💾 Archived View for spam.works › mirrors › textfiles › computers › crtstuff.txt captured on 2024-02-05 at 11:24:44.
⬅️ Previous capture (2023-06-14)
-=-=-=-=-=-=-
????????????????????????????????????????????????????????????????
? ?
? The Quest for the ?
? Ultimate Display System ?
? ?
? by ?
? Steve Gibson ?
? GIBSON RESEARCH CORPORATION ?
? ?
? Portions of this text originally appeared in Steve's ?
? InfoWorld Magazine TechTalk Column. ?
? ?
????????????????????????????????????????????????????????????????
I remember those simple days not so long ago when a purchaser of
a brand new IBM Personal Computer had only one choice to make
when it came to choosing the display system for his computer.
Blessedly, the only choice to be made back then was between
either a monochrome display and adpater, the so-called MDA
solution, or a color screen and adapter, and CGA route. Needless
to say, things are not so simple or straightforward these days!
There are so many choices and options open to a purchaser or
upgrader of a PC that I'd be crazy to even ATTEMPT to offer any
clarification or guidance.
Okay, so call me crazy. It's time to choose the display
subsystem for Steve's Dream Machine and I've got some real
surprises in store for you this time! I've spent most of the
past six weeks (when I haven't been reconfiguring my system
between VM/386, Desqview, and Omniview) researching, probing,
and digging for the best possible contemporary display solution
for the least possible money.
Surprisingly, my research has uncovered some truly startling
facts which I'll be sharing through the next few weeks as we
explore THE DISPLAY SYSTEM FOR STEVE'S DREAM MACHINE. We'll see
things like why the new 16-bit display adapters are generally
not worth a dime more than their older 8-bit predecessors, how
few, if any, VGA adapters on the market are REALLY register
level compatible, and what risk that represents in light of
IBM's unknowable future plans. We'll see what extra display
memory DOESN'T buy for you because many of the manufacturer's
device drivers don't even use it, why the highest resolution
modes can be much more trouble than they're worth, and what NOT
to pay for a great high resolution display.
Given the incredible variety of available choices (not at all
like in the old days) it may surprise you to know that I have
found ONE set of choices which delivers far more bang for the
buck than any other! In order to give these conclusions a proper
perspective, let's first step back for a moment to review the
technological fundamentals and constraints which give our
decision meaning. Then we'll see where we've been and where
we're going.
No matter what style of display screen and interface adapter is
in use, several fundamentals always apply. In the first place,
the image displayed by a CRT screen is not at all the static
image which it appears to be. In fact, there's never really an
image being displayed on the screen at all! If you were to
photograph a computer display screen with a high speed camera
you'd only see a single bright dot of light rather than an
entire image.
The illusion of a screen full of information is created with the
aid of some incredible technology. The screen is actually
painted by a single madly whizzing dot of light which traces
successive horizontal lines across and down the face of the
display tube. When I say "madly whizzing" I'm not exaggerating
because a typical CRT screen paints horizontal scan lines on its
face at the rate of 350,000 inches per second, which is 20,000
miles per hour! This furious speed is required in order to fool
our eyes into believing that the entire image is being
continuously displayed when in fact it's mostly NOT being
displayed!
A typical display screen consists of about 450 of these
horizontally scanned lines, each of which must be redrawn or
"refreshed" at least every sixtieth of a second. This means
scanning across 27,000 lines per second, every second. If this
is not done our eye will perceive that the lines are not being
continuously illuminated and the illusion we've tried so hard to
achieve will fail.
As the single dot of light traces its furious course it changes
color thus tracing out the full screen image which is stored in
the display adapter's DISPLAY REFRESH MEMORY. On a monochrome
screen which is limited to displaying a single color, the dot of
light varies in brightness only, whereas a color display system
allows the dot's instantaneous color to be varied as well.
So we're left with a number of important concepts: In order to
eliminate the overall "refresh flicker" of a display screen, the
entire screen must be redrawn or repainted approximately 60
times each second. Since the scanning dot traces lines from left
to right as it moves more slowly vertically from the top of the
screen down, the downward motion is referred to as the screen's
Vertical Refresh Frequency and the very rapid horizontal left to
right scanning is referred to as the display's Horizontal
Refresh Rate.
The Display System Adventure Continues
So we've seen that the display screens of our computers rely
entirely upon our eye's persistence of vision to assemble the
illusion of an image on the screen. Each dot which composes the
display must be redrawn, or refreshed, at least sixty times each
second in order to appear continuously illuminated. Now we'll
examine the evolution of our display screens, giving some
perspective to where we've been and where we are today.
The original Color Graphics Adapter (CGA) traces its ancestry
directly from commercial television. Commercial TV refreshes its
screen exactly 60 times per second with a horizontal scanning
frequency of 15,750 cycles per second. The IBM CGA display
utilizes this timing to support the images it generates. The
total number of horizontal scanning lines traced onto the screen
by a CGA system can be determined simply by calculating how many
horizontal lines are scanned out during one vertical scan. Since
the screen is scanned vertically 60 times per second, we divide
15,750 by 60 to yield the horizontal line count of 262. Two
hundred of these scan lines are used to display actual image
data, with the balance used to illuminate the CGA screen's border
region.
As we all remember, the CGA was not known for producing highly
legible text (for some it's not yet a memory). The prime
determiner of text quality is the number of individual pixel
dots which are available to display individual characters.
Dividing 200 total image lines by 25 lines of text yields just 8
scanning lines available per text line. Then since the CGA
adapter was able to display 640 dots across a horizontal line,
dividing this by 80 characters per line yields 8 pixel dots
horizontally per character.
So CGA technology yielded a budget of 8 by 8 pixels per
character. Since it is necessary to separate characters by at
least one blank pixel, and since characters are taller than they
are wide, CGA characters were designed to fit within a rectangle
of dots 5 wide by 7 dots high. If you have a few moments to
spare some time with some graph paper, try designing an entire
upper and lower case alphabet where each character fits within a
5 by 7 pixel cell. It's not simple, and there is no really great
solution.
Driven by the concern that serious business computer users would
be very unhappy with the appearance of CGA text, IBM decided to
provide a better text display alternative. The Monochrome
Display Adapter (MDA) was the result. In order to deliver more
legible characters, more pixels are required both horizontally
and vertically. Where the CGA fits 5 by 7 characters into 8 by 8
"cells," the IBM monochrome display provides much higher
resolution: 7 by 9 characters within a 9 by 14 space.
Changing the character resolution from 5 by 7 to 7 by 9 results
in a tremendous improvement in character legibility, and a 2-
pixel horizontal inter-character spacing with a 5-pixel vertical
spacing leaves the display's characters feeling quite uncrowded.
The question is, where did IBM get all those extra scan-lines?
25 lines of text with 14 scan-lines per line means a total of
350 active scan lines compared to the CGA's 200! The scan-line
count can be increased by increasing the horizontal scan
frequency so that more lines are scanned per second, or by
decreasing the overall vertical refresh rate thus allowing more
time to scan the horizontal lines.
IBM did both of these things to create the MDA standard. The MDA
refreshes its screen at only 50 cycles per second with a
horizontal scan rate of 18,432 hertz. Now, dividing 18,432 by 50
yields about 368 scan lines. Since 350 of these are required for
text display, the MDA is not able to display a border.
But how can IBM refresh the MDA screen at only 50 cycles per
second if we begin noticing a flicker as refresh frequencies
fall below 60 cycles per second? IBM compensated for our lack of
vision persistence by designing a highly persistent green
phosphor into their monochrome display. Many people immediately
noticed a "smeary" effect whenever the IBM monochrome display
scrolled text. This smearing was created by the use of a long
persistence phosphor which continued to glow long after the
screen's electron beam stopped refreshing the region.
If you've ever noticed an annoying continuous flicker from an
inexpensive clone monochrome display, now you know why. Most
clone monochrome displays use less expensive standard short or
medium-length phosphors... which are inadequate for masking the
very noticeable effects of the MDA's lower refresh rate. Also,
since the flicker-perception phenomenon is extremely subjective,
many people perceive flicker where others don't. I've
learned that I don't see flicker where other people are being
driven nuts by it.
With an understanding of the interactions of horizontal and
vertical scan rates and display resolution we're ready to
explore the EGA, VGA and multisync technologies.
The role of Hercules Graphics,
and the evolution of the EGA
We've seen how IBM designed their MDA monochrome display system
to deliver extremely well-formed characters by increasing the
display's horizontal scanning rate and decreasing the vertical
refresh rate. Before continuing our discussion of EGA, VGA, and
multisynchronous monitors, it's important to understand another
quite well established and significant display standard,
Hercules.
Perhaps IBM simply overlooked the idea of monochrome graphics
altogether, or underestimated the demand for the display of
graphic information. More likely though, IBM felt that the word-
processing market toward which they were targeting their
monochrome display system had no need to display graphics. How
could IBM, or anyone for that matter, have anticipated the
phenomenal effect Lotus' 123 spreadsheet product would have upon
the IBM compatible market?
While columns of numbers are indeed informative, the ability to
graphically display, correlate, and view the results of
spreadsheet calculations is extremely useful. The folks at
Hercules Computer quickly recognized this and designed a
wonderful solution which, with the early support of Lotus, became
a solid standard.
Since the Hercules high resolution mode was designed to operate
with an IBM or compatible monochrome monitor, at a horizontal
sweep rate of 18,432 cycles per second and a refresh frequency
of (only) 50 hertz, it could directly leverage the extremely
high resolution which IBM had designed into their monochrome
text system. The Hercules monochrome display resolution of 720
by 350 pixels made the CGA's 640 by 200 look quite sad when
compared side by side, and suddenly people could have both
readable text and great looking graphics at the same time and
from a single system.
IBM's next move demonstrated that they'd been listening to their
user's complaints about the low resolution of the CGA standard.
They were also watching the guys at Hercules make money like
crazy and were attempting to serve the always mixed blessing
requirements of full backwards compatibility. The IBM Enhanced
Graphics Display was IBM's second generation solution, and
it rapidly became a new standard for the industry.
By recognizing the CGA system's crying need for better text, IBM
saw that it had to crank up the scan line count to something
more like their monochrome display; however, since full-color
long persistence phosphor monitors are barely affordable by
small countries, IBM knew that it couldn't play the trick of
getting the scan line count up by lowering the system's overall
refresh rate below 60 cycles per second. The only alternative was
to push the system's horizontal scanning frequency higher than
the monochrome system's.
This would mean that their new EGA display system would not be
backwards compatible to the existing installed base of 200 scan
line resolution CGA software. (The non-optimal solution crimes
which are continually committed in the name of backwards
compatibility is probably my single
biggest pet peeve. It directly accounts for the unprogrammablity
of the Intel microprocessor instruction set!) So, in order to
achieve CGA compatibility from their new EGA system, IBM
invented the "bi-synchronous" display system.
By inverting the polarity of the EGA monitor's Vertical
Synchronization signal, the EGA adapter is actually able to
switch the EGA monitor between two separate modes: The CGA's
horizontal sweep rate of 15,750 cycles per second and the newly
invented EGA horizontal rate of 21,800 cycles per second. The
15,750 hertz rate yields a CGA software compatible resolution of
200 lines, while the 21,800 hertz rate results in a full
Hercules-type resolution of 350 lines. In EGA graphics mode, this
results in a significant, Hercules-similar resolution of 640
by 350 pixels.
Since IBM seems determined not to kick the horizontal resolution
of these systems up above 640 pixels, we don't quite get the full
character separation beauty of MDA and Hercules text. On the
other hand, the EGA's character resolution budget
of 8 by 14 pixels is significantly better than the CGA budget of
8 by 8 and allows lower case characters with descending tails
like "g," "p," "q," and "y" to be imaged cleanly. The EGA's
resulting well-formed characters made most people happy.
The EGA's final addition to the CGA standard was the provision
for additional colors. Where the CGA display could display 8
colors in either of two intensities, bright or dim, the EGA
display, when operating in EGA mode, allowed each of its three
primary colors, Red, Green, and Blue, to be mixed together in
any of four intensities. Therefore 4 times 4 times 4, or 64
total colors could be displayed by IBM's EGA display. Though
technology has passed the EGA monitor by, it represented an
adequate, backward compatible, unification of the CGA, MDA, and
Hercules standards.
IBM's recognition of the EGA's shortcomings
with the creation of the VGA "standard"
On our journey toward the goal of selecting the best possible
display system for the least possible money for Steve's Dream
Machine, we've traced the evolution of IBM compatible display
system technology from the original CGA and MDA standards
through the development of the Hercules and EGA standards. IBM's
announcement of its new generation PS/2 machines offers yet
another display system to "the standard" throne. Oddly named
after the integrated circuit chip which implements it, the Video
Graphics Array, or VGA, has provided enough new cleverness and
innovation to displace the prior EGA standard.
With graphic user interfaces gaining ever more market
recognition and IBM's own OS/2 Presentation Manager on the
horizon, IBM needed to push their graphics resolution offering
above the EGA's 640 by 350. At the same time, IBM wished to
further enhance the system's color capabilities, probably to
further differentiate itself from Apple Computer's
monochrome Macintosh products and to better compete with Apple's
newer colorful Mac II. To further confuse things, this was all
happening at a time when IBM was determined to lower its
manufacturing costs.
While the EGA display was innovative with its split-personality
dual-frequency horizontal sweep rate in order to deliver both
350 line vertical resolution without sacrificing 200 line CGA
compatibility, is was more expensive to manufacture than IBM was
now happy with. IBM made a brilliant move in their VGA system
which completely eliminated the need
for the expensive frequency changing display while actually
enhancing the appearance of older CGA-style text and graphics.
The VGA's fixed horizontal sweep rate of 31,500 cycles per
second offers several wonderfully clever savings. In the first
place, dividing the horizontal rate of 31,500 hertz by the 60
cycle vertical rate yields 525 total horizontal lines scannable
during one screen. This high scan line count delivers even
better legibility from VGA text which now has a text character
pixel budget of 8 by 16, while the EGA's barely adequate high
resolution line count of 350 jumps up to a very respectable 480.
The excess line count (the difference between the 525 total and
the 480 used) even allows a tidy 1/4 inch border in all modes.
The VGA's cleverness stems from two additional things which IBM
did in order to deliver backward compatibility to the CGA and
VGA. The VGA monitor's very fast horizontal scan rate put IBM in
the enviable position of actually having, in some cases, too
many scan lines, rather than too few. So in such cases IBM
slightly INCREASES the vertical refresh rate (to above 60 hertz)
in order to trim back on the number of lines displayed when they
need fewer.
Secondly, rather than slowing the display's HORIZONTAL rate
drastically down to the CGA's 15,750 cycles, in order to deliver
just 200 horizontal scan lines, the VGA raises its VERTICAL rate
just slightly up to 70 hertz which yields 400 scan lines. Then a
clever double-scanning approach is used to emulate the CGA's 200
line mode. Double scanning simply repeats each of the CGA's
lines twice and results in a higher resolution appearance while
maintaining complete software backward compatibility.
The only remaining "tweak" required involves keeping the VGA's
displayed screen height constant which the IBM VGA monitor
achieves by sensing the polarity of the Vertical Synchronization
signal sent to it by the VGA adapter. The monitor uses the
Vertical Sync signal polarity to adjust the spacing between
successive scan lines so that the VGA's image is kept almost
uniformly sized throughout the increasing jungle of new, old,
and older display modes.
Thus the VGA system scans 350, 400, and 480 lines to achieve
CGA, EGA, and VGA compatible display modes while leaving the
horizontal scanning rate set to a constant 31,500 hertz and only
tweaking the vertical refresh rate between a happy 60 and 70
cycles per second. The result is a simpler and far less
expensive VGA monitor which exceeds the EGA's capabilities and
delivers far cleaner CGA emulation.
The other major change presented by the VGA system is an
expansion of the system's color capabilities. The original CGA
monitor utilized one signal each for Red, Blue, and Green
colors, and an additional single signal for intensity which
delivered 16 total possible colors. The EGA expanded upon this
by providing two signals each for the Red, Green, and Blue
colors, thus delivering four intensities of each color, with 64
color mixtures possible. The VGA's color system operates in an
ANALOG rather than DIGITAL fashion where varying voltages, rather
than ON/OFF signals are provided for each color for mixing.
Software and memory limitations pare the resulting infinite
color possibilities down to a maximum of 256 colors chosen from
a total palette of 262,144 in some display modes.
NEC's Brilliant Creation of the Multisync,
and 800 by 600 Resolution Graphics
We've taken a detailed look at the evolution of IBM compatible
display systems, focussing almost exclusively upon the multitude
of standards which have first been set then soon superseded by
IBM. We've seen that the various display adapters have always
been "tightly coupled" to their display monitors and have
frequently employed fancy "kludge" solutions (like conditional
inverting of synchronization signal polarities) when necessary
to maintain backward compatibility to the multitude of prior
standards.
Amid the wilderness created by the incredible array of vertical
and horizontal scan rates, a solid alternative to the eternal IBM
lock-step frenzy has arisen. Originally conceived by Nippon
Electric Corporation (NEC) as an answer to just this problem, the
so-called "multi-synchronous" display monitors are now selling in
the hundreds of thousands for a very good reason.
In what could only be called a truly astounding leap of insight,
the designers at NEC integrated the past and predicted the
future when they invented their original NEC Multisync, a single
unified display monitor solution for all adapter technologies
past, present, and future. Rather than following IBM with yet
another tightly coupled clone display monitor, NEC invented a
single monitor which quietly displayed anything it might be
handed by the system's display adapter. By accepting an unheard
of range of vertical and horizontal synchronization frequencies,
as well as both digital and analog RGB intensity signals, the
NEC Multisync became virtually obsolescence-proof.
While IBM was busily requiring all of its EGA owners to
completely scrap their "yesterday's solution," EGA monitors which
would no longer be compatible with the VGA of today (and
tomorrow?), and purchase the all new VGA displays, proud
Multisync owners only needed to change their monitor's cable
then flip a couple of switches at the rear of their displays.
That's what I call truly brilliant engineering!
Of course it wasn't long until everyone else recognized NEC's
brilliance and began cloning multisynchronous monitors like mad.
Today's mail order ads are drenched in "generic multisynch-ness"
because it's simply the right way to go.
However, there's something else which makes multisynching the
right solution, and after extensive experimentation and
comparison it has become an INFINITELY CRITICAL COMPONENT of
Steve's Dream Machine: Support of the wonderful 800 x 600 pixel
super high resolution modes which are now available from all
state-of-the-art EGA and VGA display adapters.
Many of you will remember that Steve's Dream Machine and I have
been holding onto monochrome display technology for dear life...
looking to monitors such as the Wyse-700/Amdek-1280 and MDS
Genius to provide the truly useful bit-mapped graphics
resolution which is, and will be, required by today's and
tomorrow's desktop publishing, MS Windows, and OS/2 Presentation
Manager applications. Until many months of searching yielded the
incredible, ultimate, adapter/monitor combination, I didn't
believe that a color system could really deliver "truly useful"
(and in fact wonderful) high resolution bit-mapped displays. It
can. I'll tell you about the results of my quest, but first we
need a bit more foundation...
It turns out that truly useful bit-mapped resolution requires
stepping above even the VGA's new 640 by 480 resolution up to
800 by 600. By cranking the horizontal sync up to 35,100 and
sneaking the vertical refresh just a tad below 60 hertz to about
56, any solid multisynchronous monitor can readily display 600
lines of 800 full color pixels per line.
There's something magical about the difference between 640 by
350, 640 by 480, and 800 by 600. It's a staggering difference.
The prior two resolutions simply pale by comparison to 800 by
600. Trying to understand why things get so incredibly better as
the resolutions are increased, I've decided that it's because
the total pixel count increases with the PRODUCT of the
horizontal and vertical resolutions. This is a powerful
relationship. For example, on a screen with square resolution,
the total pixel count would increase with the SQUARE of the
screen's edge resolution, so a DOUBLING of edge resolution
produces a QUADRUPLING of the total pixel count. Consequently
the standard EGA resolution of 640 by 350 contains only 46% of
the pixel count of 800 by 600, and even the VGA offers only 64%.
800 by 600 resolution delivers 156% of the VGA's pixel count.
So at this juncture we must leave IBM in the dust. Only enhanced
EGA and VGA adapters are able to generate 800 by 600 pixels, and
only multisynchronous displays can lock onto the extreme
synchronization frequencies required for the delivery of this
stunning and readily available resolution.
The Incredible SONY CDP-1302A...
Steve's Dream Machine Monitor of Choice!
Having decided that Steve's Dream Machine monitor had to be
multisynchronous in order to deliver the most resolution
possible, the next obvious question was: Which one was the best?
After staring endlessly at, and touching and feeling, just about
every available candidate, I determined that no other monitor
comes anywhere NEAR the quality of the Sony "Multiscan" CDP-
1302A. The Sony Multiscan is solidly entrenched as the Steve's
Dream Machine video display monitor. After purchasing several, I
couldn't be more pleased.
The single feature which distinguishes the CDP-1302A from the
crowd, placing it heads and shoulders above the rest, is its
image quality. Based upon Sony's legendary Trinitron color
picture tube, the 1302A packs its primary red, blue, and green
phosphors so closely together that white text actually looks
white, rather than appearing as an ugly island of white fringed
with red on one side, green on top, and blue on the other side.
Coming from the purely monochrome character coloring of
monochrome displays as I did, I just wasn't willing to sacrifice
text color purity for the sake of color. The Sony 1302A is the
ONLY monitor in the industry which doesn't compromise text
appearance for color capability. As I write this column with PC-
Write, I'm staring at white text on a blue background. With my
nose one inch from the screen, aside from being cross-eyed, I
absolutely cannot see anything but white text on a blue
background. No other monitor delivers this quality.
All contemporary color monitors operate through a process known
as "SPATIAL COLOR MIXING." Though from a distance the screen
appears smooth, homogeneous and continuous, it's actually
composed of thousands of individual red, green, and blue
phosphor regions. When the display's electron beams strike the
phosphors from behind they fluoresce and glow in one of the three
primary colors. By controlling the instantaneous voltages applied
to each of the three electron beams at the back of the CRT, the
red, green, and blue color phosphors in the region
where the beams are striking are made to glow in proportionate
brightness.
Our eyes, having somewhat limited resolution, don't see the
individual red, green, and blue phosphors in the region, but
instead spatially mix these colors into a single composite.
(It's rather incredible to realize then that the first thing our
eyes do is to re-separate this composite color back into its
red, green, and blue color levels since our eyes are built from
light sensitive rods and cones which selectively respond only to
red, green, and blue light!)
However, our eye's ability to convincingly spatially mix the
screen's primary colors is a function of the center-to-center
inter-color spacing, which is also known as the display's "DOT
PITCH." Not only does the Sony have a significantly tighter dot
pitch than any other large display in the industry (0.26
millimeters versus 0.31 or coarser for everyone else), but the
Sony's Trinitron'ness seems inherently better suited to the job
of helping our eyes to perform this mixing. It's almost as if
the individual colors are being pre-mixed behind the screen
before leaking out onto the tube's glass faceplate.
This dot pitch also means quite a lot when the monitor is being
called upon to display higher resolution images. As the number
of displayed pixels per inch begins to approach the number of
phosphor dots per inch a strange interaction known as "SPATIAL
FREQUENCY BEATING" occurs. You can most easily see this by
drawing single pixel wide horizontal, vertical, or slanted black
lines against a solid white background. Rather than appearing as
black, the line's width is so much smaller than the surrounding
illuminated pixels that these too-fat pixels bleed their colors
into the supposedly black line, rendering a non-black dimly
colored line. In practice, high resolution black on white
applications such as desktop publishing end up appearing
disturbingly multi-colored rather than pleasingly black on
white. The 800 by 600 pixel resolution which multisync displays
provide at no cost requires the dot pitch to be as tight as
possible.
If you care about your eyes, I urge you to check into the Sony
Multiscan CDP-1302A. This is NOT a place to compromise.
And the Paradise VGA Plus Card,
The Ultimate VALUE in VGA Adapters
Having answered the burning question of the ultimate video
monitor for Steve's Dream Machine with my enthusiastic ravings
about the marvelous Sony CDP-1302A multiscan monitor, the final
question to be answered for our display sub-system project is:
What's the ultimate display adapter?
Determining the correct answer to this question was complicated
substantially by the simple fact that the VGA marketplace is
filled with an incredible degree of clutter, misdirection,
overstatement, and outright lies. What you see and hear is
almost always FAR FAR different from what you actually get. Wild
claims made by VGA adapter manufacturers abound, the ads are
largely full of baloney, and it's quite hard to really know
what's true. It's also quite hard to know what really makes a
DIFFERENCE in VGA adapters, so consequently even the normally
shrewd buyer will wind up guessing.
As my research into VGA adapters progressed, and I learned more
and more, I became increasingly upset by the state of affairs and
committed a disproportionate amount of time and energy to
the task of finding out what's REALLY going on. Getting
underneath the covers to substantiate or debunk various claims
required the creation of special benchmarking software to
directly measure critical adapter parameters such as horizontal
sweep rates, overall vertical refresh rates, and raw low-level
adapter data bandwidths. What I discovered amazed me, and even
though the results of this research may upset some significant
players in the industry, I feel compelled to share what I found.
Since I don't want to tease you any more than necessary, I'm
telling you right up front, here and now, that for my money,
there is no adapter in the industry which delivers more overall
value than the inexpensive, analog-only, 8-bit, incredible
Paradise VGA Plus. Though the VGA Plus is currently in very
short supply, being affected both by its own popularity as well
as by our industry's current dynamic RAM shortage, it's an
incredible value at its current street price of between $230 and
$260.
I urge you not to purchase any other display adapter, VGA or
otherwise, until you've heard me out. Though you might have to
struggle and/or wait a while to find one, it'll be a decision
you couldn't regret.
The various VGA adapters in the industry may be differentiated
by applying the following tests and comparisons: raw low-level
data bandwidth, companion software drivers, display monitor
compatibility, IBM VGA register level compatibility, system-
level hardware compatibility, and to a lesser degree backward
compatibility with prior display standards.
Of all these characteristics, only video display compatibility
and backward compatibility are obvious from the surface. Every
other characteristic must be determined through actual use and
testing. The only negative feature of the Paradise VGA Plus in
this regard is it's total lack of support for the older digital-
only monitors including the original IBM monochrome, CGA, and
EGA displays. You won't be able to use the VGA Plus if you have
one of these, though Paradise has stated that they will make a
version of their card for sale to large OEM customers which will
support both digital and analog monitors. This liability is
shared by the Compaq and Video Seven Fastwrite and VRAM cards,
so the Paradise is in good company. Of course this is no problem
if you already own or intend to purchase any multisync monitor
like Steve's dream monitor, the Sony CDP-1302A.
Almost every VGA adapter in the industry is a so-called "five-
in-one" card. Five-in-one refers to MDA, CGA, Hercules, EGA, and
VGA, and means that such cards can run virtually any software
ever written to any of these major standards. The two notable
exceptions are IBM and Compaq which lack support for the
Hercules standard. Even though Compaq's VGA adapter utilizes the
Paradise PVGA1A VGA chip, and could thus have easily implemented
Hercules backwards compatibility and the useful extended
resolutions as do the Paradise VGAs, Compaq chose not to bring
these features to their purchasers, apparently preferring to
remain more strictly IBM compatible. For this reason, and
considering its high price, you'd have to really love the Compaq
name in order to intelligently purchase Compaq's VGA adapter.
It's a very nice adapter, but the Paradise Plus or Pro do more,
cost less, and are otherwise identical, all being based upon the
same VGA chip.
Display System Performance
It's hardly surprising that the single hottest issue in the VGA
marketplace is performance. People want machines that don't
slow them down, and since our video display screens are the
windows into the souls of our machines, it's only natural to
want a screen that can keep up with the CPU which lurks behind.
Being a performance fanatic myself, the first thing I did was to
write a machine language benchmarking program to determine the
fundamental raw machine-level data throughput of VGA adapters.
As a low-end reference point, the true Blue IBM VGA adapter can
accept text data at 569 Kbytes per second and graphics data, when
in 640 by 480 resolution, at 592 Kbytes per second.
The IBM's raw text throughput of 569 Kbytes per second means
that the entire 4000 byte text screen could be re-written 142
times per second. Since display screens are only displayed 60 to
70 times per second, anything faster than this is completely
invisible and represents wasted performance. The point is, when
displaying a 25 line by 80 column text screen, even the SLOWEST
VGA card on the market (which the IBM VGA is) is twice faster
than is even visible! Those "8 times faster" performance claims
being made by several VGA competitors are based upon their
card's text-mode throughput and are about as useful as a jet
engine on a skateboard. I ignore such nonsense and the companies
behind it.
However, what's true for text mode performance is not
necessarily true for bit-mapped graphics. While an entire text
screen is specified by just 4000 bytes of data, a 16-color 800 by
600 high resolution bit-mapped image requires 240,000 bytes of
data! Even so, IBM's 592 Kbytes of graphics throughput can still
paint an entire VGA image in four-tenths of one second. That
really isn't bad.
So how do the other boards in the market compare? Well any board
based upon the Tseng Labs (pronounced sang) chipset will deliver
approximately IBM-grade performance. Tseng Labs based boards
such as those from Genoa, Orchid, Sigma, STB, and Tecmar have
throughputs of 591 Kbytes for text and 588 Kbytes for graphics,
which is actually a bit slower than IBM. The advantage these
boards have over the IBM is 5-in-1 backwards compatibility.
Unfortunately, this comes with an expense of yawning performance.
Several also utilize the Tseng Labs 1024 by 768 resolution mode.
This requires display screen interlacing which halves the overall
refresh rate and produces completely unacceptable display flicker
when using Ventura or with Window's color mixing scheme known as
dithering. One positive feature of these cards is their full
support for the digital-only MDA, CGA, and EGA monitors, but
since such monitors aren't state-of-the-art anyway, it would be a
shame to choose a poor performing VGA adapter for the sake of
running a poor performing display. For these reasons, I don't
recommend Tseng Labs chip based VGA adapters.
Video Seven has been generating quite a lot of press attention
lately with their FastWrite and VRAM VGA adapters. Having
studied these boards at length with the hope that they would
turn out to be real screamers, I have to admit to being less
than fully impressed. I had significant hardware and software
incompatibility problems with the FastWrite and VRAM boards and
none with any others. Though I've heard that newer revisions
have solved many of the earlier problems, I still feel shy toward
them. Also, the incompatible way their video BIOS was designed
prevents multitasking software from freely and properly
switching tasks between various extended modes. This alone would
keep me away from Video Seven's products.
However, it can't be denied that the Video Seven pair are
uncontested winners when raw throughput alone is considered. In
640 x 480 mode, the FastWrite came in with 1.812 megabytes per
second throughput, and the VRAM delivered a screaming 2.885
megabytes per second.
I was puzzled at this point because my favorite little Paradise
Plus board, with its 1.139 megabytes per second throughput, just
didn't SEEM to be any slower than the VRAM. It occurred to me
that the board's raw throughput was being "watered down" by
"software overhead" which would tend to equalize performance.
After writing a new set of benchmarks to test performance
THROUGH their respective Windows drivers, I found what I
expected. Despite the fact that the VRAM board could accept raw
bit-map data 153% faster than the Paradise Plus, the software
overhead in the Windows drivers resulted in a performance
difference of only 54%! When the application's own overhead was
factored into this, the VRAM edge was even further blunted.
Due to architectural characteristics of the Paradise PVGA1A VGA
chip, Paradise's 16-bit boards actually deliver NO MORE
PERFORMANCE than the inexpensive 8-bit Paradise Plus, Steve's
Dream Machine VGA board.
The Display System Series
Loose Ends
Let's finish our study of the state of the art in IBM video
display technology by tying down a variety of loose ends. As
we've seen, my display adapter of choice is Paradise's 8-bit VGA
Plus. Surprisingly, the architecture of the PVGA1A chip, which
forms the heart of every VGA adapter from Paradise as well as
the VGA systems produced by AST Research and Compaq, gains
NOTHING from a 16-bit bus connector when the boards are used in
their high resolution bit-mapped modes. This means that except
for the additional memory on the Paradise VGA Pro board, there's
absolutely no benefit to purchasing it over the less expensive
8-bit Paradise Plus. In fact, the temptation would then be to
run the Pro card in its 256 color mode, but my benchmarks
revealed that display performance suffers with higher color
counts. This is hardly surprising since additional colors depend
upon the use of additional memory which must be managed by the
driving software.
After declaring the Sony "Multiscan" CDP-1302A to be today's
ultimate video display, I was contacted by many competing vendors
who wanted me to believe that their displays were better. As a
result of entertaining several such possibilities I'm more
certain now than ever that the Sony blows EVERYTHING else away.
As I acquire increasing experience with 800 by 600 resolution,
which you get "free" when the Sony is paired with the Paradise
VGA Plus, I'm becoming more and more certain that it's ultimately
the best general purpose resolution. When running at 800 by 600
resolution, the Sony produces an active image area which is 10
inches wide by 7.5 inches tall. Dividing each of these lengths
into the pixel resolution in that dimension yields exactly 80
pixels per inch IN EACH DIRECTION. This beats the Macintosh's 72
ppi resolution with a much larger screen while delivering the
Macintosh's popular "square" pixels which are exactly as wide as
they are tall. It's nice to have a system on which circles
appear circular and squares really are square!
While I'm thinking about high resolution under Microsoft
Windows, I really need to make sure you know about Micrografx's
incredible Designer product. Designer feels to me like a highly
evolved CAD package with an exquisite state-of-the-art Windows
user interface. Using Designer has become fast and reflexive. It
has that rare easy-to-learn feeling which results from several
generations of detail polishing. While Designer completely
answers my desire for the lightning fast creation of structured
graphics, I've been surprised and delighted to find that several
of my died-in-the-wool traditional "CAD freak" friends have
completely switched to Designer after seeing me mouse my way
around it. If you have any need for PC based drawing, I'd urge
you to take a peek at Micrografx's Designer.
I'm addicted to Ventura Publisher for the creation of all manner
of high grade hard copy, so the quality and legibility of
Ventura's displayed image has profound importance for me. If
you've been reading this column for long, you probably know that
I tend toward perfectionism, always needing to get the most out
of my system. So I've been irked by Ventura's three fixed
display screen zoom factors. At each zoom setting the image is
always either too small, leaving an unused "grey zone" to the
right of the page's image, or too large, requiring a horizontal
scroll to see everything.
Bitstream Inc. has developed and sells a fabulous technology
called FONTWARE which generates any size and resolution of
ultra-high-quality typefaces from a set of sophisticated
typeface outline masters. Since the EGA's pixels aren't square,
the EGA-compatible screen fonts which are shipped with Ventura
aren't specifically tailored for 800 by 600 resolution. So I
decided to used Bitstream's Fontware to regenerate an entirely
new set of Ventura screen fonts with SQUARE pixels, and while I
was at it, to choose a screen font resolution which would give
me EXACTLY the Ventura zoomed sizes I wanted.
After some experimentation, I'm delighted to tell you that I now
have exactly what I want from Ventura. By asking Bitstream's
Fontware technology to rebuild Ventura's screen fonts at 100 by
100 pixel resolution the text of a standard 8.5 by 11 inch page
with one inch margins EXACTLY FILLS the screen in Ventura's
"normal" viewing mode with Ventura's mode selection icons
displayed. The result is an incredibly clear and legible image
in 800 by 600 resolution which puts the VGA's defacto 640 by 480
image to shame.
Micrografx can be contacted about Designer at (800) 272-3729 and
Bitstream can tell you more about Fontware at (800) 522-3668.
- The End -
Copyright (c) 1989 by Steven M. Gibson
Laguna Hills, CA 92653
**ALL RIGHTS RESERVED **
Another file downloaded from: NIRVANAnet(tm)
& the Temple of the Screaming Electron 415-935-5845
Just Say Yes 415-922-1613
Rat Head 415-524-3649
Cheez Whiz 408-363-9766
Reality Check 415-474-2602
Specializing in conversations, obscure information, high explosives,
arcane knowledge, political extremism, diversive sexuality,
insane speculation, and wild rumours. ALL-TEXT BBS SYSTEMS.
Full access for first-time callers. We don't want to know who you are,
where you live, or what your phone number is. We are not Big Brother.
"Raw Data for Raw Nerves"