured system, does it have FCC certification?
  While it's not necessarily the case that a non-certified system is going
  to spew a lot of radio-frequency interference, certification is legally
  required --- and becoming more important as clock frequencies climb.  Lack
  of that sticker may indicate a fly-by-night vendor, or at least one in
  danger of being raided and shut down!

X. Things to Check when Buying Mail-Order

A. Tricks and Traps in Mail-Order Warranties

Reading mail-order warranties is an art in itself.  A few tips:

Beware the deadly modifier "manufacturer's" on a warranty; this means you have
to go back to the equipment's original manufacturer in case of problems and
can't get satisfaction from the mail-order house.  Also, manufacturer's
warranties run from the date *they* ship; by the time the mail-order house
assembles and ships your system, it may have run out!

Watch for the equally deadly "We do not guarantee compatibility".  This gotcha
on a component vendor's ad means you may not be able to return, say, a video
card that fails to work with your motherboard.

Another dangerous phrase is "We reserve the right to substitute equivalent
items".  This means that instead of getting the high-quality name-brand parts
advertised in the configuration you just ordered, you may get those no-name
parts from Upper Baluchistan --- theoretically equivalent according to the
spec sheets, but perhaps more likely to die the day after the warranty expires.
Substitution can be interpreted as "bait and switch", so most vendors are
scared of getting called on this.  Very few will hold their position if you
press the matter.

Another red flag: "Only warranted in supported environments".  This may mean
they won't honor a warranty on a non-DOS system at all, or it may mean they'll
insist on installing the UNIX on disk themselves.

One absolute show-stopper is the phrase "All sales are final".  This means you
have *no* options if a part doesn't work.  Avoid any company with this policy.

B. Special Questions to Ask Mail-Order Vendors Before Buying

* Does the vendor have the part or system presently in stock?  Mail order
  companies tend to run with very lean inventories; if they don't have your
  item in stock, delivery may take longer.  Possibly *much* longer.

* Does the vendor pay for shipping?  What's the delivery wait?

* If you need to return your system, is there a restocking fee? and will the
  vendor cover the return freight?  Knowing the restocking fee can be
  particularly important, as they make keep you from getting real satisfaction
  on a bad major part.  Avoid dealing with anyone who quotes more than a 15%
  restocking fee --- and it's a good idea, if possible, to avoid any dealer
  who charges a restocking fee at all.

Warranties are tricky.  There are companies whose warranties are invalidated by
opening the case.  Some of those companies sell upgradeable systems, but only
authorized service centers can do upgrades without invalidating the warranty.
Sometimes a system is purchased with the warranty already invalidated.  There
are vendors who buy minimal systems and upgrade them with cheap RAM and/or disk
drives.  If the vendor is not an authorized service center, the manufacturer's
warranty is invalidated.  The only recourse in case of a problem is the
vendor's warranty.  So beware!

C. Payment Method

It's a good idea to pay with AmEx or Visa or MasterCard; that way you can stop
payment if you get a lemon, and may benefit from a buyer-protection plan using
the credit card company's clout (not all cards offer buyer-protection plans,
and some that do have restrictions which may be applicable).  However, watch
for phrases like "Credit card surcharges apply" or "All prices reflect 3% cash
discount" which mean you're going to get socked extra if you pay by card.

Note that many credit-card companies have clauses in their standard contracts
forbidding such surcharges.  You can (and should) report such practices to
your credit-card issuer.  If you already paid the surcharge, they will usually
see to it that it is restituted to you.  Credit-card companies will often stop
dealing with businesses that repeat such behavior.

XI. Which Clone Vendors to Talk To

I went through the March 1992 issue of Computer Shopper calling vendor 800
numbers with the following question: "Does your company have any
configurations aimed at the UNIX market; do you use UNIX in-house; do
you know of any of the current 386 or 486 ports running successfully
on your hardware?

I didn't call vendors who didn't advertise an 800 number.  This was only partly
to avoid phone-bill hell; I figured that toll-free order & info numbers are so
standard in this industry sector that any outfit unable or unwilling to spring
for one probably couldn't meet the rest of the ante either.  I also omitted
parts houses with token systems offerings and anybody who wasn't selling
desktops or towers with a 386/33DX or heavier processor inside.

After plundering Computer Shopper, I called up a couple of "name" outfits that
don't work direct-mail and got the same info from them.

The answers I get revealed that for most clone vendors UNIX is barely a blip on
the screen.  Only a few have tested with an SVr4 port.  Most seem barely aware
that the market exists.  Many seem to rely on their motherboard vendors to tell
them what they're compatible, without actually testing whole systems.  Since
most compatibility problems have to do with peripheral cards, this is a
problem.

Here's a summary of the most positive responses I got:

A --- Advertises UNIX compatibility.
C --- Has known UNIX customers.
I --- Uses UNIX in-house.
T --- Have formally tested UNIX versions on their hardware.
F --- Have 486/50 systems
* --- Sounded to me like they might actually have a clue about the UNIX market.

Vendor        A C I T    F *    Ports known to work
---------------    - - - -    - -    -----------------------------------------------
ARC        . . X X . .    SCO XENIX 2.3.2, SCO UNIX 3.2.1
AST        . X X X X *    SCO UNIX 3.2.4, ODT 2.0 Microport V/4
Allegro        . . X X . .    SCO XENIX 3.2.4
Altec        . X . X    . .    XENIX (no version given).
Ares        . X X X X *    AT&T 3.2, ISC (version unknown)
Basic Time    . X X X X *    SCO XENIX 2.3.2, have in-house UNIX experts.
Binary Tech    . X . X X .    Claims to work with all versions.
CCSI        X X . . X .    They've used SCO XENIX, no version given.
CIN        . X . . . .    SCO UNIX (version not specified)
CSS        . X . X . *    SCO 3.2.2, ISC 3.0, SCO ODT. See Will Harper.
Centrix        X . . . . .    No specifics on versions.
Compudyne    . X X X X .    Couldn't get details on which versions.
Comtrade    . X . X X .    Couldn't get details on which versions.
Datom        X X X X X .    SCO XENIX 3.2.
Dell        X X X X X *    See Dell SVr4 data.
Desert Sands    X X . X X .    SCO UNIX 3.2.4
Digitech    . X . X . .    SCO UNIX 3.2.1, XENIX 2.3.1
EPS        X X X X . *    SCO XENIX 3.2.4, ISC & AT&T (versions not sp.)
Everex        X X X X X *    Esix 4.0.3
Gateway 2000    X X X X    X *    SCO UNIX 3.2.0. XENIX 2.3.4 ISC 3.0, ESIX 4.0.3
HD Computer    . X . X X .    SCO UNIX 3.2, SCO XENIX 3.2.2
HiQ        . X . X . .    SCO UNIX (version not specified)
Infiniti    . X . X X .    SCO UNIX (versions not specified)
Insight        . . X . X .    SCO XENIX 3.2.4.  No tech support for UNIX
Keydata        X . X X X *    SCO version 4, ISC 3.2
Legatech    . X . . X .    SCO UNIX, ISC (versions not specified)
MicroGeneration    . . X . . .    Uses XENIX.
MicroLab    X . . . . .    SCO UNIX, SCO XENIX
MicroSmart    X X . X . .    SCO XENIX (version not specified)
Microlink    X . . X X .    SCO XENIX (version not specified)
Myoda        X X . X X .    SCO XENIX 3.2.2, ISC 3.2
Naga        . X . X X *    SCO & XENIX 3.2.
Northgate    X X . X X *    SCO UNIX 3.2
PC Brand    . X X X . .    SCO XENIX, ISC UNIX
PC Professional    . X . X . .    ISC 3.2
PC-USA        X X . X . .    ISC 5.3.2 and SCO 3.2
Profex        . X . X    . .    SCO XENIX 3.2.
Royal Computer    . X . . X .    No details on versions.
SAI        X X . X X .    SCO UNIX 3.2.2.
Santronics    . . X X X .    SCO XENIX 3.2.4
Solidtech    . X . . . .    Dell (no version given), ISC 3.2.
Strobe        . . . X X .    SCO, Microport, ISC (no version numbers given)
Swan        X X X X X *    SCO 2.3.1, UNIX 3.2, ISC 3.2v2.0.2
TriStar        . X X X X *    SCO UNIX 3.2.2, XENIX 2.3.2, ISCr4
Zenon        . X . X X *    SCO UNIX (version not specified)
Zeos        . X X X X *    SCO XENIX 3.2.4, AT&T 3.2

Special notes about a few vendors who appear to have a clue:

Ares targets some of its systems for UNIX CAD use.  They have a house wizard
name Ken Cooper (everybody calls him "K.C.").

EPS targets some 486 EISA configurations for UNIX.

Swan doesn't know the UNIX market very well yet, but their project manager
wants a bigger piece of it and is interested in doing some of the right
things.  They have a house wizard, one John Buckwalter.

Dell, of course, supports an industry-leading SVr4 port.  They're a bit on
the pricy side, but high quality and very reliable.  Lots of UNIX expertise
there; some of it hangs out on the net.

Zeos is on the net as zeos.com, with a uunet connection; they host a UNIX BBS.
They have an in-house UNIX group reachable at [e-mail].  There are
biz.zeos.general and biz.zeos.announce groups on USENET.

Special notes about a lot of vendors who appear to have *no* clue:

Vendors where I couldn't get a real person on the line, either because
no one answered the main number or because I couldn't raise anyone at
tech support after being directed there: Sunnytech, Quantex, AMS, USA
Flex, Lapine, Syntax Computer, MicroTough, PAC International, The Portable
Warehouse.

Vendors where the question met with blank incomprehension, puzzlement,
consternation, or "We've never tested with UNIX": Allur, AmtA, Aplus, HiTech,
Locus Digital Products, LodeStar, Ultra-Comp, UTI Computers, PC Turbo Corp,
Evertek, Microcomputer Concepts, Jinco Computers, UWE, ToughCom, System
Dynamics Group, Terribly Fast Bus Systems.

Vendors who understood the questions but had no answer: Bulldog Computer
Products, LT Plus, Standard Computer, JCC.

Vendors who said "Yes, we're UNIX-compatible" but had no details of any tests:
CompuCity.

Vendors who said "Go ask our motherboard vendor": Ariel Design, Lucky Computer
Co., V-com, Professional Computer, MicroLine, MileHi.

Vendors who sent me to a toll number: Absec, Hokkins, New Technologies, Mirage.

Vendors that believe they have UNIX customers, but can't be any definite than
that: Austin Computer Systems, PC Professional, Treasure Chest Computer
Systems, CompuAdd Express, FastMicro, MidWest Micro.

Final note:

If you order from these guys, be sure to tell them you're a UNIX customer
and don't need the bundled DOS.   This will shave some bucks off the system
price, *and* it may encourage them to pay more attention to the UNIX market.
--
    Send your feedback to: Eric Raymond = [e-mail]