Panzer Dragoon Saga

[Anarchist420]

Is it true that it could not have looked as good while running at the same frame rate on the PlayStation?

A lot of PS1 ports to the Saturn looked awful, but I can see how PDS could not have been done as well on the PS1.

That said, the 3DO, PS1, and the Saturn each had their own strengths and weaknesses, but all in all I think the Saturn was the best hardware of the 3 (best 2D quality, good enough 3D quality for its time, and reliable as well as the fact that most of its games loaded faster than their PS1 counterparts). The 3DO's biggest strength was its alpha blending, although the transparency in Astal looked pretty damn nice. The 3DO's biggest weaknesses were the 9 pin male controller port, the fact that the FZ1 models had to use a fan, the small cache buffer for the CDROM drive, and the upscaling it did. It would've been nice if used triangle based setup engine, but that's a minor problem for its time.

 

[Sonikku]

The reason a lot of 3D PS1 games looked bad on the Saturn was because developers were just aghast at the "horrible idea" of packing in multi-cores into a gaming machine at the time and didn't take advantage of them. Many developers didn't even USE some of the Saturns cores because they just either couldn't figure it out or didn't want to spend the time to do a game right before shipping to retail. I doubt Panzer Dragoon Saga would have worked on the PS1 given how deeply PDS drank of the Saturns hardware configuration.

Also, I just read your little WW2 sig, and I think it's bologna. He does not so much as acknowledge the slaughter of the Jews that was happening in Germany at the time when explaining how we'd all be better off never going to war. It leaves his essay wide open to counter argument.

 

[Ricochet]

Of all the games that I'd hope would see a current Gen update, PDS is top on my list. Everyone talks about FFVII but really I thought PDS was the better game.

 

[Anarchist420]

The reason a lot of 3D PS1 games looked bad on the Saturn was because developers were just aghast at the "horrible idea" of packing in multi-cores into a gaming machine at the time and didn't take advantage of them. Many developers didn't even USE some of the Saturns cores because they just either couldn't figure it out or didn't want to spend the time to do a game right before shipping to retail. I doubt Panzer Dragoon Saga would have worked on the PS1 given how deeply PDS drank of the Saturns hardware configuration.

Agreed. I had always thought that myself, because there aren't any PS1 games that look better IMO.

 

[nismotigerwvu]

Also don't forget the fundamental difference in the approach to 3d these two systems made when comparing the two. The Saturn used quads for thier primitives while the ps1 (and effectively the rest of the industry) used triangles. Porting engines between the two was far from trivial and almost every asset would need to be redone (assuming you wanted to actually optimize).

 

[crisium]

If you have an import cart, try Bulk Slash. Probably the best use of 3D on the system, and a hella fun game.

The Saturn's best 3D technology games, imo (no particular order):

Virtua Fighter II
Nights (amazing use of colors, decent texture detail, low draw distance is its only problem)
Bulk Slash
Panzer Dragoon Saga
Radiant Silvergun (2D and 3D technology hybrid)
Shining Force III (mainly its impressive how they used loading technology to make battles instant)

Close runners up:
Many other Sega Arcade ports
Panzer Dragoon 1 and 2
Burning Rangers (potential to be the best looking, but the graphics glitch, fade in and out - don't know how to describe it, Nights engine so it uses great colors, but the game seems rushed)

The Saturn has a total of 5 processors iirc, so that made it hard to program for. 2D games occasionally suffered too from poor programming, but overall 2D was this systems prime.

 

[Sonikku]

Too bad for Sega this was the generation when everyone was coming off from the 2D sprites of previous generations and going ZOMG OMG OMG 3DEE!!!1111

For the first few years Sony wouldn't even LET you make a game for their system unless it was 3D. A generation or two later, gamers start realizing 2D games can still be classy. But for Sega, having a console that did 2D best was hardly an asset for the casual rand and file gamer that make or break a consoles success.

 

[Anarchist420]

The Saturn used quads for thier primitives while the ps1 (and effectively the rest of the industry) used triangles.

I thought the 3DO used quads. Was I mistaken?

low draw distance is its only problem)

Most of it I remember had short draw distance, but I seem to recall the last level (Twin Seeds) having rather large draw distance. The last boss was pretty cool. I could get an "A" on every course using the 3D Analog Controller... however, using the D-Pad makes the game hard as hell.

 

[nismotigerwvu]

You're right, after I looked it up it appears the 3DO used quads as well. For some reason I always thought it used triangles. Kudos

 

[Exophase]

The 3DO's biggest weaknesses were the 9 pin male controller port, the fact that the FZ1 models had to use a fan, the small cache buffer for the CDROM drive, and the upscaling it did. It would've been nice if used triangle based setup engine, but that's a minor problem for its time.

You don't think the $699 launch price tag was its biggest weakness..?

The whole quad vs triangle thing is really interesting because the rendering approach was really fundamentally different. Saturn and 3DO were "forward renderers" as opposed to PS1 (and pretty much everything else) which could be called backwards renderers. What this means is Saturn's VDP1 walked over a sprite one pixel at a time and transformed its coordinates to something on the screen, while PS1's GPU walked over the triangle's area on the screen and transformed each screen coordinate back to a texture coordinate.

Triangles work okay with the reverse approach because the interpolation is derived with nice constant linear gradient values that can be inverted and still be nice constant linear gradient values. But with quads, the "correct" approach - bilinear interpolation - has a really ugly inversion that's very computationally expensive to get right. Even the forward method requires more work than it does for triangles, though.

So it lets you do quads, which is the upside, and really it makes sense because the hardware design started with 2D sprites. Quads can be more efficient for some geometry, and they also have the advantage of not looking quite as distorted w/o perspective correction. But there are a couple really big downsides to how Saturn did it. The first is that you have to fill gaps between pixels on the screen, and prevent them from drawing multiple times. This is really hard to get right, and Saturn didn't actually get it totally right, which is why the blending primitives give you artifacts. The other problem is that since Saturn transforms rectangular sprites, you can't arbitrarily clip them against the screen. In fact sprites are multiples of 8 pixels in width so you get even less freedom to resize them. I don't remember what kind of guard band the Saturn framebuffer has but I don't think it's that big so some quads are bound to give you clipping problems under some camera angles.

I kind of wonder if Saturn's DSP being much weaker for 3D transforms than PS1's GTE was ever a problem either. The MAC throughput was a lot lower and it didn't have hardware divisions so you had to synthesize them with Newton Raphson iterations or something.. but maybe the rasterizer couldn't handle enough quads to make a difference.

IMO both systems had really different strong points.. I don't remember Panzer Dragoon Saga that well but it's not hard to imagine that if you design a game around Saturn's strengths you won't be able to match it graphically on PS1. But the same thing is probably true with PS1. Really hard to call a winner.

 

[Anarchist420]

You don't think the $699 launch price tag was its biggest weakness..?

Well kind of, but 3DO charged lower licensing fees (3 dollars only in fact, if what I read was true)... however, I don't recall the games being really inexpensive when they had just come out, even if they cheaper than the average Super NES game.

I always thought it was amazing how Sega could be the first of a generation and the hardware would still have strengths over the competition in all of those 3 generations. The Dreamcast's graphics co processor was actually a lot better than the Gamecube's GPU in my opinion even if it was an order of maginitude slower. The Flipper's fixed function T&L engine and the embedded 1TSRAM were nothing to write home about in my opinion... I know Nintendo wanted a small form factor and low TDP, but when looking at the games of the generation, I think that they made far too many bad choices. To Nintendo's credit, Metroid Prime ran at a great overall frame rate, but artifacts and aliasing with the digital video output in progressive scan mode are as bad as those DC games taking advantage of and running on the VGA box. I think MP would've actually looked better on the DC's graphics processor, but it probably would've run too slowly to be playable.

 

[Exophase]

Well kind of, but 3DO charged lower licensing fees (3 dollars only in fact, if what I read was true)... however, I don't recall the games being really inexpensive when they had just come out, even if they cheaper than the average Super NES game.

Lower licensing fees than who.. the companies making cartridges? It's hard to separate the cost of licensing from the cost of media when the console maker has to make your games for you. So I don't know if they were trying to make more money from console sales than game sales, but it goes without saying that they struck a bad balance and the launch price doomed the console. And they took way too long to lower it. Justifying it by calling it more than a gaming machine made things even worse.. what did they think people were using these for, CD audio and VCDs?

Had 3DO cost < $400 from the start it might have done a lot better. But I don't how well their design license model could have supported it. Would have thought someone with deeper pockets would have tried entering with a cheaper one earlier.

I always thought it was amazing how Sega could be the first of a generation and the hardware would still have strengths over the competition in all of those 3 generations. The Dreamcast's graphics co processor was actually a lot better than the Gamecube's GPU in my opinion even if it was an order of maginitude slower. The Flipper's fixed function T&L engine and the embedded 1TSRAM were nothing to write home about in my opinion... I know Nintendo wanted a small form factor and low TDP, but when looking at the games of the generation, I think that they made far too many bad choices. To Nintendo's credit, Metroid Prime ran at a great overall frame rate, but artifacts and aliasing with the digital video output in progressive scan mode are as bad as those DC games taking advantage of and running on the VGA box. I think MP would've actually looked better on the DC's graphics processor, but it probably would've run too slowly to be playable.

Enh.. if anything Sega showed over and over again that the underdog doesn't gain traction by releasing first in a new generation... The sudden push to release Saturn much earlier was an especially bad idea. Dreamcast had some smart hardware, very much to IMG's credit.. but I don't think Gamecube had serious problems and was also a pretty good design.. IMO a lot fewer than N64 anyway..

 

[Anarchist420]

Enh.. if anything Sega showed over and over again that the underdog doesn't gain traction by releasing first in a new generation... The sudden push to release Saturn much earlier was an especially bad idea. Dreamcast had some smart hardware, very much to IMG's credit.. but I don't think Gamecube had serious problems and was also a pretty good design.. IMO a lot fewer than N64 anyway..

The gamecube delivered a lot of performance per watt and form factor/size, but the internal rendering precision was only done at RGBA6. The Dreamcast used a slightly lower precision back buffer than the GC, but it rendered internally at RGBA8. The Gamecube also was limited to a 24 bit fixed point Z-buffer (unless the CPU's FPU were to be used) while the dreamcast did 32 bit floating point depth precision with infinite clip planes, which may not have been perfect, but it was more versatile overall compared to what the GC could do in that regard.

The Dreamcast also relied on its CPU for its T&L. While that was way slower than the GC's HW non-programmable T&L, at least it could be done with more precision and more programmability (although I may be mistaken). Finally, the GC had awful texture compression.

Due to those issues, Metroid Prime could've actually looked better on the Dreamcast than how it was actually done on the Gamecube, but it would've probably only ran 1/4 the frame rate, which would've been unplayable. It's a shame that the Dreamcast didn't do rotated grid sample pattern for its AA, that it only used a 16 bit back buffer, and that it's filtering/mipmapping wasn't that great (and was costly to even use due to low texel fill rate), but the Gamecube actually wasn't any better in those regards either. The Dreamcast had much better, although not perfect, textures.

Lower licensing fees than who.. the companies making cartridges?

I had thought Nintendo always had disproportionately higher licensing fees than Sega in the Genesis/Super NES era and they were both carts. I had also thought that 3rd party Nintendo 64 games were more expensive than PS1 games not just because the former used ROM chips, but also because Nintendo wanted to make more profit off of the ROM chips than Sony did off of the discs.

Was I mistaken? Just wondering

 

[Exophase]

The gamecube delivered a lot of performance per watt and form factor/size, but the internal rendering precision was only done at RGBA6. The Dreamcast used a slightly lower precision back buffer than the GC, but it rendered internally at RGBA8. The Gamecube also was limited to a 24 bit fixed point Z-buffer (unless the CPU's FPU were to be used) while the dreamcast did 32 bit floating point depth precision with infinite clip planes, which may not have been perfect, but it was more versatile overall compared to what the GC could do in that regard.

Do you have a reference that shows a 6-bit precisions limitation anywhere in the Flipper's pixel pipeline? I can't find anything that'd suggest this but I can't find thorough documentation either.. however I can at least confirm that it supports 24-bit framebuffer output and the TEV stages are 8+bits precision. I wouldn't call 24-bit Z depth much of a limitation for this level of geometry but of course you have to be careful with it.

The Dreamcast also relied on its CPU for its T&L. While that was way slower than the GC's HW non-programmable T&L, at least it could be done with more precision and more programmability (although I may be mistaken). Finally, the GC had awful texture compression.

This doesn't seem like a fair comparison; if you're going to talk about what DC could do with software T&L you should look at what GC could do with purely software T&L and not compare it to its hardware T&L capabilities. I don't know the exact floating point throughput of Gekko but it seems like it can at least do 2x32-bit FMADD per cycle, which would be half the throughput of Dreamcast's SH4 but over twice the clock frequency. So I'm sure it can transform more in software. Gekko's engineers talked about sustaining 2 FP2 SIMD instructions per cycle but I don't know what the exactly capabilities would have been.

And is S3TC really awful compared to Dreamcast's VQ compression?

Due to those issues, Metroid Prime could've actually looked better on the Dreamcast than how it was actually done on the Gamecube, but it would've probably only ran 1/4 the frame rate, which would've been unplayable. It's a shame that the Dreamcast didn't do rotated grid sample pattern for its AA, that it only used a 16 bit back buffer, and that it's filtering/mipmapping wasn't that great (and was costly to even use due to low texel fill rate), but the Gamecube actually wasn't any better in those regards either. The Dreamcast had much better, although not perfect, textures.

The problem with this kind of comparison is you can't just scale the polygon indefinitely on Dreamcast. The more polygons in the scene you have the more memory you need for binning them into tiles. Once you run out of memory you have to split it up into multiple passes and you start losing a lot of the advantages the GPU has, plus adding a ton of additional performance overhead.

Plus you're not really looking at the whole picture here, Dreamcast had a very simple pixel pipeline compared to GC's much more robust 16-stage TEV combiners.. most per-pixel effects you'd see there simply couldn't be done on Dreamcast without again resorting to multiple passes and outside of the precision impact this brings the performance hit is of course dire.. At this point you'd may as well be talking about what you could do with software rendering..

I had thought Nintendo always had disproportionately higher licensing fees than Sega in the Genesis/Super NES era and they were both carts. I had also thought that 3rd party Nintendo 64 games were more expensive than PS1 games not just because the former used ROM chips, but also because Nintendo wanted to make more profit off of the ROM chips than Sony did off of the discs.

Was I mistaken? Just wondering

No idea, but I don't remember SNES games costing more than Genesis games on average. All I really remember was games like Phantasy Star 4 (24mbit cart) cost $100 and comparably sized SNES games (say, FF6) didn't cost that much.. but of course Sega wouldn't be subject to licensing fees on games they themselves were publishing so who knows what the reason here was.

No idea for N64 vs PS1 either, I have to wonder if this is even information anyone without a lot of inside knowledge at Sony and Nintendo could ascertain? All I really know is that third party N64 games weren't more expensive than first party ones, so no one was trying to recoup licensing costs.

 

[Anarchist420]

And is S3TC really awful compared to Dreamcast's VQ compression?

Both consoles had pretty terrible textures.

This doesn't seem like a fair comparison; if you're going to talk about what DC could do with software T&L you should look at what GC could do with purely software T&L and not compare it to its hardware T&L capabilities.

Good point.

Do you have a reference that shows a 6-bit precisions limitation anywhere in the Flipper's pixel pipeline?

Wikipedia once said it did unless alpha isn't used.

 

[magomago]

I played Grandia on both the PSX and the Saturn simultaneously (I had a japanese version and needed english assistance haha), and the PSX veresion looked nicer to me; there was a definite color shift though, and I don't know which version rendered the original vision more nicely. I could actually SEE through the water.

 

[Dari]

The Saturn was built from the ground up to be a 2D system and that is where it excelled. Its 3D was inferior to the PSX.

 

[crisium]

Actually that isn't quite true. The accepted history among Sega scholars (I'm being half funny and half serious with that label) is that the Saturn was built originally to be a system with 3D. However, they were probably more focused on very margianlly surpassing the 1993 consoles: 3D0 and Jaguar. When they saw the specs for the PSX, they knew the Saturn was no match, panicked, and hastily added more chips (which is why it ended up with 5). This of course makes it hard to program for.

It's quite true the 3D on the PSX was still better in practice though. Still, when you get all those chips running to their full potential the Saturn could have been something amazing for 1994 technology. Perhaps stronger than the PSX, judge for yourself:

http://www.youtube.com/watch?v=Gl1Y53W89bw&feature=player_embedded

If only something with those good of graphics actually came out.