SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : S3 (A LONGER TERM PERSPECTIVE) -- Ignore unavailable to you. Want to Upgrade?


To: Bill Lin who wrote (7954)11/17/1997 10:04:00 PM
From: Ski  Read Replies (2) | Respond to of 14577
 
Bill, A few points per your post:

-There isn't that much difference between OpenGL and Direct3D from the chip perspective. Different drivers, of course, but the equations are the same equations for most of the 3D pipeline. Some variation, but not insurmountable. Those two APIs define the 3D space for PCs. So, the current direction is fairly well set. Someday, a different paradigm/architecture may come up, but the current one seems safe for the next few design cycles.
-As you note, ZD benchmarks are very, very important. One number is as deep as most people can go. Fortunately, the ZDBop (Ziff Davis Benchmarking Operations) benchmarks are, frankly, quite good at what they do. So, it's not a bad measure.
-You may very well be right on the 1-2 dominant graphics companies. ATI's goal is to be either the one, or one of those two companies. The market could shape up that way. The deciding factor may not be the hottest implementation of 3D in silicon. Rather it may be a combination of support infrastructure and low cost. In that sense, ATI is fairly well positioned, along with being competitive on the design side. Next five years should tell the tale. After that, I'd bet things will start to firm up and the barriers to entry, particularly on the I.P. level, will be too huge for many start ups to try to enter the biz.
-TXN or the Rock, huh? Think they'd buy a Canadian company? hee-hee-hee.

Ski

P.S. See the December "Upside" magazine for an article by Peter Glaskowsky on 3D graphics. Some good future ideas in there too.



To: Bill Lin who wrote (7954)11/18/1997 2:24:00 PM
From: Bill Lin  Respond to of 14577
 
phil, ski, jan, etc.

neat site:
tomshardware.com
tomshardware.com

Introduction

I receive a lot of mail nowadays from people who don't know
which graphic accelerator card to get and I can understand
that there are a lot of choices for all kind of different needs.
There's a lot of hype thrown at us from all the different card
and chip manufacturers on the graphic market too and you
can easily face a huge disappointment if you should make
the wrong choice.

The difficulty in choosing the right video accelerator card
comes from the different needs we have for this piece of
hardware. As usual we'd prefer getting a card that can do
everything at an excellent level and this if somehow possible
for a low price as well. However the miracolous cheap
all-round card isn't out yet and I guess that it will possibly
take forever until all our needs will be pleased. Hence we
have to make our mind up what is most important to us and
also how much money we are willing to spend.

The first question we have to ask ourselves is if we will use
our system mainly for professional work or mainly for games.
Most professional cards are not great at games and vice
versa. If you've already got a video card in your system, ask
yourself if you're pleased with its 2D performance at
professional work and if you just want to purchase an add on
card for games. In this case you still have the professional
performance of your current video card and add some real
good gaming performance with the add on 3D card. You will
need an additional PCI slot though.


Considerations for Gamers

In case gaming is most that you do on your system and you
couldn't care less for Windows NT, true color and OpenGL,
you want to go for a pure 3D gaming card or get an add on
card.

Direct3D or Proprietary 3D Engine?

You'll now have to decide what kind of games are important
to you. Currently the graphically best games are often
designed for a special graphic chip, or at least they look best
with this one chip. The number one supported 3D graphic
chip is nowadays the 3Dfx Voodoo, found on add on cards
like the Orchid Righteous3D, Diamond's Monster3D and
several others. It looks as if upcoming games will still support
this particular chip and since the Voodoo 2 is already on the
horizon, you can expect 3Dfx's 'Glide' engine staying
supported by many games for a long time. Alternatively to a
special 3D chip support, many new games are using
Direct3D's new features quite heaviliy, so that it depends on
how well the 3D card's driver translates Direct3D to their
proprietary engine. PowerVR's PCX1 and PCX2 chips are
quite powerful 3D chips, but the cards that use them are
highly incompatible. I've seen only very few games that run
on this chip properly. If the PCX engine is used directly, the
games look awesome though. The only 3D chip to my
knowledge, that doesn't have a dedicated 3D engine, but is
using Direct3D as its API directly, is NVidia's RIVA 128 chip,
currently the fastest Direct3D chip available on the market.
The RIVA 128 is wonderful for Direct3D games, but games
that are only supporting a bunch of proprietary 3D engines
will not run on the RIVA 128. The future will bring almost any
game in Direct3D, which will help NVidia's RIVA a lot.

3D Performance

It is not easy to measure pure 3D performance, because
there are so many different ways a 3D engine can be used.
Most official benchmarks are using the Direct3D engine of
DirectX, like e.g. ZD's 3D Winbench or VNU's Final Reality.
These benchmarks can only show you the card's Direct3D
performance, hence how well the driver translates Direct3D
into the chip's own 3D engine. NVidia's RIVA 128 doesn't
need this 'translator', it uses Direct3D as its own API. This is
only one reason why the RIVA scores by far best in Direct3D
benchmarks. However some games written for that specific
3D engine of a chip can run much faster than the 3D
Winbench score would let you expect them to. VQuake for
Rendition's Verite 1000 is one good old example. The Verite
1000 was never scoring well in 3D Winbench, but VQuake
looked good and ran fast.

3D Quality !

Now 3D performance is only one thing, 3D quality is another.
There are a lot of 3D features used nowadays, most of them
supported and used by DirectX 5, but there will be even more
3D features implemented in DirectX 6. A 3D chip can only
support a special amount of 3D features, others are either
not supported at all, or special drivers are used that emulate
these features. In my latest test I came across only one chip
that supports virtually every current 3D feature properly and
this is 3Dfx's Voodoo chip. The big let down of the Voodoo
chip leads to the other aspect for quality, the 3D screen
resolution. The Voodoo chip can only do 640x480 in case of
2 MB frame buffer memory (4 MB cards), as in most of the
Voodoo cards, or maximal 800x600 in case the card comes
with 6 MB RAM (e.g. Quantum 3D Obsidian 100SB) , 4MB
hereof as frame buffer. NVidia's RIVA 128 chip has got a
simular problem, it can't support more than 4 MB onboard
memory, only good for a 3D resolution of maximal 800x600.
Now it doesn't have to be that bad, since we are quite
pleased with our good old television as well, which has a
lower resolution than 800x600. The 3D chip and the system
CPU have to be powerful enough for running smoothly at this
resolution as well. However, I've seen 'Forsaken' at
1024x768 on a PII 300 with an ATI XPERT card and it looks
pretty awesome.

How Powerful is Your CPU?

Some 3D chips are taking a lot of workload from the CPU,
others want decent CPU performance for its operation.
PowerVR's PCX chips want at least a Pentium MMX 166 for
decent quality, 3Dfx's Voodoo lets games run fast even on
systems with weak CPUs and Rendition's new Verite
2100/2200 chip gives a huge improvement to slow CPUs,
but fast CPUs are reaching its limit and don't really benefit of
this chip anymore. NVidia's RIVA chip seems to scale
linearily from 6x86 CPUs up to Pentium II CPUs. Under
Direct3D its always the fastest chip.

Price !

Another thing you obviously want to take in consideration is
the price you've got to pay for the card. Many cards that
have good 2D performance as well are pretty expensive.
This is often due to the more expensive memory they are
using, but it could also be the additional features like e.g. TV
out or video compression. Cards with more memory are also
more expensive, but they offer higher resolutions in 3D,
higher color depth and higher resolutions in 2D. Make sure
you don't pay for something you won't need.


Considerations for Professionals

Picture Quality

If you are working on your computer professionally one of
the most important things is the picture quality. This is
achieved by a high quality and high clocked RAMDAC. Most
of the new graphic chips have included the RAMDAC
internally, thus saving cost, but the best picture quality is still
produced by an external RAMDAC. The most popular cards
with external RAMDACs are Matrox Millennium I and II and
Number Nine's Revolution 3D. These cards are still offering
you the sharpest and cleanest picture on the screen. If you
have got an expensive monitor, you want to use the high
refresh rates your monitor supports. As a simple rule you
should at least have a refresh rate of 85 Hz available for all
the reolutions you want to use. Refresh rates of 120 and
more sound nice, but they won't give you much of an
advantage anymore. Responsible for this is again the
RAMDAC. The higher its clock rate, the higher are possible
refresh rates.

2D Performance

The 2D performance was what used to determine the quality
of a graphic card in the past. Now 2D acceleration seems
pretty close to the limit and almost all cards are offering a
good 2D performance, at least at 16 bit color modi. Good 2D
performance in true color is a virtue that's pretty rare still
though. Matrox and Number Nine always used to fight about
the 2D performance crown and it hasn't changed much still. If
you are working really professionally at your computer, you
can impossibly use Microsoft's mouse driver collection called
Windows 95. Hence you are either using Windows NT or
some really good OS that's not from the monopolist. NT
drivers are very important for professional cards and the NT
performance should be more important than the Windows 95
performance. There is often quite a bit of a difference
between NT and 95 performance.

OpenGL, Heidi

For people that use a real graphic workstation with CAD
and/or 3D rendering, SGI's OpenGL as well as Heidi are of
major importance. Nowadays if you hear 'PC' and 'OpenGL'
one company comes into your minds ... 3DLabs. I will not
discuss the real high end chips of 3DLabs, since this is off
topic on this website, but 3DLabs' new Permedia 2 chip is
one of the most impressive graphic chips on the market
today in my opinion. For 3DLabs the Permedia 2 is nothing
but a low end chip for the mass market, but amongst its
competitors it's quite a gem in terms of professional work. In
mid to higher prized systems cards with the Permedia 2 will
offer you the best OpenGL performance combined with a
good 2D and a fairly impressive Direct3D performance.


AGP or PCI

Since the new advanced graphics port (AGP) has been
released, you may wonder what interface to go for. I think
you should certainly get a board with Intel's new LX chipset,
in case you want to use a Pentium II system. In this case
you won't do much wrong going for an AGP card. However
please be aware that there's currently hardly any
performance advantage of AGP over PCI. In case of Socket
7 there's still the problem that there aren't many AGP
motherboards available yet, but this will hopefully change
soon, now that the PA-2012 was released.