Lockheed Real3D demo footage [Pro/1000 ?] from Comdex 1997

[Megadrive1988]

Look what I've found - Pro 1000 developer's guide.
http://web.archive.org/web/19980610062119/http://www.real3d.com/support/pro/devguide4.pdf

Wow, very cool find.

 

[Putas]

Siggraph 96: http://youtu.be/BI_puH7gYR8

It really was not until the GeForce 256 and GeForce 2 GTS / GeForce 2 Ultra came along in 1999/2000 that PC graphics cards rivaled Model 3. The major shortcoming of Model 3 was its low resolution.

It really was something given that the heart of Model 3 (two R3D/Pro-1000s) is really 1995 technology which powered arcade games that only started getting released. in the 2nd half of 1996. It took PC 3D hardware about 4 to 5 years to catch up.

By rivaling you mean completely eclipsing in every possible way? Is it because of the 'T&L'? Do we know if their Geometry Processor was at least functionally equal to the T&L? It is not something CPUs are terrible bad at, and PowerPC was cheap FP king. So how big is the benefit in multichip arcade machines? 100 MFLOPS for geometry looked strong in 1996 game technology, by 1998 it was low end. Do you now if Model 3 chips were actual Pro-1000 and not derivatives? Claimed polygon rates of Real3D Pro chips are very very high, but games were using still what, >10,000?
I think you should consider i740 as a balanced (geometry to fillrates) integration of Model 3 engine of effectively similar speed. Of course some of the performance is lost in PC environment.

 

[swaaye]

I just read yesterday that Model 3 Pro 1000 chips were derivatives in some way.
http://www.supermodel3.com/Forum/viewtopic.php?f=7&t=313


By the way Putas, I rediscovered your Vintage3d site the other day and have read most of it. Very insightful and interesting.

 

[Laa-Yosh]

Damn, I feel nostalgic and old at the same time

 

[Megadrive1988]

I just read yesterday that Model 3 Pro 1000 chips were derivatives in some way.
http://www.supermodel3.com/Forum/viewtopic.php?f=7&t=313

Yeah no doubt. It's not as if Real3D just stuck a couple of stock Pro 1000 GPUs into a board along with a host PPC 603 CPU and some RAM and VRAM/VROM for Sega and called it day. Although it wasn't flat out stated by publications at the time, it was pretty clear the solution was customized.

The Model 3 board was originally supposed to be powering 3 arcade games due for release in Japan by end of '95 It suffered delays. Correct me if I'm wrong but Model 3 wasn't shown off to the public until sometime in late Q2 1996. Virtua Fighter 3 wasn't released until something like late Q3 or maybe even Q4 1996 in Japan. I'm not sure but I don't think Scud Race / Sega Super GT released before the end of the year in Japan and I personally did not see VF3 in the Chicago area until the very beginning of '97.

I have no idea what specific changes were made to Pro 1000 for Model 3 but one thing was probably the clockspeed. I think the later Step 1.5, 2.0 and 2.1 versions used more raster unit / modules.

 

[Flack]

Hello, I am the person who uploaded the footage Megadrive1988 found on Youtube. I was searching Google for some more information about Real3D, which led me to this post. It's a small world, and all that.

In 1997 my buddy and I (we were both LAN administrators at the time) convinced our bosses they should send us to Comdex to learn about new network technology. As you can tell from the footage, we spent the majority of our time checking out new gaming technology instead.

In 1997 I didn't own a digital camera, but I did own a small camcorder, which I took. I also owned a Snappy, a parallel port device that allowed you to capture still frames from a video source. My plan was to shoot video at the show, pull digital stills from the footage, and then share the pictures online. Who could have imagined that 10 years later it would be just as simple to share the video?

That Real3D demo was one of the most amazing things I have ever seen in person. The footage (especially in the space demo) almost looked like they had shot footage of a miniature and were overlaying videogame graphics (the shoddy explosions you see) over the top of it. At one point in the footage you can hear my friend say "it's skipping." Every now and then the video would skip or jerk, which led us to believe at the time that it was being played off of a laserdisc or DVD.

I hope you guys found the footage useful or at least interesting. I've been dumping all my old camcorder tapes to AVI and thought that someone might get a kick out of seeing that one from Comdex. Looks like I was right!

 

[Putas]

The explosions are animated textures of low framerate, common at that time. The shooter could be just as well interactive movie, can't tell if they are in control of movement.

 

[Megadrive1988]

Hello, I am the person who uploaded the footage Megadrive1988 found on Youtube. I was searching Google for some more information about Real3D, which led me to this post. It's a small world, and all that.

In 1997 my buddy and I (we were both LAN administrators at the time) convinced our bosses they should send us to Comdex to learn about new network technology. As you can tell from the footage, we spent the majority of our time checking out new gaming technology instead.

In 1997 I didn't own a digital camera, but I did own a small camcorder, which I took. I also owned a Snappy, a parallel port device that allowed you to capture still frames from a video source. My plan was to shoot video at the show, pull digital stills from the footage, and then share the pictures online. Who could have imagined that 10 years later it would be just as simple to share the video?

That Real3D demo was one of the most amazing things I have ever seen in person. The footage (especially in the space demo) almost looked like they had shot footage of a miniature and were overlaying videogame graphics (the shoddy explosions you see) over the top of it. At one point in the footage you can hear my friend say "it's skipping." Every now and then the video would skip or jerk, which led us to believe at the time that it was being played off of a laserdisc or DVD.

I hope you guys found the footage useful or at least interesting. I've been dumping all my old camcorder tapes to AVI and thought that someone might get a kick out of seeing that one from Comdex. Looks like I was right!

Indeed Flack, I found it very interesting. Thanks for uploading your footage to Youtube.

I also think it's pretty incredible that those demos had to have been made way back in 1995, as Next Generation had coverage that year.

 

[swaaye]

The footage reminds me of the demos SGI showed for N64's hardware. Except with Real3D the games actually ran on the demoed hardware. SGI demoes were only accurate from a basic featureset angle.

I get all tingly thinking about how hyped up I was for N64 and 3DO M2....

 

[Megadrive1988]

The footage reminds me of the demos SGI showed for N64's hardware. Except with Real3D the games actually ran on the demoed hardware. SGI demoes were only accurate from a basic featureset angle.

My thoughts exactly.

[

I get all tingly thinking about how hyped up I was for N64 and 3DO M2....

Was pretty hyped for N64, really hyped for M2.

 

[Stiletto]

Look what I've found - Pro 1000 developer's guide.
http://web.archive.org/web/19980610062119/http://www.real3d.com/support/pro/devguide4.pdf

Hah, MAMEDEV's and other Model 2/3 emu devs had a copy of that for years. My backup dates from January 2004, it may even have been before that. Anyhow, there's a few more PDFs there, but they're not that interesting. Example:
Pro-1000 Multigen Database Developer's Manual 1.1e Feb. 1997
Pro-1000 Product Description 3.1e Nov. 1996
Pro-1000 Product Description 3.1 Dec. 1996
Pro-1000 Product Description 4.0e Dec. 1997
R3D-100 Architecture Overview 2.2 Nov. 1996

BTW, there's other footage on their site. I forget which of these work and don't, but I archived a few off there months ago in order to upload to Youtube for Megadrive1988 to obsess over.

http://web.archive.org/web/19970102122401/http://www.real3d.com/images/spacec1a.mov

http://web.archive.org/web/19980610030421/http://www.real3d.com/cool/pyramid.mov

http://web.archive.org/web/19980610030511/http://www.real3d.com/cool/town.mov

http://web.archive.org/web/19980610030558/http://www.real3d.com/cool/water.mov

http://web.archive.org/web/19970102121507/http://www.real3d.com/images/spacec1a.avi

http://web.archive.org/web/19980610030135/http://www.real3d.com/cool/pyramid160.avi

http://web.archive.org/web/19980610030226/http://www.real3d.com/cool/town160.avi

http://web.archive.org/web/19980610030333/http://www.real3d.com/cool/water160.avi

http://web.archive.org/web/19970102115515/http://www.real3d.com/images/dbayou1m.avi

 

[motke]

Hah, MAMEDEV's and other Model 2/3 emu devs had a copy of that for years.

By the way, it would be superb if you could share with the world that Model 2 tech manual that you guys have. Maybe in the future, if it's not appropriate to do so now. Also, if you don't want to post it here, it's enough to quietly leave it at some googleable site.

 

[swaaye]

I've been messing with a Riva 128 card lately and reading about its various shortcomings and secret design compromises that make it render in its special fugly way. Intrigue such as the seemingly fake per-pixel mip mapping box in the driver control panel....

I stumbled upon a giant Real3D interview from the ancient times. I figured Megadrive1988 and others would probably enjoy it. A lot of interesting stuff. Sega, i740, competitors and more.

Hardcore Gaming Exclusive Interview with Real 3D
Part I

The folks at Real3D were kind enough to take some time to answer a few
questions for us. First I would like to thank Shog for helping out
with some of the questions. While we couldn't get them to talk about
Model4, they were very open and willing to talk about Starfighter,
3DFX, Power VR and the state of 3D. So enough with the yapping from me
here is part one of our 1 on 1 with Real3D

HGN: Given the nature of your relationship with Sega, we'd like to
know if Real3D was ever approached to work on Sega's next home system?
If so what happened? If not, than why do you think you weren't asked,
especially considering the work that you've done with their arcade
boards.

Real3D: We had some preliminary discussions with Sega regarding the
home console market, but both companies agreed the business model for
the console market didn't create a win-win situation like we have in
the arcade market. The two markets are very different and the
companies decided not to pursue any developments in the console market
space.

HGN: Right now it seems like the 3D card race is getting very crowded.
Why is it that none of the card makers have made any real attempt at
the business model? You know using the cards for presentations and
such. They all seem aimed at the same market (games) and the one how
could make a serious dent in the corporate side of this would have a
big advantage.

Real3D: The simple truth is that business productivity software that
takes advantage of the 3D hardware capabilities is just not available
yet. However, we see this changing over the next six to eighteen
months. Real 3D certainly wants to see the corporate market start
demanding 3D graphics hardware. We are starting to see some potential
in applications such as Viewpoint's LiveArt, which is 3D clipart that
can be inserted in presentations. Real 3D has established our
RealPartners developer program, which is a program that gets software
developers early specifications and releases of our hardware. This
way, the software developers can develop 3D content that takes
advantage of the 3D hardware capabilities and helps make 3D in general
more widespread.

HGN: I'd like to talk about the competition a bit. What would you say
are the strengths and weaknesses of the following chipsets: Voodoo 2,
PVRSG, and Riva 128? How would you rank the Starfighter against them?

Real3D: The Voodoo2 chipset is certainly an impressive alternative for
the serious gamers out there. It seems to be well-matched for QuakeII
and similar Quake-based games, beating the frame rates we get with the
StarFighter and everyone else gets with the various chips/boards.
Performance under Glide-based games is also impressive. But when you
turn to Direct3D game performance, the StarFighter/Intel740 is about
on-par with the Voodoo2. The image quality of the Voodoo2-based boards
leaves little to be desired, and we believe the StarFighter definitely
holds its own in the image quality match with the Voodoo2. But the
higher cost of the Voodoo2 boards , and the lack of an integrated 2D/
3D solution, makes their board a lesser value to the average consumer
and commercial user in our opinion.

We don't want to comment much on the PVRSG, having not seen it. The
press it has gotten is certainly impressive. But we can remember being
similarly impressed with their PR campaign on the PCX1 and PCX2, which
turned out to be pretty unspectacular compared with modern parts.
We'll hold our judgment until we see the PVRSG in action.

The RIVA 128 is a fair piece of silicon; however, it comes with
annoying compromises we never would have made. It has serious
rendering flaws, as we're sure you've seen and heard descriptions of.
Polygon cracks and Z-buffer poke-throughs point to design errors in
the silicon which simply can't be patched with drivers. We never would
have resorted to the "texture mip-level per polygon" simplification
they chose to implement, as we had been using per-pixel LOD
determination for over 15 years in computer image generation solutions
for the military and Sega. The RIVA's rendering speed on Direct3D apps
is no better than 10-15% better than StarFighter/Intel740, at best,
even though they tout their 128-bit bus and 100 megapixel throughput.
In our estimation, if the average consumer can live with these
rendering anomalies, the RIVA boards are reasonably priced.

However, in our humble opinion, mainstream consumers do not have to
sacrifice image quality for performance with the StarFighter. You get
the best combination of both performance and image quality with the
StarFighter/Intel740. We have attempted (some say to a fault) to bring
high fidelity, high performance graphics technology to the mainstream,
and believe the StarFighter/Intel740 offers greater value to the
informed consumer.

HGN: What is the biggest advantage to supporting D3D instead of coming
up with custom API for the Starfighter? You know, something like
Glide?

Real3D: The biggest advantage to D3D and OpenGL are that they are
standard APIs that all hardware and software developers can relate to
and develop for.Custom APIs are somewhat of a nightmare for software
developers because they are forced to incorporate many special
features in their code. From the hardware side, we want as many
applications as possible to be able to take advantage of the
StarFighter/Intel740. The best way to do this is to use standard APIs.

As you are aware, the industry is moving very quickly to standards. It
used to be that software vendors could get 10-15% more out the
hardware by writing to a custom API. Today, Microsoft has much more
competitive code and the software resources that used to be applied to
APIs can now be applied to making the games better. For this reason,
you are seeing more and more software developers supporting the
standard APIs and de-emphasizing the custom APIs.

Glide came out at a time when there were no graphics standards in the
commercial marketplace. Now that all new hardware and software coming
out are supporting either D3D or OpenGL (or both), we think you'll
find that hardware-specific APIs such as Glide will become less
pervasive. This is not to say that D3D or OpenGL are better or worse
at this point, but the cost and complexity in continued support for a
hardware-specific API will be prohibitive.

HGN: PowerVR has Z-buffer-less infinite planes, Voodoo has separate
pixel and texel chips, Riva has 128 bit wide data pipeline, What does
R3D have in Starfighter/Intel740 that give it a competitive advantage
over other 3D architectures? In another words, what is it's
technological trump card? How do you exploit it?

Real3D: In short, the trump card is "AGP Done Right." There are many
AGP cards on the market that are basically PCI cards running across
the AGP bus. These cards will take textures out of AGP memory and copy
them to local memory on the board in order to use them. Even with the
bandwidth of the AGP bus, at some point there will be a bottleneck in
the size of the textures that can be accessed without slowing the
application down. This is why you will find that most cards only
support texture maps up to 256x256 (or maybe slightly larger).

The StarFighter, on the other hand, has no local memory on the board
for texture storage. All textures are accessed directly out of AGP
memory on a texel by texel basis. When an application is running, the
texels that are needed to render the current frame are copied from AGP
memory directly to the on-chip cache to render the image. There is no
wasted copying of the entire texture map. This allows the StarFighter
to run with texture maps up to 1024x1024 with no performance hit. It
also allows the StarFighter to run with huge quantities of texture
(i.e. 40MB or more).

What's all this mean for the gamer? Game developers will be able to
use bigger textures and lots of them (and we're encouraging them to do
so). This basically gives you more stunning and realistic visuals and
an overall better gaming experience, with no performance degradation.
We've seen lots of companies claim "arcade-like experience on the PC",
which we get a chuckle out of. Well, we know a little about both and a
true AGP implementation like the StarFighter/Intel740 actually does
approach an arcade-like experience.

HGN: Starfighter/Auburn's raw polygon output performance and pixel
fill-rate is about on par with the first generation Voodoo (Voodoo1)
chip and is far below the output of the current generation of Voodoo
chip (Voodoo2). Knowing that fact, is it safe to assume that the high
end 3D accelerator market was not what R3D and Intel was going after
with Starfighter/Intel740? Then what market are you going after with
Starfighter/Intel740?

Real3D: The Intel740 was designed as a performance part for the volume
mainstream PC market. The comparison to the Voodoo is somewhat of a
misleading comparison. The Voodoo boards are aimed at a narrow market
and even 3Dfx acknowledges its Voodoo chip set is narrowly focused to
meet the avid gamer market. The Voodoo chip set from 3Dfx is an
excellent product and offers high performance, but it is a 3D only
solution that carries a premium price tag. The StarFighter board with
the Intel740 is a complete, integrated 2D/3D/video board aimed
squarely at the mainstream market. With the StarFighter, you do not
need any additional graphics accelerators for 2D or 3D in your system.
The StarFighter is ideal for a range of markets and users, including
corporate desktop configurations, small business desktop systems,
small office/home office systems, computer game enthusiasts and entry-
level workstation systems.

HGN: What would you consider the optimum system to run the Starfighter
on? What would push it to its fullest potential? Given how spec crazy
everyone is these days, will you have a chip that can compete with
Voodoo2 and PVRSG in polygon output and pixel fill-rate in the near
future?

Real3D: 3D graphics performance is highly correlated to system
performance, for all graphics cards. To get the most out of the
StarFighter, put it in the latest and greatest Pentium system you can.
How about the 700MHz Pentium Intel just demonstrated at CeBit in
Germany? Since this is not an option for most of us, the StarFighter
AGP will perform great in any AGP system. And for those without an AGP
system, the StarFighter PCI will give you comparable performance in
your Pentium, Pentium PRO, or Pentium II system.

We can't give you any specifics about future developments, but let's
just say we're not sitting still with the Intel740.

HGN: How did it come about that you ended up working with Intel? What
exactly is R3D's relationship with Intel regarding the Starfighter/
Intel740 chipset? How much design influence did Intel have on the
project?

Real3D: There are several factors that led Intel to partner with Real
3D in developing the Intel740. First, Real 3D has a great deal of
experience in developing high performance, high fidelity image
generation and 3D graphics technologies. We've done so for the
military and space program for over 30 years. Second, Real 3D has a
very successful relationship with Sega for developing arcade graphics
chips and boards. Finally, Real 3D holds almost 40 patents related to
3D graphics technology. All of these factors led Intel to the decision
that Real 3D would be the ideal partner in bringing very high-end 3D
graphics technology into the mainstream.

The development of the Intel740 was truly a team effort between Intel,
Real 3D, and Chips & Technologies (who did the 2D core). Real 3D was
responsible for system level definition, hardware design, and 3-D
pipeline simulation. Real 3D assisted Intel with validation of the
Intel740, which included system design of the 3-D pipeline and
development of the test suites. Real 3D also led the development of
all driver software for the chip.

Real 3D's relationship with Intel was further cemented earlier this
year when Intel purchased a 20 percent equity interest in Real 3D
(Lockheed Martin still owns the other 80 percent).



Part II

Thanks for checking back in. This is the second part of HGN's 1 on 1
with Real3D. We just touched on a few things regarding Sega and
Voodoo2 in this portion of the conversation.

HGN: Starfighter and the Intel740 has already been acknowledged by
many as having very high quality textures in it's architecture. What
is so special about the R3D architecture that allows this?

R3D: First, StarFighter/Intel740 is AGP done right. With textures
stored in system memory, there is virtually unlimited texture memory
so what you see on the screen is what the author intended, not some
scaled down MIP level. Second, using AGP for texturing frees up local
memory bandwidth for drawing. Our efficient chunking engine makes the
most effective use of memory and internal caches. Third, the
hyperpipelined architecture includes algorithms from our military and
Sega work - algorithms that are tried and true. Combine these with
great pixel accuracy and what you see is what a great image. Also, we
don't want to underestimate the inherent nature of our engineers to
pay strict attention to detail. Remember, our heritage is in
developing simulations that had to be incredibly realistic to meet
military specifications. Experience counts for a lot in many things
and 3D graphics is no different.

HGN: Ok I'm glad you touched on this because I wanted to ask about it.
I'm talking about the scalability of hardware or software. What I mean
is, is it the software's job "to scale up" to the hardware? Or is it
the hardware's job to scale up to the code of the software? Thats one
of the reasons I asked about custom API's. I'll use Moto Racer as an
example. In the D3D version of the game the smoke is like a patch of
white, while in the Power VR and 3DFX versions its more transparent.
Not being a developer I would think that it was because of D3D. Then
you see a game like Incoming running D3D and you ask what gives? Why
such a difference in the quality of the effects? Shouldn't the
hardware ( if it can do the effects like transparent smoke) recognize
the code and "scale up" or if the software sees that the hardware can
do the effect properly, execute it better? Am I making any sense? I
guess I'm asking where does the responsibility lie? Hardware,
software, or both? People come down hard on Power VR but I've seen it
do some really incredible graphics. Likewise I've seen some real bad
stuff on 3DFX.

R3D: The short answer is that it's both the hardware and software
responsibility to make up for the other ones deficiencies. The
hardware itself does what it does and that's it. There is flexibility
within the hardware drivers, but ultimately the driver just receives
individual polygons with no knowledge of the scene that they are a
part of. The application on the other hand has complete knowledge of
the scene and also of the hardware it is running on (by checking the
capability (CAP) bits at startup).

The comparison of MotoRacer to Incoming is a good one. They both use
D3D, but yet even on the same hardware the smoke in MotoRacer looks
horrible and the effects in Incoming look great. This is due to the
quality of the art work in the texture maps and the game developers
ability to effectively use D3D. (or any API) For example, there are 15
different alpha blending modes in D3D. A good developer knows which
modes can best produce the desired effect. Then if the hardware can't
support the selected mode the app should default to something the
hardware can support. (at the expense of the image quality)
So to summarize: visual quality is mostly due to good texture design
and effective use of D3D/OpenGL/etc. But of course the more features
the hardware supports the easier it is for the application to look
good. (this is why PowerVR often gets blasted. Lack of features, like
most of the alpha blend modes.

HGN: I know the Starfighter supports D3D and OpenGL, but how long
before we start seeing titles that are designed to exploit the i740
chipset? Is that where RealPartners would come in? The differences in
the way D3D is used in games like Moto Racer and Incoming is drastic
to say the least. How do you make Incoming the standard?

R3D: Yes, RealPartners is exactly the answer. To exploit the i740, we
encourage developers to support high screen resolutions, use high
resolution texture maps, use lots of texture, etc. We also talk to
them about how they could most effectively take advantage of their
target platform and D3D or OpenGL. For example, a lot of extra speed
can be obtained from the PII processor by keeping your data properly
aligned for the CPU's cache. And with AGP systems there is no longer a
need to use sub-maps for textures. And the list goes on.

HGN: How exactly did LM go from doing simulators for the military to
making arcade boards for Sega? What made you decide to go that route?

R3D: In 1991 we started to look for commercial adjacent market
applications of our Compu-Scene real-time 3D graphics technology.
Market surveys of various industries indicated a growing need for more
realistic video graphics in entertainment applications. This analysis
led to the initial contact with Sega of Japan. Early discussions
revealed that Sega was about to embark on a two-year internal
development effort to improve their polygon based arcade graphics
hardware ( Model 1) with texture mapping so that video screens would
present images which were more realistic. By adapting our proven real-
time 3D graphics technology specifically for Sega's needs, we were
able to get Sega to the market 14 months early.

HGN: On the subject of Sega, how did you work with them on the Model X
boards? Did they come in with ideas or specs that they wanted or did
you just show them your product? Also what can we expect from Model 4?

R3D: Our relationship with Sega is very much a joint effort where both
companies discuss ideas, market requirements, specifications --
basically everything that will go into producing the world's best
arcade games. Since the graphics chips and boards we do for Sega are
custom for Sega, we really couldn't just show them our "product",
because it's not a product until the specs and requirements are set.
When we first started with Sega, what we showed them was the
sophisticated visual systems we had done for the military. Sega
believed (and we believed) we could adapt that technology to the
arcade and if this could be done, the level and sophistication of
arcade graphics would take a quantum leap forward. Well, as you know,
we were able to adapt the technology, which resulted in Model 2 and
later Model 3. And Sega continues to enjoy a market leader position in
the arcade space.And you ask about Model 4. I know you and your
readers are probably anxious to know what's on the drawing board, but
I really can't comment on Model 4 or future arcade development efforts
with Sega. Hope you understand.

HGN: Is there any real advantage to adding an additional 3D card to
run with the Starfighter? I mean if you have a higher quality graphic
to start with, would that take care of the short comings of the other
3D cards? So if you were to take a Starfighter card and matched it
with something like the Voodoo 2, which from what I heard is just more
of Voodoo 1, would it get you the best of both worlds or would they
cancel each other out so to speak?

R3D: The Voodoo 2 is pretty much just more of Voodoo 1. Basically it
is a low-res 3D-only card that has the highest fill rate of anything
currently on the market. What does this mean to the user? It means
that any application with a high depth complexity or that uses
multiple textures per polygon will still run fast. (i.e. Quake II) For
all other applications the Voodoo 2 is comparable to StarFighter or
other top-performing 2D/3D boards for raw performance. What you don't
get with Voodoo 2, but do get with StarFighter is the ability to run
2D, 3D up to 1280x1024, 3D in a window and video. You also only need
to spend $200 for your complete graphics solution versus $250 for a
Voodoo 2 plus a 2D card to go along with it. So, if you spend a lot of
time playing games like Quake II and cost is not much of a factor,
then you might consider a Voodoo 2 to go along side your StarFighter.

HGN: On the subject of Voodoo2, do you see any real advantage to
running two of those things at the same time? That seems to be a big
sell point for the chipset. Is that something other companies would
see as a viable video solution?

R3D: Having two Voodoo 2's gives you an even higher fill rate than
having one. In an application like Quake II, where there are two
textures per polygon, each Voodoo 2 card would process one of the
textures simultaneously so that the scene can be rendered in one pass
instead of two. There aren't many applications (if any) on the market
that are pushing high enough fill rates to take advantage of this. All
it does for Quake 2 is push the frame rate from 60+fps to 100+fps. To
us, when you start exceeding 60fps, it's doubtful the user would be
able to tell the difference and the cost for the two board combo is
around $500.

HGN: A while back the R3D/100 was announced as a consumer product, and
then we kind of never heard anything else about it. Whatever happened
to that product? Is it related in anyway to the i740?

R3D: The R3D/100 was a graphics chip designed for the high-end
workstation markets such as CAD and 3D modeling. This product was not
really related to the Intel740 because the i740 is targeted at the
performance mainstream PC market. As a company, Real 3D has decided to
focus its chip-design efforts in the mainstream PC market through our
co-developments with Intel and our own designs (in addition to the
custom work we do for Sega). The primary business model for Real 3D is
as a board company. We are still involved in selected business
opportunities with the R3D/100, but it is not a product we are
actively marketing any longer.

HGN: Given that Intel is Intel how long do you think it will be before
the i740 starts showing up on Intel's motherboards? That would be a
huge advantage to Real3D in the "3D wars". How would you avoid all the
criticisms that S3 took by being on the motherboards?

R3D: I really try to answer whatever questions are posed to me by
reporters, but for this one, I need to defer to my colleagues at
Intel. I hope you understand, but Intel's efforts in the graphics
market has been and continues to be scrutinized by the Federal Trade
Commission. For this reason, any decision to bring the i740 down onto
motherboards will be watched closely and I'd rather have Intel comment
on the prospect of this happening. As for the second part of your
question, I'll speak for Real 3D from strictly a business perspective
and that is we certainly have a vested interest in the widespread
acceptance and sale of the i740 because of the royalty stream it
generates for Real 3D. And I think if/when the i740 moves onto
motherboards, there will be other performance mainstream graphics
chips offering much more so consumers will still have the choice of
basic graphics through the chip on the motherboard or better graphics
through an add-in card.

HGN: I think I recall hearing about a Model 2 game called "Hummer"
being developed by Real3D. Can you give us a little information about
this title? Also is this an area that you are interested in getting
into more in the future?

R3D: The Model 2 game you speak of called "Hummer" is actually called
"Behind Enemy Lines". This is a title developed by Real 3D -- not only
the hardware (which is in all Model 2 and Model 3 games), but also the
software. It is the second game we have developed for Sega (Desert
Tank was the other one). One point I would like to make is that Real
3D is not in the business of developing games or game software. We
have done a couple at the request of Sega, but Real 3D is a graphics
hardware provider and the primary business of the company is designing
chips, boards, and high-end graphics systems.

HGN: Where do you see 3D going? Its got to be more than just an
increasing polygon count, which seems to be the current trend.

R3D: 3D is a lot more than polygon count - it's realism. That means
adding new features and improving quality in addition to more polygons
and higher fill rates. The next generation of 3-D graphics
accelerators will enhance pixel fidelity with greater color and z-
buffer depths, as well as offer higher quality texture and blending
modes and improved illumination and depth effects. Realism will reach
new heights by the use of more and higher resolution textures enabled
via full utilization of AGP textures. Higher resolution frames at real-
time update rates and improved anti-aliasing will also enhance the
realism of the user experience. More polygons and pixels are
important, but to truly do "real 3D", you have to do much more. That's
why one of the things we're trying to evangelize is to not simply
consider performance (fill rate, polygons), but look at a combination
of performance and image quality.

 

[fellix]

Aside from the revision mess, AGP was also problematic for most of its lifetime due to many half-baked implementations -- cheap motherboards, chipset and driver bugs. In many cases, most of the AGP's advanced features had to be turned off or dialed down to provide stable operation either manually or through the default device driver settings, and simply act as a high-speed conventional PCI bus for the graphics cards. Gladly, the on-board video memory grew fast enough in all market segments to offset any need for "remote" texturing. It was a good riddance when the much better PCI-E came around and put things in order.

 

[swaaye]

Aside from the revision mess, AGP was also problematic for most of its lifetime due to many half-baked implementations -- cheap motherboards, chipset and driver bugs. In many cases, most of the AGP's advanced features had to be turned off or dialed down to provide stable operation either manually or through the default device driver settings, and simply act as a high-speed conventional PCI bus for the graphics cards. Gladly, the on-board video memory grew fast enough in all market segments to offset any need for "remote" texturing. It was a good riddance when the much better PCI-E came around and put things in order.

Yeah until about 2000 you needed to go with Intel chipsets if you wanted to get AGP that worked.

It's interesting to see how these Real3D people were so excited about AGP texturing. With memory prices dropping regularly you'd think they would've seen where things were going. Maybe this is partly their Intel partnership and some AGP brainwashing from that. Less than a year after i740 launched there were 16MB cards making AGP texturing much less valuable.

 

[Blazkowicz]

You could go for 3dfx if you wanted a lack of AGP. Virge PCI or AGP + Voodoo2 PCI or Voodoo2 SLI, Banshee AGP, Voodoo3 AGP ; all aforementioned AGP cards really are PCI cards in an AGP slot. (even the Vooodoo5 !)

I did that by accident, had a K6/2 fsb 100 rig at some point and I wasn't aware of the AGP problems back then.

Lack of proper AGP got annoying on Voodoo5. 3.3 volt was the bigger limitation, need to hunt for specific motherboards with a universal AGP slot. Then in texture heavy games (such as UT2003 or even 2004!) running out of memory got you very slow texture swapping going on at 1 or 2 fps whereas a real AGP card and a 4x at that could possibly do two-digit framerate.
But it's a bit amazing it did not crash and did 1 fps instead!

UT2K3 was playable for real, UT2K4 not so much and totally unusable with any "terrain" map.

 

[Megadrive1988]

.

I stumbled upon a giant Real3D interview from the ancient times. I figured Megadrive1988 and others would probably enjoy it. A lot of interesting stuff. Sega, i740, competitors and more.

I remember that interview quite well.

 

[DeanoC]

Starfighter PCI had an AGP emulator on board, as the i740 could only 'talk' AGP.

But they didn't change the PCI ID, so the AGP and PCI driver would both detect the card but were completely incompatible.

In Win 95 this had the affect that if you swapped cards (like us devs might do for testing...) and forgot to forcible remove the old cards driver before, windows would die in a horrible (sometimes reinstall only fixable) way on boot.

I loved working with the Starfighter PCI!