dreamcast graphics comparison

[chao2]

Could the dreamcast compete with late ps2,xbox,and gamecube graphically? I know the dreamcast had a couple amazing titles like shenmue,shenmue 2, and sonic adventure 2 but I dont think they compare well graphically to games like kingdom hearts,pyschonauts,and ninja gaiden. Did the dreamcast have much left to show graphically ? We probably never know but I wonder if you guys have found or discovered something extraordinary on the dreamcast that could have improved its graphics?

[PH3NOM]

Well, I hate to be a downer, but the DC really is much weaker than the Xbox. The xbox really was a generation ahead, even compared to the PS2. The gap between the PS2 and DC is much smaller, though.

The Xbox features a GPU that handles customizable vertex shaders, offloading transform and lighting to the GPU so the CPU can handle other code.

The DC has to perform all transform and lighting on the CPU, even though the SH4 features some nice vector math functions, that alone can consume the entire SH4 CPU with even some relatively low polygon count scenes, depending on the number of light sources.

As a programmer, I would have to say Shenmue II really must be close to maxing out what the DC can do, as the code I have written comes nowhere close in terms of polygon count.

[chao2]

How small is the gap between the ps2 and gamecube? would the dreamcast be able to receive ports kingdom hearts or sonic heroes without too much being taken away from the game?

[DaMadFiddler]

This is all off the top of my head, so please excuse any technical errors.

Well, that's kind of a moot point, because it's doubtful any of those titles will ever be open-source.

However, one thing to bear in mind is that while the DC has more memory than the PS2, the PS2 can push *significantly* more polygons. So, all other things aside (and that's a LOT of other things to set aside), either of those games would likely have to be completely redesigned for Dreamcast. (Sonic Heroes looked like garbage on the PS2, anyway.)

Another important difference is the base architecture.

The Dreamcast used an SH4 CPU, which is similar in in ability (though not in function) to other RISC processors like the higher-end models of the ARM series. In that regard, it's closer to what you'd find in a smartphone than in a PC. The PS2 used a custom (and slightly complicated) Sony chip. The GameCube used a derivative of the PowerPC G3, and the Xbox used a modified Pentium 3. (In fact, you can swap out the Xbox CPU for an off-the-shelf P3 for an easy upgrade. The Xbox was basically just a locked-down Wintel machine.)

In terms of the GPU, the Dreamcast used a PowerVR chip, which uses some clever tricks to do more with less and was quite good for the time (1998), but is lacking many features that would be considered standard just a few years later. Sony, again, used custom hardware, which was a bit of a mixed bag. It could push a lot more polygons than the Dreamcast's GPU, but it was oddly lacking some hardware features that even the Dreamcast had. The GameCube used an ATI GPU similar to the Radeon 8500 series, and the Xbox used an nVidia chip similar to the GeForce 4. (In PC terms, both would be considered DirectX 8.1-compliant hardware.) The Xbox's was more powerful than the GameCube's, but they had a comparable feature set whereas the PS2 and DC GPUs both lacked several features compared to the others, meaning a number of now-common effects had to be either faked, left out, or handled by the CPU.

Memory was another issue. The Dreamcast had more RAM than the PS2 (RAM was a real bottleneck for the PS2, significantly limiting its performance), but still tiny compared to what the GameCube and Xbox had.

However, what this all neglects to address is the artistic side. Visual design counts for a lot, and the longer developers have to really work their way through the ins and outs of a given system, the more clever little ideas they come up with to improve the look and feel and to try out new techniques. A top-quality developer making a Dreamcast game today would undoubtedly be able to produce a better-looking game than a top-quality developer in the Dreamcast's heyday, simply because the modern developer would have an extra decade-plus of design evolution to draw upon.

That's not the same as being able to port a game that looked good on another system (say, Metroid Prime) to the Dreamcast, but a NEW game being made for the Dreamcast--with the right tools and talent--could still blow away most if not all of what we've seen on the platform. However, in that same hypothetical, a modern developer making a new game for the GameCube or PS2 would also be able to make a better-looking game for either of those platforms, and be able to make that game look better than anything the Dreamcast could handle--because those platforms were more powerful in various ways.

As a side thought, though: the DC died pretty much at the beginning of "modern" gaming. A number of new graphics techniques and gameplay elements were developed over the course of that console generation, many (though not all) of which would have worked for the Dreamcast as well. As a result, we never really saw many of those newer techniques applied to the Dreamcast; most of what we saw was built off of what people knew from the PlayStation/N64 era.

Of course, the Dreamcast's limited processing power and GPU functionality would still have created bottlenecks, but had DC development continued, it probably could have held its own for at least most of that generation, specifically thanks to the PS2. The PS2 was a mixed bag in comparison to the Dreamcast, but was significantly less powerful than the GameCube and Xbox. As a result, it kind of held back game development, as game developers wanted to get their titles on the "popular" console, so they had to design around the less-powerful PS2 and port upward, not taking full advantage of the more powerful systems.

Since the PS2 was around so long, though, and got so much attention, its hardware is pretty well understood and it has probably been utilized more thoroughly than just about any other platform. Evolved design aside, I doubt you could make a PS2 game that looks too much better than some of that platform's later titles. The Wii is basically a spec-boosted GameCube, so you can get a pretty good idea of what it could have done by imagining scaled back versions of Wii titles. But the Dreamcast... well, it's all speculation since all we have to test the theory with is indie developers and KOS (which, great as it is, is not the same as a properly maintained professional dev kit in the hands of a AAA developer), but I guess the best way to picture the "best possible" Dreamcast game would be to imagine one of the late-period PS2 games, with sharper textures but MUCH blockier geometry. (The more limited CPU, however, will also constrict what you can do in terms of physics and AI... so some games just flat-out wouldn't fit the hardware.)

So there's a rambling... thing. More technical folks, feel free to correct anything that's off-base or inaccurate.

[chao2]

so what techniques have you speculated or know that could be used in dreamcast games to improve the graphics ? give me an example from a game ? thank you Fiddler and ph3nom I can't believe I forgot to consider artistic styles in the equation.

[PH3NOM]

You make a lot of good points there.
It really is a shame the DC was cut short, we never will know what could have happened if the DC had received commercial support for as long as the PS2 had.

The only thing I dont agree with is the statement

A top-quality developer making a Dreamcast game today would undoubtedly be able to produce a better-looking game than a top-quality developer in the Dreamcast's heyday, simply because the modern developer would have an extra decade-plus of design evolution to draw upon.

I stand by my original statement, the DC's potential was fully realized with Shenmue II.
Yu Suziki is not just a top-notch developer, but I would argue he is a brilliant game developer.
Not just that, but his team was fully backed by SEGA to ensure all tools were available to leverage what the hardware could offer. Also, the game had begun development on the Sega Saturn, so we know the game had spent plenty of time in development.
Yes, things could be changed graphics wise, but there are always sacrifices to be made. I believe those sacrifices were carefully balanced in Shenmue II, pushing things pretty much as far as they can go in that respect.

Sure, things have changed in game development since the DC's lifetime, but the reality is current development is based around hardware the DC does not have, and that is Vertex Shaders on the GPU.

For example, the lighting in my build of OpenGL very closely resembles that of lighting used in modern vertex shaders. The big difference, is my code is running on the SH4 CPU, instead of a close-to-modern GPU.

Although the pipeline could be optimized to some extent: the lighting algorithm could perform some backface-culling checks to eliminate redundant execution among other things, the numbers should give a baseline. The fact is there are many things that need to be done on the DC's 200 mHz CPU, clipping polygons alone can take a good amount of time, something that is done natively by some modern GPU's.
At any rate, here is a screen of my OpenGL API, using dynamic vertex lighting with 2 diffuse +1 ambient light sources. All of the lighting equations are using the SH4's fast vector math functions, written with inlined assembly.
Real-Time, I was able to achieve 400k verts/sec at 20fps on the DC, see the post here
viewtopic.php?f=29&t=102506


When it comes to Vertex operations, I have made a few benchmarks some time ago. The benchmarks were made without any output to the PVR, so it measures CPU only performing such operations.
I marked the SH4 somewhere around 16 million perspective divided vertex transforms per second. In a 3D space, every vertex on screen must be transformed. Divide that by frames per second, at 60 fps that gives us 266666 vertices per frame.

Lighting a Vertex real-time uses the dot-product operation quite heavily.
I marked the SH4 at

Code: Select all

/* Throughput: Benchmarked at ~28,410,000 dot operations every 1,000 msec */

Sounds like a big number, over 28 million dot-product ops per second. But when you divide that by frames per second, and then subtract time spent transforming the vertex, those numbers become increasingly smaller.

Although, I would really like to hear what Tapamn has to say on this topic...

[RyoDC]

MadFiddler, all the things you said are true, especially about the part with developers and etc, but you forgot one thing: all those factors doesn't make Dreamcast hardware any better. Katana libraries, I think, already (at the state of it's release) had maintained most of the hardware features in DC. Even if they utilized all the powers of dreamcast hardware (applied all optimisations, removed all bottlenecks, max usage of DMA etc.), there's still a top border, which you can't jump above.
Xbox have much more powered hardware (more computations per sec, more features, that are by the hardware, and that means those features are done faster).

[DaMadFiddler]

I think my comment about a modern developer being able to do better was a bit misunderstood. It's not that I was trying to say the Dreamcast would be able to pull off newer effects; it's more a matter of art design having evolved.

For example, we have much better software and hardware at our disposal now, which should result in better-looking prerendered elements (textures, sprites, videos, and other 2D assets).

We've also learned through experience where certain types of effects do and don't work (using yellow and orange polygons for fire, fore example, looks awful).

There are also some newer effects that can be implemented in software, though as Ph3n0m pointed out, the Dreamcast doesn't have a lot of horsepower to spare, so the trade-offs for these may not be practical. Some examples of things that were implemented in a lot of second-wave PS2/GC games include basic puddle dynamics, cloth movment--which is only used in a couple of DC games--more extensive/mature use of particle effects, etc. And some artistic decisions would be made differently today, in terms of the design and animation of models, environment and object design, when to use models vs pseudosprites (where to place those pseudosprites strategically so they don't stand out as much), etc. Even things as small as the fact that Sonic Adventure 2 used Comic Sans as its primary font, or how the life bars are designed in a fighting game, can have an impact on how polished the game can feel.

More importantly, though, game design matured a lot over the last 10 years. 3D games were still pretty young and rough when the Dreamcast got started, and as a result, gameplay design was clunky. Again, you have some pretty severe hardware limits to consider, which will limit your AI, world size, and amount and degree of detail in the things that are going on to comprise the gameplay world... but other things are just results of outdated, clunky design. Decisions made regarding level design, HUD & interface design, how characters control, and even framing elements all seem small but add up to a lot.

That's what I meant by saying a game developed now by a AAA developer would look better. It's not necessarily that the Dreamcast could be pushed to do MORE now (as mentioned, a few things could probably be implemented in software, but they'd come at such a significant trade-off that they'd only be useful in very specific circumstances), it's that game design has had over a decade to mature since the big boys stopped making Dreamcast games.

[chao2]

would have there been better ways to implement new effects for the dreamcast with out sacrificing too much from the game?

[DaMadFiddler]

would have there been better ways to implement new effects for the dreamcast with out sacrificing too much from the game?

No. That's pretty much what PH3N0M was saying. Anything that isn't natively supported on the hardware would have to be done in software, which is VERY expensive in terms of system resources. Remember, the Dreamcast only has a 200MHz RISC processor. Having to load that down with various lighting effects and transforms in ADDITION to running the game logic would be quite taxing.

You might be able to get away with it in games that have very simple engines (puzzle games, space shooters, maybe some fighting games, and possibly some 2D titles), but not in any title that has very much going on logic-wise.

[chao2]

So the dreamcast wouldn't be able to have a game(rpg for example) with 5 million polygons with effects added ? This would be impossible right?

[RyoDC]

So the dreamcast wouldn't be able to have a game(rpg for example) with 5 million polygons with effects added ? This would be impossible right?

Yeah, something like that.

[chao2]

So if the dreamcast survived until 2003 or 2004 we would not be seeing games not that much different from 1999 in terms of graphics. So top notch developers wouldn't be able to push the dreamcast anymore than shenmue 2 did already? wouldn't other top notch developers improve things a little more than shenmue 2? I mean shenmue 2 came out only three years after the dreamcast.

[legit nyck]

I'm a big fan of sega and dreamcast light I have one here at home also have the play2
Unfortunately the truth has to be told the playstation is a lot better than the dreamcast in terms of graphics

Note: please no me ban

[BlueCrab]

So, this doesn't really have anything in particular to do with programming, so off to a more appropriate forum we go...

[dark2]

A fun way to ponder this question is to look at late period (post 2003) dreamcast and naomi games - naomi, because this sega arcade hardware was similar in spec to the dreamcast and various developers continued making games for the system until 2008 or so.

I think a few naomi titles are good to mention:
-Senko no Ronde (released in 2005) (by GRev, makers of Border Down and Under Defeat among other titles)
-Dynamite Deka Asian Dynamite (2006) (by Sega, enough said )
-Illvelo (2008) (by Milestone, makers of Chaos Field, Radirgy and Karous -3 games on the naomi hardware)

*Each of these games were created for the naomi/dreamcast spec rather than being a port from different hardware.

*Each of these games featured 3D graphics and were created by developers who already had significant years of experience developing titles for the naomi/dreamcast hardware.

*Since each was created in the mid-late 2000s, the developers had more modern tool sets to use for graphics creation - the additional knowledge gained from being a developer in the mid 2000s as opposed to the late 90s - and the feeling of competition with modern games of the era on other platforms.

Imo, these are several reasons why these games in particular would be good representations of late period dreamcast games.

So how do they look? That's all subjective... I think they look good and not entirely obsolete if compared to PS2 or GC games of the era (heck, two of these three titles were later released on the Wii and the xbox 360 so the developers and publishers considered them marketable in that later generation). Nevertheless, I would say that any of these titles would have been possible from a graphical standpoint on the PS2, GC, and Xbox hardware, and I would not say these titles put other titles on the other hardware to shame. I would also say these titles do look a little blockier/lower polygon count, and simpler in terms of lighting and particle type effects than one would expect for a straight-up GC or Xbox game.

So I think my overall conclusion is that dreamcast games made by knowledgeable developers with good resources and modern know-how around 2005-2008, as represented by these titles, could have made games that would have been close to the best that the dreamcast hardware could do - and while they would not have looked exceedingly dated compared to competing games from other systems - from a graphical standpoint, these titles would probably appear fairly average looking when compared to straight-up GC and Xbox titles.

Some screenshots of the titles for your reference:
http://gamesdbase.com/Media/SYSTEM/Arca ... _G-Rev.jpg
http://gamesdbase.com/Media/SYSTEM/Sega ... -_Sega.jpg
http://www.hardcoregaming101.net/dynami ... aex-12.png
http://img185.imageshack.us/img185/716/163ae2.jpg

[TheRedFox]

I remember reading a thread somewhere a while back where someone had done polycounts of various dreamcast games using nullDC, and found that, in terms of polycounts Dead or Alive 2 (unsurprising) and Triggerheart Exelica (much more surprising) had the highest. I don't know which games he tested though. I believe Shenmue II was one of them as well as Test Drive LeMans.
Also, apparently the polys aren't clipped (I'm not totally sure what that means) in Triggerheart, which is, I guess less efficient?
The information was strewn about in this thread somewhere, it's been a while since I read it.
Interesting stuff.

Another thing that people like to mention is the tile-based rendering, whereby only the visible polygons are onscreen at any given time, which can conserve vram for more important things (at least if I'm understanding this correctly, I don't have any hands on experience with this stuff, only have read a lot about it). That would lead to a higher degree of detail without actually increasing polycount, a higher "apparent polygon count" one might say.

Of course, as I'm sure any Dreamcaster knows, polygons only go so far, there's a lot more that leads to nice looking graphics, like textures, for example, bright colors etc.

Another thing that people like to mention is the PVR2DC's texture compression support that leads to it being able to keep a lot of texture data in RAM compared to, for example, the PS2. I think others in this thread have touched on that a little bit. I read somewhere that DoA2 on the DC uses mipmapping, as opposed to the PS2 version, which I assume is only possible due to the texture compression (and higher VRAM more generally). I don't have any links to that, though.

I haven't done very extensive comparisons or anything, but it seems like the DC was able to be relatively efficient with its meager resources, to get to the level of the first couple of years, at least, of PS2 games. Also, devs often times needed to make pretty severe redesigns of games, such as, lets say Resident Evil 4, to port them over to PS2 from GC and/or Xbox. If games had been redesigned from the ground up for Dreamcast, like they were for the PS2, to take advantage of the Dreamcast's strengths and efficiencies, I don't see why they couldn't have been ported. But that being said, the games that were ported down to PS2 probably looked a lot worse then their GC and Xbox originals, and the same would probably be true for the Dreamcast.

as far as art design goes, I think the recent Sturmwind is a good example of this. Technically, it's probably not very impressive at all, using sprites made of prerendered 3D models, most likely (apparently there's some polygons in use, but I don't think very many), yet the impressive parrallax, and general art design makes it look WAY smoother than probably any other game that uses that art style, at least that I've ever seen (then again if it is done really well, I wouldn't notice it, anyway). I mean, compare that to like DKC or something (not that that's a bad game, but the graphics haven't aged all that well). I think Duranik might have even used some photographic elements in the backgrounds, which usually looks severely tacky (think of that one Zelda game for the CDi that uses photographic backgrounds).

anyway, I've probably said enough. Does anyone know if any modern graphics chips use anything similar to the PVR's texture compression or tile-based rendering? seems like it would be a good idea.

[chao2]

I just found about this game for atomiswave called Kenju. It was like a dbz/soul calibur with colorful graphics. It was cancelled but here is a clip of the game.

http://www.youtube.com/watch?v=WhG9LCPT7iY

http://www.youtube.com/watch?v=WhG9LCPT7iY