SEGA - 3DFX circa 1997

[Megadrive1988]

very interesting thread http://tinyurl.com/jy4h

The 3DFX-SEGA dispute over Sega dropping 3DFX in favor of PowerVR.


The 6th reply to the original post is what caught my eye

BinaryCompatible wrote:

"When Sega and 3Dfx signed that deal, Sega had at least two other
chipsets under development, and that there was no guarentee that
Sega would choose 3Dfx's chipset(In fact, the other two, PowerVR
Highlander and Real3D, were more favored). It was simply a contest of
who would provide Sega with most bang for its buck, and Highlander
won that contest, fair and square. I doubt Sega would not have
included an exit clause in the contracts it signed with all three
parties. At least I don't see Lockheed threatening to sue Sega, even
though their chipset was the fastest(and most expensive) of three."




another insightful response:


"A little known fact is that Sega of America had specified not just a 3DFX
but a Motorola 603e (both American designed chips) for their preliminary
spec of the next-generation console, versus Sega Japan's spec of
Japanese-only hardware. (Note that more powerful variations of the new
design would also be used in next-gen Sega Arcade hardware.)

Sega Japan, which controls all of Sega, immediately rejected the Motorola
CPU in favor for an unavailable, untested, and unproven Hitachi processor
design. Sega's primary reason for rejecting the much more powerful and
well-proven Motorola PPC chip was primarlily political; they wanted a full
license to be made in Japanese fab under Sega's control, something I doubt Motorola would grant. So both the Sega America (specifying 3DFX for graphics processor) and the Sega Japan designs (specifying a yet unchosen Japanese chip for a graphics processor) were in competition for the next console design, and Sega America was forced to change their specification for this unproven and as-yet-unavailable new Hitachi processor as well.

Note that there are other reasons they'd want to go with a proprietary
processor; they didn't want to lose their tight control over developer
releases either; if the 603e and 3DFX combo was used, then it would be a
fairly simple matter to emulate the nextgen console on a PPC equipped with a 3DFX Voodoo card... it would also open the door for developers to easily port the next-gen Sega console games to a powermac equipped with a 3DFX card; and if there's one thing Japanese corporations will not tolerate, it's an open specification and the intense competition that comes with it. They *want* the system to stay proprietary and closed; that's how they do business as a rule (name any Japanese computer system that follows an open spec; even their laptops use proprietary RAM modules, CDROM modules, docking stations....). If someone wanted to clone the console, you can bet they'd have a very difficult time getting the processor chips from Hitachi.

So it comes as no surprise whatsoever that Sega Japan would overrule Sega America's design and go with a Japanese graphics processor as well. Since the only Japanese processor that currently approachs the 3DFX's power was the NEC chip (which of course is a major player in corporate Japan), it seemed *invitable* that they would go with this. Indeed I had been quite surprised when I first heard that they were considering the 3DFX at all."


It'll be interesting to see if anything similar happens between MS, ATI and Nvidia, with XBox2

 

[MfA]

603e more powerfull than the SH4? Only in America

It is all fine and well accusing them of nationalism, but with those kind of reality distorting glasses on it becomes rather humourous.

 

[archie4oz]

Well in scalar maths yes since they were available at higher clocks... But that's about it (at the cost of larger memory footprint as well).

 

[Fox5]

Eh, I thought the powervr chipset was more powerful than what 3dfx was offering at the time. At least whenver these auguments came out in the past, people always said 3dfx was offering voodoo 2 rather than a dual chip solution or a more heavily featured(but not faster) voodoo 3, which would be dubious anyway in outpowering the powervr chip.(I believe an overdraw of 3 would easily put it past a voodoo 3, something which most later games could achieve)

 

[Megadrive1988]

PowerVR2 (PMX, ARC, CLX) was only ment to compete with a single 3Dfx Voodoo2.

Not only did it out do Voodoo2 SLI (two Voodoo2's) it outdid Voodoo3 in many ways to. incredibly impressive. hats off to Videologic.


I still would have rather seen a Lockheed Martin Real3D chip in DC.
say a hypothetical R3D-500 ^__^


but for cost/performance, nothing could beat PowerVR back then.

 

[Fox5]

If the powervr chip was only meant to compete with a single voodoo 2, how come I remember early pr about about how it could get incredible graphics with an overdraw of 10.(that was probably at least after the dc had released though)
It'd be interesting if even its engineers had no idea what it could do.

Anyhow, didn't the power vr(in its most extreme cases) slightly outperform the real3d chip?

 

[Megadrive1988]

Depends on which Real3D chip/chipset we are talking about. the only ones that I know of are the Real3D-100, (lowend/midrange) the Real3D/Pro-1000 (two were used in Sega Model 3) and the single chip i740 Auburn used by Intel and in Real3D Starfighter cards (which sucked)

The Real3D chip(s) developed for a Sega console, assuming there really was, i have no idea how that compared to the PowerVR chip in Dreamcast because it was kept totally out of the publics view and no specs are available on the internet (web or usenet) and believe me, I have searched.

If anyone can provide any links, specs or info of any kind, about the chip(s) or chipset(s) that Lockheed Martin developed for either a Saturn upgrade or a Sega console, I would be really grateful to see anything

see this:

 

[zurich]

One would think that Sega of America was already well versed in Japanese politics, no? If so, then why did they even bother?

 

[Fox5]

Maybe they hoped for some dual system release like the 32x and the saturn? Why did sega of america even have the power to design systems, I don't think nintendo of america can do that....it's it because sega was once sort of an american company?

 

[Megadrive1988]

I look at it like this. Sega of Japan has two console prototypes built to see which one would be the best to replace the aging Saturn. this would make the winner better than if only one was built. I see it much like how the USAF had two Advanced Tactical Fighter prototypes built (YF22 & YF23) to compete against each other. one of them would win and go on to replace the aging F-15 Eagle.



the funny thing is, Lockheed Martin was involved in both! the F-22 for the USAF and the Sega console, in design proposals and holding talks with Sega, at the very least, if not actually building a chip for them, even though it was ultimately rejected.

 

[Fafalada]

Sega's primary reason for rejecting the much more powerful and
well-proven Motorola PPC chip was primarlily political;

That is akin to saying that Sony should have used the well-proven and much more powerfull Celeron chips instead of EE (the whole chip, not just r59k).

Well in scalar maths yes since they were available at higher clocks... But that's about it (at the cost of larger memory footprint as well).

Last I checked 603e's went up to 200mhz and the improved 603ev stopped at 240mhz, which isn't exactly a groundbreaking difference in clock speed.

 

[archie4oz]

Last I checked 603e's went up to 200mhz and the improved 603ev stopped at 240mhz, which isn't exactly a groundbreaking difference in clock speed.

Well they got up to 300MHz for a bare 603e. You can find faster cores in (upto 400MHz) in some of Moto's 82xx controllers... Of course the 603 was dated at THAT time, and I imagine if they would have continued, 750's would have been in order... Of course once you start talking about the FPU it's a whole different matter...

That and PPC has generally benefitted from something you always bitch about in RISC processors... Better CACHE hiearchy... Of course SH again has the better instruction footprint...

 

[Sonic]

No, any such Lockheed Martin chip was scrapped long before the 3dfx and PowerVR deals came to fruition. SEGA would have liked something better than the Pro 1000 to be put in the Dreamcast. But since the manufacturing costs would be so exorbitantly expensive they said bye bye to Lockheed a long time ago.

Some of the people don't have their facts straight. A PowerPC isn't exactly what SoA was looking for in their project. And did anyone really want a Voodoo Banshee as the graphics processor for the Dreamcast?

Good thing it was left to the hands of SoJ. Still, the company as it stands now is far better off than it was years ago.

If Kalinske never left, the company would still reign in the hardware market today.

 

[Fafalada]

Well they got up to 300MHz for a bare 603e. You can find faster cores in (upto 400MHz) in some of Moto's 82xx controllers... Of course the 603 was dated at THAT time

Ok so I don't have the benefit of PPC knowledge that of a mac user 8) But are you sure they got to 300+ back in 97 already? That'd suggest they scaled better then good ole P5 MMX
Anyway, speaking of P5, the benchmarks I remember put 603 performance very close to par with them (at least FPU), and with SH4 being rated as 800mhz P2 you'd need what, one 1ghz 603e to compte?

Or yeah, they could have modded it and bolt an extra coprocessor on but that'd make it an unproven and more expensive design...

That and PPC has generally benefitted from something you always bitch about in RISC processors... Better CACHE hiearchy... Of course SH again has the better instruction footprint...

That I can't argue with, I think they already had out of order execution too right? But yeah, what about SH4 huge register stack and instruction set !!

 

[Megadrive1988]

"No, any such Lockheed Martin chip was scrapped long before the 3dfx and PowerVR deals came to fruition. SEGA would have liked something better than the Pro 1000 to be put in the Dreamcast. But since the manufacturing costs would be so exorbitantly expensive they said bye bye to Lockheed a long time ago."

That would have probably been around 1995-1996 that Sega said goodbye to Lockheed Martin then, because the 3Dfx and PowerVR deals came into being around 1996-1997. Although I still think it was LM's stubborn nature to not compete with the consumer 3D chip makers. Aside from the half-hearted i740, LM decided not to enter the low cost chip industry, even though they had the capability to produce a "R3D-500" or some such chip, on .25 micron, for PCs and consoles, that could have blown away the Pro-1000 like PowerVR2 did in terms of cost/performance.


"Some of the people don't have their facts straight. A PowerPC isn't exactly what SoA was looking for in their project. And did anyone really want a Voodoo Banshee as the graphics processor for the Dreamcast?"

I don't know what SoA was looking for (in BlackBelt) but I thought they were using a PPC 603e and then maybe changed over to the Hitachi SH-4. not sure about that, though. I've also read SoA used a MIPs CPU in BB with the 3Dfx chip. that would have made sense, since MIPs was often used in conjuction with arcade games that used 3Dfx for graphics (i.e. SF Rush)

And, no way in hell would I have wanted a Voodoo Banshee or Banshee2 as the graphics processor for the Dreamcast. thankfully 3Dfx's inferior chips never got inside a released Sega console. Although I wanted LMC R3D more than anything else, PowerVR was an infinitely better choice than 3Dfx.

"Good thing it was left to the hands of SoJ. Still, the company as it stands now is far better off than it was years ago."

Amen to SoJ having the final say in what went into its last(?) console ever.


"If Kalinske never left, the company would still reign in the hardware market today."

yeah <sigh!> too bad Kalinske was no longer there. I miss him so badly!

 

[Lazy8s]

I believe the proposed 3Dfx design was custom-made for SEGA's console, somewhere between Voodoo 2 and 3 and definitely more powerful than the War Gods arcade hardware.

Lockheed's Real3D solution was taken out of the running when SEGA focused on a lower US$199 equivalent target price point. The chipset was supposedly unparalleled in delivering IQ properties like truly high-quality and useable anti-alaising and texture filtering.

For SEGA's needs, the NEC/VideoLogic PowerVR choice was the winner for both performance and political (control, pricing advantages on other supplied parts) reasons. The CLX was literally years ahead of its time (ultimately leading to some of its potential going unexplored)... I run through some Sonic Adventure 2 levels on occasion in VGA, and it's still the most graphically impressive platformer out there.

With regard to Hitachi, the iterations of their SuperH line which were contracted for inclusion in consoles (SH-2 for Saturn and SH-4 for Dreamcast) have always turned out stellar. Hitachi really went all-out creating those powerful and efficient CPUs, which seem to represent the strongest generational leaps within that family line, and delivering in good time for SEGA's launch schedule. It's really no wonder that Dreamcast's core can be modified and made into such a good mobile solution considering how smart the designs from both VideoLogic and Hitachi were.

 

[Vince]

I believe the proposed 3Dfx design was custom-made for SEGA's console, somewhere between Voodoo 2 and 3 and definitely more powerful than the War Gods arcade hardware

If I was another 2% sure, I would say this is Blatantly false. From what I remember, the project for Sega of America was spearheaded by Sellers and it's work was ultimatly manifested in the Voodoo2. Which should never have existed, but that's a story for another time.

 

[darkblu]

I believe the proposed 3Dfx design was custom-made for SEGA's console, somewhere between Voodoo 2 and 3 and definitely more powerful than the War Gods arcade hardware.

the custom-mage design 3dfx offered to sega* later materialized as the pc/mac consumer's voodoo2 product sans the original cream on top (read: SGRAM and TV out signaling), in which form it had a considerable success. which comes to show how the console and pc videocards markets differ.

* effectively an sst1 on steroids.

 

[Vince]

* effectively an sst1 on steroids.

Heh, everything that company actually brought to market was a "SST1 core on steroids."

 

[darkblu]

Heh, everything that company actually brought to market was a "SST1 core on steroids."

yes. still, we don't know whether those were the original intentions, or 3dfx plainly stumbled at delivering new designs (the mythical rampage); come to think of it, a single design reigning the 3d cards consumer market for, erm, 4 years could be quite a cleverly devised business scheme.