Should the SEGA DC and Saturn have launched with these alternative designs?

[swaaye]

How much do you want your Sega console to cost? 3DO M2 in 1996!? I saw ERP say in another thread years ago that in ~1997 or so he couldn't figure out how M2 could ever be priced competitively. And then it became obsolete anyway. The M2 games that are out there aren't much better visually than N64 IMO. Quite a few have been showing up on Youtube.

And a Voodoo1 card alone cost ~$300 in 1996. I've read that $300 happened because they lucked out when the DRAM market did its periodic implosion in '96. 4MB used to be the realm of those $600 2D cards. In '96 Voodoo1 hardware would have been an awfully expensive portion of a console.

I've been thinking that they should have pushed Saturn for a few more years and then made something a bit better than Dreamcast and competed with Xbox/PS2/Cube. But I don't know if the market can support 4 players so maybe it was a lost cause for Sega regardless of what they did.

 

[Nesh]

The M2 games that are out there aren't much better visually than N64 IMO. Quite

?? We ve got M2 games available??
I am interested to see

I've been thinking that they should have pushed Saturn for a few more years and then made something a bit better than Dreamcast and competed with Xbox/PS2/Cube. But I don't know if the market can support 4 players so maybe it was a lost cause for Sega regardless of what they did.

I think it would have been a bad idea to prolong Saturn's life considering how much money it bled. They had to replace it with something healthier as soon as possible to start getting money back or they would go bankrupt.

I think it would have benefited them more if they decided o go multiplatform from there on. I doubt they had much choices given their position. It was either DC in 1998 or multiplatform

 

[Squeak]

Or they could have slapped the Saturn's CD drive on the N64. Increase the size of the texture cache (they'd already enhanced the design once for SoA) and add the Saturns sound processor too and you'd be laughing.

You'd be laughing when you saw the price.
The N64 should have launched with the 64DD as standard storage device. 64Mb was more than enough for games of the time and the loading time was allegedly very fast.
It would have made memory cards look ridiculous.
It would have made games much cheaper. Cheaper still if they had employed a download feature where you could get games either from a modem connection or vending machines. All the you'd need would be a blank disc. No more worries about over or understocking.

The texture "cache" shouldn't have been bigger, it should have been a real cache and not just a tiny piece of software managed scratchpad mem.

And while you're at it, split the memory in VRAM and normal RAM.

 

[swaaye]

?? We ve got M2 games available??
I am interested to see

The unreleased IMSA Racing was apparently almost complete. One of the original developers released it to the web recently.

And there are the arcade games that ran on M2

two more
http://www.youtube.com/watch?v=XaCDPTF7nTM
http://www.youtube.com/watch?v=syq1zYffO0k

 

[Crazyace]

A Saturn related question:

Does anyone know about the Saturn SCU's DSP, and what kind of 3D work it could assist with? There are some documents here, but I can't find any details (perhaps I don't know what I'm looking for):

http://koti.kapsi.fi/~antime/sega/docs.html

System 16 says the following about the ST-V: "SCU DSP : fixed point maths coprocessor, up to 4 parallel instructions". Does this mean you could do matrix multiplication for 3D transforms?

It was a DSP - so it could load 2 32 bit values , perform an integer 32x32 -> 48 bit MAC , and store the result in a single cycle. ( using a small internal memory )
No division though - although you could lookup a table in external memory, and it was clocked at ~14MHz

 

[Megadrive1988]

?? We ve got M2 games available??
I am interested to see

Already posted but yeah there are several unreleased M2 console games which you can see on Youtube, as well as at least 5 (perhaps 6 or 7) released Konami arcade games using M2 technology.


M2 arcade games:

*Battle Tryst
*Evil Night / Hell Night
*Heat Of Eleven '98
*Polystars
*Total Vice

there may be others released or unreleased.

 

[Megadrive1988]

How much do you want your Sega console to cost? 3DO M2 in 1996!? I saw ERP say in another thread years ago that in ~1997 or so he couldn't figure out how M2 could ever be priced competitively. And then it became obsolete anyway. The M2 games that are out there aren't much better visually than N64 IMO. Quite a few have been showing up on Youtube.

M2 hardware was superior to N64 in every way except in AA. Some developers said M2 was 2-3 times the polygon performance of N64, and slightly better than the best PC 3D accelerators at the time including Rendition Vérité 1000, 3DFX Voodoo Graphics and PowerVR PCX1/2.

Just because the M2 games seen don't look far better than N64 doesn't mean the M2 hardware wasn't capable of more, I firmly believe it was. The only hardware superior to M2 at the time was non-consumer arcade boards such as the Sega/Lockheed Model 3.

I've been thinking that they should have pushed Saturn for a few more years and then made something a bit better than Dreamcast and competed with Xbox/PS2/Cube. But I don't know if the market can support 4 players so maybe it was a lost cause for Sega regardless of what they did.

I partly disagree. I think Sega should've trashed Saturn's 1994 design and gone with either 3DO M2, or a Lockheed-designed console in 1996. Then allow THAT platform to succeed beyond 1998/1999 and release a more powerful console in 2000 or 2001 to compete with PS2 and GameCube (and perhaps Xbox if MS still went out on their own).

 

[swaaye]

M2 hardware was superior to N64 in every way except in AA. Some developers said M2 was 2-3 times the polygon performance of N64, and slightly better than the best PC 3D accelerators at the time including Rendition Vérité 1000, 3DFX Voodoo Graphics and PowerVR PCX1/2.

Just because the M2 games seen don't look far better than N64 doesn't mean the M2 hardware wasn't capable of more, I firmly believe it was. The only hardware superior to M2 at the time was non-consumer arcade boards such as the Sega/Lockheed Model 3.

Voodoo Graphics is in another class compared to Verite and PowerVR PCX2. Voodoo1 has several times the pixel fillrate of Verite V1000. Rendition couldn't match Voodoo1 until their V2200 chip and then of course they got squashed by Voodoo2. Voodoo1 had competition but not from a performance or quality perspective.

The M2 games, the end results of people who worked with the hardware and had to actually produce a real game, look to me like higher-framerate higher-resolution N64-level graphics. It doesn't really look like Dreamcast level stuff. I'm sure if Matsushita had had the guts to market another game console and go up against Sony that having some skilled developers work on it would've brought out some interesting stuff. But then again we have the "new millennium console situation" with PS2/Cube/Xbox thoroughly outclassing it.

BTW, why are we discussing M2 again? As an alternative to Dreamcast? That seems....suicidal.

There isn't much of any info on the 3D hardware in M2. It wasn't designed by any of the major 3D companies so who knows how broken or malformed it was. That's the real question. You'd need to chat with someone who worked with it. Spec sheets don't say anything about reality. The 3DO M1/"Opera" hardware wasn't particularly astonishing from a realtime 3D perspective so I'm skeptical about that little team.

 

[Megadrive1988]

I was talking about M2 as an alternative to Saturn, *not* Dreamcast. No, of course M2 did not even remotely approach Dreamcast-like performance, as M2 was 1995/1996 hardware whereas Dreamcast was 1997 hardware released in 1998. The M2 was closer to N64 than it was Dreamcast but, M2 was better than N64 in almost all areas, and significantly better in polygons & textures. M2 was at least as good as 3DFX Voodoo1 and PowerVR PCX1/2 if not slightly better, and miles ahead of Saturn & PS1.

The BDA (Bulldog ASIC), the graphics chip, in M2 was designed at 3DO by ex-SGI engineers as well as the guys that designed the Atari Lynx. It was pretty well documented and understood by developers. The twin PowerPC 602 CPUs offered alot of general purpose and floating point performance. One PowerPC 602 was dedicated to be the geometry engine for the BDA / graphics chip which also housed the sound hardware amoung other things.

The original 3DO M1/Opera hardware was never designed to be a 3D polygon machine like PS1 was. It was closer to Amiga hardware in design. Developers struggled to get decent 3D performance out of the machine which was outclassed by even the Saturn. The M1 was designed in the very early 1990s

 

[TapamN]

One alternate design for the Dreamcast that I think might have been reasonable to do (compared to doubling the clock rate) would be to have a unified memory architecture. Switch from having a 32 bit wide 200 MHz CPU bus and a 64 bit wide 100 MHz video bus, which the real hardware has, to a single 64 bit wide 200 MHz combined. (Sound could still keep its own bus.)

There's be a lot more flexibility to how resources are allocated. Very high polygon count games would have more RAM available for vertex data. Less intensive games would generally have more RAM for textures, from what would be otherwise be slack main RAM. The web browser would benefit from the extra RAM available to the CPU and some simplifications on how it uses the PVR to draw the screen.

A lot of the PVR's data formats seem limited to using 16 MB of memory (e.g. the texture location field is 21 bits long, shifted left 3 bits, giving 16 MB range; many vertex/binning structures support a maximum of 16 MB), so some of the PVR as it is would likely be limited to the lowest 16 MB, but texture addressing would just need an extra left shift to reach all RAM.

Bus contention would cause it to have lower maximum raw performance. I have no idea how much of a loss it would cause, but this setup would probably be OK after adding some better caches. The extra flexibility would be nice.

 

[function]

It was a DSP - so it could load 2 32 bit values , perform an integer 32x32 -> 48 bit MAC , and store the result in a single cycle. ( using a small internal memory )
No division though - although you could lookup a table in external memory, and it was clocked at ~14MHz

Thanks. Back in late 1996 I remember reading about a "previously undocumented DSP" that was supposed to allow people to squeeze even more out of the Saturn. I'm assuming it would be this.

How would you go about using this for vertex calculations? Would a maxtrix multiply of a 4D vector take 16 cycles, or would looking up values in a sin/cos/etc table cause some kind of stall?

It sounds like the two SH2s made it kind of redundant. Strange that it was clocked at half the speed of all the other graphics chips and the two SH2s. Also kind of funny that Sega had the graphics hardware strung out across three different chips with four separate memory pools.

 

[function]

I partly disagree. I think Sega should've trashed Saturn's 1994 design and gone with either 3DO M2, or a Lockheed-designed console in 1996. Then allow THAT platform to succeed beyond 1998/1999 and release a more powerful console in 2000 or 2001 to compete with PS2 and GameCube (and perhaps Xbox if MS still went out on their own).

If the M2 wasn't economical to build in 1997 then that just about rules that out. Given Lockheed Martin's delays with the Model 3 board they possibly couldn't have delivered a console in 1996 either, and it might also have been too expensive, although that would still have been a better option than the Saturn.

The Saturn was bleeding Sega dry, I don't think they could have suffered another few years of Saturn losses and then found half a billion dollars for a new super high end console. Going up against the PSOne with a console with six memory pools (excluding the CDROM drive) and six processors would have gone even more badly than it initially did!

Sega's best option was the one Sega of America was investigating - the N64 hardware. They could have launched in late 1995 (when 350 nm was available from NEC) with a 2 or 4 speed CD Rom drive and had a system that was reasonably economical to make (Nintendo priced) and competitive performance wise.

Any option involving multiple graphics chips and many pools of memory would probably not have been a good one.

 

[swaaye]

N64 hardware with a CDROM would definitely be something that I would like to see the alternative timeline on. Or perhaps with the 64DD thing from the start.

Although I really was not a fan of 2X CDROMs in consoles. N64 has just about zero loading time and that's certainly better than having FMV and CD audio IMO. But having some more storage for improved art assets would be a plus. N64 games were being squished into the size of early '90s PC games.

 

[NeoTechni]

Since it's fantasy, DC should have been a Naomi2 board.

N64 hardware with a CDROM would definitely be something that I would like to see the alternative timeline on.

Such a thing exists (kinda) Some company made a CDROM drive that plugs into the bottom, and it plays ripped N64 games off it. IIRC it came in 2 models, one with 32 MB of RAM, the other with 64 MB (since so few games were that big)

 

[I.S.T.]

Since it's fantasy, DC should have been a Naomi2 board.



Such a thing exists (kinda) Some company made a CDROM drive that plugs into the bottom, and it plays ripped N64 games off it. IIRC it came in 2 models, one with 32 MB of RAM, the other with 64 MB (since so few games were that big)

Load times must be a bitch on that.

Edit: Unless, of course, they load the game into the RAM. If they do that, load times might be faster.

 

[Megadrive1988]

In reality the N64 hardware was not overlly impressive enough for SEGA, that's why they turned to other companies tech like Lockheed Martin Real3D, 3DO M2 and PowerVR Series 1, for either an upgrade to Saturn or a replacement, or both.

 

[SG79]

What was the nature of that proposed Saturn upgrade anyway? RAM surely, but anything else?

 

[function]

In reality the N64 hardware was not overlly impressive enough for SEGA, that's why they turned to other companies tech like Lockheed Martin Real3D, 3DO M2 and PowerVR Series 1, for either an upgrade to Saturn or a replacement, or both.

According to Tom Kalinske (head of SoA at the time) the N64 was rejected for being too difficult to manufacture and SoJ went ahead with the Saturn instead. Whether it really was too difficult to manufacture isn't something I have any way of knowing, and maybe it was, but there's always been speculation that SoJ were very keen to work with Hitachi, and the excessive use of SH processors in the Saturn doesn't go against this.

Whatever, by the time late 1995 rolled around this wouldn't have been the case - in fact it was cheap enough that Nintendo went with it. Saturn was failing and losing money while the N64 giggled it's way into shops and made Nintendo money.

The Megadrive could have carried Sega until 1996, so something from Lockheed Martin might have been a possibility.

What was the nature of that proposed Saturn upgrade anyway? RAM surely, but anything else?

3D accelerators basically. The two that I remember being rumoured were one from Lockheed Martin (so probably Real3D based) and a PowerVR one. The PowerVR rumour is probably based on early DC talk, and the Lockheed Martin one either didn't get very far or was never actually intended to be an upgrade.

Here's a forum thread that I Googled across with some interesting stuff in it:
http://www.assemblergames.com/forums/showthread.php?t=9375

IMO The best "32-bit" system would have come from aborting the Saturn and getting a Real3D based console instead. The best outcome for Sega would probably have been aborting the Saturn and going with N64 hardware in 1995 and replacing it with a PowerVR based system in 1999 or 2000.

The worst outcome was actually launching the Saturn. Especially after having launched the 32X.

[Edit]A thread about the Real3D/100: http://hardforum.com/showthread.php?t=1431521

Three chips just for the Real3D/100. Still, no worse than the Saturn and at least it wouldn't need 2 CPUs. Looks like 3 memory pools for the graphics subsystem too ... so better than the Saturn then. And unlike the Saturn it could probably actually do really good 3D. Miles ahead of the N64. [/Edit]

 

[Megadrive1988]

According to Tom Kalinske (head of SoA at the time) the N64 was rejected for being too difficult to manufacture and SoJ went ahead with the Saturn instead. Whether it really was too difficult to manufacture isn't something I have any way of knowing, and maybe it was, but there's always been speculation that SoJ were very keen to work with Hitachi, and the excessive use of SH processors in the Saturn doesn't go against this.

Whatever, by the time late 1995 rolled around this wouldn't have been the case - in fact it was cheap enough that Nintendo went with it. Saturn was failing and losing money while the N64 giggled it's way into shops and made Nintendo money.

The Megadrive could have carried Sega until 1996, so something from Lockheed Martin might have been a possibility.



3D accelerators basically. The two that I remember being rumoured were one from Lockheed Martin (so probably Real3D based) and a PowerVR one. The PowerVR rumour is probably based on early DC talk, and the Lockheed Martin one either didn't get very far or was never actually intended to be an upgrade.

Here's a forum thread that I Googled across with some interesting stuff in it:
http://www.assemblergames.com/forums/showthread.php?t=9375

IMO The best "32-bit" system would have come from aborting the Saturn and getting a Real3D based console instead. The best outcome for Sega would probably have been aborting the Saturn and going with N64 hardware in 1995 and replacing it with a PowerVR based system in 1999 or 2000.

The worst outcome was actually launching the Saturn. Especially after having launched the 32X.

[Edit]A thread about the Real3D/100: http://hardforum.com/showthread.php?t=1431521

Three chips just for the Real3D/100. Still, no worse than the Saturn and at least it wouldn't need 2 CPUs. Looks like 3 memory pools for the graphics subsystem too ... so better than the Saturn then. And unlike the Saturn it could probably actually do really good 3D. Miles ahead of the N64. [/Edit]

I basicly agree with almost everything you've said. The Saturn upgrade would've been a 3D accelerator (PowerVR Series1, Locheed Martin Real3D/100 or the far worse i740 or maybe even 3Dfx. The other idea as to replace Saturn altogether. Either not have Saturn come out, or quickly replace it. The upgrade probably wouldn't have worked anyway.

Here is the November 1995 Next Generation article on Saturn 2, using a Lockheed Martin Real3D/100 graphics engine/image generator and possibly a PowerPC CPU manufactured by Hitachi. Next Generation notes that this Saturn 2 could've been a new standalone console, but possibly instead it would be an upgrade for Saturn, saving cost by using the Saturn as an I/O device, CD-ROM drive and power supply.

http://i.imgur.com/sjqprU5.jpg
http://i.imgur.com/10BH7EI.jpg
http://i.imgur.com/tsEgLM1.jpg


Still we all know add-ons/upgrades rarely succeed but it was no different than what 3DO was doing by making the M2 an upgrade for the original M1/Opera hardware, as well as a standalone console.

 

[function]

Here is the November 1995 Next Generation article on Saturn 2, using a Lockheed Martin Real3D/100 graphics engine/image generator and possibly a PowerPC CPU manufactured by Hitachi. Next Generation notes that this Saturn 2 could've been a new standalone console, but possibly instead it would be an upgrade for Saturn, saving cost by using the Saturn as an I/O device, CD-ROM drive and power supply.

http://img80.imageshack.us/i/saturn2lmc1crop983x9806js.jpg/
http://img122.imageshack.us/i/saturn2lmc2crop1033x13522rk.jpg/
http://img122.imageshack.us/i/saturn2lmc3crop1063x14140sq.jpg/

Still we all know add-ons/upgrades rarely succeed but it was no different than what 3DO was doing by making the M2 an upgrade for the original M1/Opera hardware, as well as a standalone console.

It's an interesting article, and as usual I end up thinking "poor Yu Suzuki".

I think Sega learned their lesson with add ons and using old systems as IO for new ones when the 32X launched, and when they thought about how much Neptune would cost to make. I would guess that by late 1995 the idea of Saturn 2 being an add on was really just being touted outside of Sega - Saturn 2 would have had a very, very limited userbase if it was limited to a subset of Saturn 1 owners, and integrating the Saturn with a Real3D system would have created a monster with about a million chips! Filling up 8 - 12MB or so of memory from a 2X CD drive would not be much fun for the user either. The top guys in Sega's AM departments never even heard of a Saturn ad on.

I found this link while Googling around - you probably know all this already, but I found there was a couple of interesting nuggets in there:

http://www.scribd.com/doc/246881/r3d-prod

The geometry processor requires one update cycle, the graphics and other processors require another, and the frame buffer delay is an additional field time

Texture maps can be any rectangular size, 2n x 2m , where m and n
are less than or equal to eleven.

2k textures? In 1996?