New interview talking of unreleased Sega hardware

http://www.sega-16.com/feature_page.php?id=323&title=Interview:%20Toshiyasu%20Morita

Very interesting. He talks about a lot of things, but included in there are several DC prototypes from SOA.

 

The SH4+3DFX is probably the Dural, right?

 

I'm guessing, yeah. Supposedly it was a PPC & 3Dfx, but it would seem SH4s were the in it from the beginning.

 

x

 

Japanese PowerVR based design was Dural
US 3Dfx based design was Blackbelt.

 

I think so, yeah.


However in spring & summer of 1997, it was understood (incorrectly or not) that Dural was one of the names for the Japanese-designed system that used PowerVR, later known as Katana which was given the consumer name Dreamcast in 1998.

In spring/summer 1997 it was reported online and in magazines that it came down to a choice between American-designed Black Belt with 3Dfx and Japanese-designed Dural with PowerVR.

At the time, the name Dreamcast was totally unknown. Also, the name Katana did not surface until late 1997. So Dural was most often used as the name of the Japanese system that competed with BlackBelt.


I think only later was it found out (again, correctly or incorrectly) that Dural was one of the names for the 3Dfx-based hardware designed in the U.S. which was most usually known as BlackBelt.

Also, it was first reported by Next Generation Online, on March 13th, 1997 that Black Belt was using PowerVR PCX2. Not 3Dfx.







At the time, NGO had not yet discovered that Sega was developing 2 consoles in parallel. It wouldn't be long before they found this out and reported on it, though.

That said, this is confusing for two reasons.

1.) Black Belt was actually the name for the 3Dfx-based system.

2.) The other system that *was* using PowerVR (Katana/Dreamcast) was using PowerVR2. The PowerVR PCX2 is -not- the same as PowerVR2.

Both PCX1 and PCX2 were first-generation PowerVR Series 1 chips.

It would later be reported that Sega would be using PowerVR Highlander, the original codename for the 2nd-generation PowerVR2 family. Highlander was actually the codename for the multi-media (2D, 3D, video) variant of PowerVR2
which later became known as PMX-1, which was what the PowerVR 250 is, used in the Neon250 PC card.

The Black Belt / Dural (assuming Dural was a name for the U.S. system) was based on a 3Dfx chip, some variant of Voodoo2 or Banshee. The CPU was finalized as SH4, however Sega was also considering the PowerPC 603e.

Also, there were some unconfirmed reports of Black Belt using MIPS CPU which was typical in 3Dfx-based arcade games, but AFAIK it was between SH4 and PPC 603e.

PowerPC 603e was considered for the unreleased American 3Dfx-based BlackBelt/Dural and possibly also for the released Katana/Dreamcast.


http://web.archive.org/web/19970605165511/www.next-generation.com/news/050297b.chtml

Here's an article from Next Generation magazine, August 1997. By this time the decision on what console & chipset would be used had already been made ( it happened in July 1997), however since this issue had to have been written in May or June, it wasn't know at press time

 

In April 1997 there was word (unfortunately IMO) that Sega would not be using a Lockheed-Martin Real3D chip. It was assumed that 3Dfx was the frontrunner at that point. Lockheed commented on the Black Belt and them not being the graphics provider

http://web.archive.org/web/19970605161903/www.next-generation.com/news/042997b.chtml

 

Yeah 3Dfx was definitally the graphics provider for Black Belt.

However, the Saturn 2 project seems more connected to what AM2 and the other AM divisions at Sega of Japan wanted: a totally 100% Lockheed Martin designed console. The higher ups at SoJ agreed only to let Lockheed do the graphics subsystem of Saturn 2, however that was the main point anyway. The CPU would've come from Hitachi, probably a PowerPC CPU if not a CPU from the SH family. The important thing was that Lockheed handle the graphics, as reported in this NG article:



The unreleased Lockheed Martin Real3D based Saturn 2 was in development (1994, 1995 upto 1996) well before Black Belt/Dural and Katana/Dreamcast (1996/1997) was into development.

 

Damn, I miss Next Gen magazine :crying:

 

that's a cool thread jim ,nice one! very interesting stuff indeed!

 

Sorry if this has been discussed before, but why was the PowerVR chosen over the Lockheed Martin Real3D? From the games that I've seen, it would seem that the Lockheed Martin Real3D is actually more powerful than the PowerVR. Was it an issue of price or something?

 

Real3D was weaker than the PowerVR chip actually - and by a large margin. Intel used Real3D in their i740 graphics cards and I can confirm that it royally sucked. Google intel 740 and you'll see that it didn't really cut it even back in 1999, which was when I bought my first ever PC with my own cash. The PC cost me 700 GBP and the graphics card was considered uber budget at the time. Voodoo 2 was close to top of the run back then I believe.

My my, how things in the PC world have changed (to the point of ridicule). When I consider what you can buy now for the same money it makes me shake my fist into the air and shout "Why God, why? Look at the monsters you have created!"

 

Pricing. Lockheed Martin's unwillingness to push the true Real3D chips into the consumer market which would include mass producing the chips to bring down the cost. The Lockheed Martin/Intel 'i740' chip used on Intel motherboards and Lockheed Starfighter cards doesn't really count because its such a downgraded accelerator without a geometry processor as used in Real3D/100 and /Pro 1000 which could be considered complete GPUs. It's a real shame Lockheed was unwilling to push Real3D/100 and the other chips they probably had into the consumer/gamer space. Or creat cheaper single-chip versions with smaller manufacturing process. they would've beaten 3Dfx and the others so badly in both polygon performance and rendering quality.

PowerVR was selected because of its low-cost (lower than 3Dfx), Japanese manufacturing partner (NEC), and performance/quality which came the closest to the Real3D-powered MODEL 3, even though Real3D was still better.

I still would've wanted Lockheed Martin Real3D to have been used in 2 generations of Sega console hardware

1.) A Saturn released in 1996 instead of the one released in 1994/1995
or a Saturn upgrade in 1996-1997 of the same generation as N64 and 3DO M2, but better because of Real3D/100. A standalone console would've been the best choice (because a faster bus/architecture and faster CD drive could be employed, whereas an upgrade to Saturn would've caused hardware & userbase problems. Standalone console is what Matsushita was going to do with M2 before it was canceled as a console.

2.) A new next-gen system in 2000-2001 or 2002 instead of Dreamcast using a new custom Real3D/XXXX with at least as much power as the NV2A GPU in Xbox, if not more powerful. Something to ensure superiority over PS2 and Gamecube.

 

Only the i740 was weaker. As I said, it doesn't count. It was a watered down accelerator mainly for Intel's needs, without geometry processing. It came out in late 1997 or early 1998 with Voodoo1+ like performance and slightly better image quality. It wasn't as fast as Voodoo2.

However the older Real3D/100 and /Pro-1000 were absolutely, totally different animals. Real3D/100 (PC, workstations) and Real3D/Pro-1000 (Model 3, workstations) were better overall than PowerVR PCX1, PCX2, PowerVR2DC, PowerVR250. Riva 128, TNT, Voodoo2 and Banshee.

It seems the Pro-1000 used in Model 3 was better than TNT2 and Voodoo3 released in 1999. It was not until the GeForce came out in late 1999 that Model3/Pro-1000 performance was seen on PCs.

Also note that consumer/gamer 3D providers quoted the max peak theoretical polygon figures for their chips and this did not reflect realworld in-game performance. Lockheed Real3D specs were very good, very high for 1996, yet, were not inflated, but what could be achived in the realworld, and, they were better than later-releaed chips from 3Dfx, ATI, Nvidia, etc with higher paper specs. Again this does not include the Real3D i740 which cut corners to get the cheapest possible chip.

 

Real3D/100 or Pro-1000 were way out of reach for a ~$200 games console. The only Real3D based hardware that would be available at a price Sega could afford was the i740. That's why it wasn't even a contender.

 

And yet, the following article (and other print+online articles of the time, in 1995) implied that Real3D/100 would be avaliable in $180 ~ $200 graphics cards:



The R3D/100 was an entire chipset with 3 main chips plus some supporting chips, in 1995. Just think, the PlayStation1 was probably a bunch of chips in 1993. Yet Sony and LSI Logic integrated most things onto a few chips by late 1994. If Lockheed had done the same with R3D/100 by late 1996 or early 1997, putting R3D/100 onto one chip, it could've been feasible for console use. Perhaps even cheaper than the 2-chip Voodoo Graphics. IIRC R3D/100 cards cost $500 to $2000. Most of that was probably memory, upto ~20 MB. I'd imagine an 8 to 16 MB version with a single logic chip and mass production could've massively reduced the cost.

 

Just an opinion... I always have considered that Dural was the name for an early stage of the Katana project. Well, me and many of the media which I've read articles from.

 

Same here. As I've said, in 1997, the Dural was known to be the name for the PowerVR2 based console. later known as Katana and Dreamcast.

Yet it seems that Dural may have been another name for the Black Belt.

I can see this being true. there was alot of confusion going on in 1997. One example of that confusion is Next Generation Online said, in their first report on Black Belt, that it was based on PowerVR PCX2, when actually Black Belt was based on 3Dfx.

 

The 3Dfx hardware probably would've been weaker than the PowerVR, right?

 

yes, the Banshee is lame.