Had Sega built Saturn based around the Genesis, Sega CD and 32X

[gamegenie]

SEGA would still be rocking it big today, and they would have crushed SONY back in 1995.

I was convinced of this as kid, and I still know it to be true today.

instead
SEGA's bad decisions drove me away from them and into the arms of SONY.

Sega Saturn: 2 x Hitachi SH-2 32-bit RISC (28.6 MHz)

Sega CD: MC68000 @ 12.5 MHz


In hindsight they should have equipped Saturn with a single 33MHz version of the Motorolla 68040, which would have easily made the Saturn BC with Genesis, and Sega CD. Perhaps include a 20MHz version of the CPU in the 32X, or not worked on the 32X at all.

[Cornugon]

Maybe it was Sony (thus indirectly Nintendo) which caused this bad decision in the first place.

But why a the 33MHz version, why not the 50 or 66MHz versions? Ofcourse an 68060 probably would have been too expensive at that time.

[Chilly Willy]

The 68040 was WAAAAAAAAY too expensive to put into a console in 1994/95. The Saturn has an 11.3 MHz 68000 in it already - they could have perhaps gone with a 28 MHz 68020 for the main CPU. The main issue with the Saturn was the complexity - developers either had to ignore half the hardware, or spend twice as long working on a game. That's why the PSX won that generation - it was the simplest hardware of the competing systems.

[Cornugon]

68030 then@28MHz? Or maybe two 68EC020's

[j_factor]

The 68x series at that point was no longer offering a very good cost-performance ratio. An 020 would have been weak (see CD32).

[Chilly Willy]

The 68x series at that point was no longer offering a very good cost-performance ratio. An 020 would have been weak (see CD32).

Which is why consoles had things like SH or MIPS or ARM. I only mentioned it as a possibility if they wanted something to make BC easier. Personally, I think they'd have been better off going with MIPS. They only went with SH because it used slightly less code space, meaning slightly smaller ROM chips or work RAM needed.

[gamegenie]

Guys I don't think it would have cost that much. Look at this link http://groups.google.com/group/misc....6eab60aaefeb0b

The 40MHz 68040 Motorola was $310, the 25Mhz version was $200. Remember the Saturn cost $400 dollars alone. The CPU price would have fit right in.

Saturn would have probably been more powerful than the PlayStation.

[kool kitty89]

The thing is the SH2's still might have been cheaper, being specifically designed for low-cost embedded/multimedia applications, depending on specific negotiations for bulk orders of course. A lower end 32-bit 68k CPU might have been an option, like the 16.7 MHz rated 68EC020 -only other speed offered being 25 MHz iirc (EC020 also uses 24-bit addressing like the 68000, which avoids some compatibility issues with software using the upper address byte for things it wasn't intended for)

I haven't seen a lot of specifics on the SH2 though, but I'd imagine it was considerably larger (in terms of # of transistors) than an 020 (due to the 4 kB cache if nothing else), though being produced on a .8 micron process rather than the EC020's 1 micron process would mean less silicon for the same transistor count. Of course, physical chip size is only one factor in cost.
The SH2 is also far more powerful, so a lot of the software rendering the 32x could do would no longer be possible.

We were recently talking in the Jaguar forum on Atariage about alternate CPUs, and the 3DO got mixed in there as well. The 68020 family tends to be the most discussed as an alternative to the Jag's 68000, and we'd already gone through discussing how the 16.7 MHz 020 would have been the cheapest option (in any case the cache and 32-bit data bus being the most important).
In any case, it was mentioned that the 3DO was hindered by its CPU lacking a cache (with both GPU and CPU fighting over main RAM), so a inquired why the 020 (or one of the ARM chips with cache) wouldn't have been an alternate choice there as well, and it was speculated that even the lower en EC020 would be significantly more expensive than the ARM60 3DO used. (the ARM being ~35,000 transistors and the 020 being ~190,000 transistors -and any of those ARMs with cache being even larger due to the large cache size -a custom one using the ARM6 core with a small cache like the 020 might have been an interesting possibility though)

So this got me thinking, with the ARM60 being such a small chip (~3/5 the size of a 68000, but still probably more expensive due to the 68k's age and high production volume), perhaps it would have been a good choice for a low-cost Sega design as well, and was a fairly well known architecture -not as much as 68k though, but should be significantly more powerful than a 68020. (as long as lack of cache wasn't a big issue -as it was on 3DO and Jaguar)



But on the main topic, I kind of like the idea, except Sega probably shouldn't have gone with another add-on with the 32x at all, but rather designed a new system (avoiding any limitations resulting from genesis interfacing) that was derived from and compatible with Genesis+CD hardware, but enhanced enough to make it reasonably competitive in the 5th gen market.
Perhaps enhance the Genesis VDP, allowing a much larger master palette (I believe it currently supports up to a 16-bit palette) along with more subpalettes and possibly more scroll layers. increase the VRAM. I'm not sure if increasing the size of the subpalettes would have been practical though, if so that would be great, and perhaps some other features like sprite scaling. (though other hardware could handle that in any case)
Modify the Sega CD video ASIC to work with more than 4-bit/16-color blocks (at least 8-bit/256 colors) and have it working in conjunction with another VDP, either a simple "super VDP" (bitmap display controller) like tomaitheus has suggested, or maybe something a bit more complex with additional hardware features like texture mapping, polygon rendering, and gouraud shading. (though perhaps those could be added features to the ASIC, whcih does have some rudimentary polygon drawing support already -or a kind of approximation)
And the Sega CD's word RAM apears to already be configured in a similar manner as the 32x framebuffers. (2x 128 kB blocks, 16-bit data bus, similar speed I believe as well -and I think the Genesis also communicates to the 32x via the framebuffer)

Probably keep the Genesis CPU the same (maybe use a faster 68k with an optional higher clock speed possible in "nextgen" mode), and either keep the CD's 12.5 MHz 68k, maybe go for a 16.67 MHz one (Sega CD's 50 MHz master clock divided by 3), or maybe a 68EC020, again the 16.67 MHz one being the cheaper option. (though there's the 25 MHz version) Or, keep the 12.5 MHz 68000 for compatibility (and perhaps as a coprocessor) and add another CPU, perhaps an SH2 (which might work out working in cache while the 68k works in main), or the low cost ARM60 I mentioned. (though you probably wouldn't be able to use the 68k much then -no cache and shared bus- maybe just for interrupts) Then expand program RAM to 2 MB (from 512), preferably with a 32-bit bus, but you could make do with 16-bit. (that's all the 32x had, and if it facilitated backwards compatibility in particular it could work)

Then include a DSP coprocessor with similar purpose as the GTE in the PlayStation, 3D calculations, albeit less powerful than the GTE. The SVP chip might be a candidate for this. (and should be able to be clocked at 25 MHz)

Finally, audio: with the Genesis and Sega CD audio hardware, along with the option of redbook CD audio, you might have enough already, perhaps expand the audio RAM for the Sega CD's PCM chip. Maybe also add a pair of DACs kind of like with the 32x, but don't use PWM, software driven like with the 32x (perhaps the DSP could be optionally used for this as well?)


Or ignore all that and go minimalist, take the stock genesis (on a single chip ASIC by 1994 I believe), and the stock Sega CD (maybe soem simple modifications to the ASIC), expand program RAM and add an SH2 or maybe SVP-ish coprocessor along with a simple VDP.


Edit:

The 40MHz 68040 Motorola was $310, the 25Mhz version was $200. Remember the Saturn cost $400 dollars alone. The CPU price would have fit right in.

Saturn would have probably been more powerful than the PlayStation.

GAH! $200 are you insane? Do you think the CPU is the most expensive component and that the Saturn's $400 price tag was selling at production cost? No it was for profit (initially), while Sony was more aggressive and took a big hit on hardware by dumping the price (they could afford to eat the cost -being a big company and planning on profiting through software/royalties) Even if they sold it at cost it would be a lot more expensive by the time it got to retail. (retailers and distributors have to make money too, plus there's the general costs involved in transportation)

That's at least 10x more expensive than you want a CPU to cost! (probably RAM and/or disc drive are the most expensive) preferably you'd get a good deal on a bulk order of the selected CPU, or buy a license and produce it even cheaper than that.

[ThugsRook]

ive pretty much come to the conclusion that if the Saturn supported SMD SCD & 32X carts/cds (even if by adapter) it would have been a winner.

most ppl had nice MegaDrive collections by the time Saturn came out... better hardware would have been nice (Neptune) but not many were interested in an all new platform.

[kool kitty89]

ive pretty much come to the conclusion that if the Saturn supported SMD SCD & 32X carts/cds (even if by adapter) it would have been a winner.

most ppl had nice MegaDrive collections by the time Saturn came out... better hardware would have been nice (Neptune) but not many were interested in an all new platform.

The EU/UK market does seem to hold on to old hardware longer (the 8-bit computers being a big example), so compatibility would probably have been particularly important there, less so in the US, but still important given the popularity of the Genesis. Europe seems like the biggest screw up given that was Sega's most established console market, the US had it's own problems as well (it really does appear they were overspending under Kalinske), but SoJ certainly didn't go about addressing those issues and maintaining their stronger markets. (and the Saturn obviously wasn't designed with those markets in mind in particular, granted neither was the Sega CD, but that was in development before the Genesis was big in the US, so a bit more understandable)

The key would have been doing that and keeping low cost, so probably not the Saturn we know. MD compatibility would obviously be most important, they might have gotten away with ditching Sega CD -but building on that hardware could have been a viable option as well -but more limited. (and should have forgone 32x entirely, just released the new system earlier instead, and more affordable, like below $300 at launch in mid '94)

It would have been simpler had there been no Sega CD to worry about, then again maybe even better if the Sega CD had been more capable in the first place. (separate VDP being the key factor, and probably not a significant cost altering one) That along with releasing an all-in one, integrated, affordable unit could have lasted a lot longer than the SCD as it was, and perhaps foregone the Saturn all together, and either ridden things out until the Dreamcast (maybe releasing it a couple months sooner, and focusing more on refinement and any possible flaws -ie piracy protection)
Or they could have released a new system, less powerful than the DC, but ahead of the mainstream 5th gen consoles around 1996/97, maybe based around an SH3 for CPU, with similar emphasis on simplicity and efficiency as the Dreamcast.

Hell, maybe not even released the CD system as an add-on, but work on it a bit longer and released it (MD compatible) ~1993 (maybe late '92 in japan and later in the west, ~1 year later than the actual MCD) as a standalone system.


And I just realized that link to the 68040 prices is a joke. http://groups.google.com/group/misc....6eab60aaefeb0b

[gamegenie]

And I just realized that link to the 68040 prices is a joke. http://groups.google.com/group/misc....6eab60aaefeb0b

I just realized that too. Even the poster back then in 1994 believes the CPU is over priced junk

[Chilly Willy]

Guys I don't think it would have cost that much. Look at this link http://groups.google.com/group/misc....6eab60aaefeb0b

The 40MHz 68040 Motorola was $310, the 25Mhz version was $200. Remember the Saturn cost $400 dollars alone. The CPU price would have fit right in.

Saturn would have probably been more powerful than the PlayStation.

Compare those prices to the SH-2, which was < $20 (in lots). Every dollar more (than twice that for two CPUs) would have been another dollar added to the price of the Saturn with an 040 instead.

You're too used to general computing processor prices. Embedded processors like the SH line and MIPS are/were dirt cheap.

[gamegenie]

If only we could remake the Saturn today under the 68040 spec, call it the Anniversary Saturn", and boast how it has more features than the Saturn from 1995, it can play Saturn, Sega CD, 32X, and Genesis games. Get SEGA to grant license of it. Price it at $199, that thing will sell like hot cakes.

[Da_Shocker]

What is everybody's fascination with backwards compatibility? Nintendo saw no need in putting it in any of their cart systems. Sega dabbled with it with Powerbase Converter but it wasn't a success at all. It didn't become a success until Sony put it in with the PS2. But it was a hit and miss with the PS3 and with the 360.

[Puffy2k316]

It's pretty convenient but I could live without it. Having the whole ps 1,2, and 3 libraries in one system under your tv is mighty nice.