The LAST GEN console capabilities thread.

[DeadlyNinja]

This is a thread for the PS2, GC and Xbox. I want to know what they can REALLY do. By that, I don't mean pointless benchmark numbers like 77 million polygons. I can find those numbers all over the internet. What I want to discuss is things like what exactly is the real polygon performance of the PS2/Xbox. I want to learn what kind of effects did the 3 systems push that totally surprised people. Thing like can the PS2 do multilayer texturing. These questions are just examples of what I (and perhaps others here as well) want to know. I want people to talk about games to give examples of blah blah blah being used of this and that.

I want a big thread that gathers all the information in one place. I searched for some of answers, but I found very little (at least very little of what I had in mind). Hopefully, this will become a useful thread.

I guess I'll start off with a few questions...

1. Can the PS2 do bump mapping? I can't think of a game that uses it, but then again, I've never played every single PS2 game known to man.

2. What's the deal with anti-aliasing on all 3 systems? I remember in an EGM article talking about the PS2's hardware problems, and one of them was anti-aliasing. It says the hardware's not capable of it, but some programmers were able to get it to work. The first Tekken 3 screens (Japanese) showed off the lack of AA, but eventually, all PS2 screens seem to have AA. The same is true with GC and Xbox. Was AA something that really required a lot of work to get done on them? I don't even remember aliasing being a big issue with any of the 3 systems, but maybe I wasn't paying attention.

3. Those example questions I asked in the beginning could also use some answers.

 

[Npl]

1. Can the PS2 do bump mapping? I can't think of a game that uses it, but then again, I've never played every single PS2 game known to man.

The PS2 cant even do multitexturing natively (Bumpmapping requires 2 textures). But alot of the effects can be done by using multiple passes and Im sure bump-mapping can be done this way too.

2. What's the deal with anti-aliasing on all 3 systems?

PS2 - edge AA support
Xbox - multisampling
GC - ?

All can do supersampling

3. Those example questions I asked in the beginning could also use some answers.

Well, those are the raw performance numbers and Im sure you can dig up the rendering features aswell. This is the only way to somewhat quantify them, but if you want subjective comparisons, just look at the games.

 

[Mobius1aic]

To answer the bumpmapping question about the PS2, yes it could be done via multiple passes. The only game I know of to do it was Hitman: Blood Money. Normal mapping could even be done as well which was proven by the game Matrix: Path of Neo. Personally I've never played Path of Neo so I wouldn't know how well done the normal mapping is, but I've played the demo to Hitman and I have to say I'm impressed that it was implemented, but it's not very heir apparent unless you look closely. There are other games on the PS2 which would make better use of such techniques and perhaps then would the damn system would've seen more games using techniques like such that had been known to be doable on the PS2 since at least 2002.

The method of doing these "shaders" involves doing the calculations via a vector unit on the Emotion Engine.

 

[archie4oz]

This is a thread for the PS2, GC and Xbox. I want to know what they can REALLY do. By that, I don't mean pointless benchmark numbers like 77 million polygons. I can find those numbers all over the internet. What I want to discuss is things like what exactly is the real polygon performance of the PS2/Xbox.

OK, now that's a problematic question for starters because you haven't defined what "real polygon performance" is... As for your pointless benchmark numbers, those aren't pointless benchmarks, those are simply specs that the hardware is capable of and they're far from pointless (how else are you going to gauge the performance of your code against the hardware's physical capabilities?).

 

[Heinrich4]

A ps2 game called Primal maybe have "some level" of bump mapping (pre baked maybe) textures.

 

[DeadlyNinja]

OK, now that's a problematic question for starters because you haven't defined what "real polygon performance" is... As for your pointless benchmark numbers, those aren't pointless benchmarks, those are simply specs that the hardware is capable of and they're far from pointless (how else are you going to gauge the performance of your code against the hardware's physical capabilities?).

Well I simply mean what game on the PS2 pump the most amount of polygons. I think that's a fair enough question. I mean, when they say 77 million, they're not really running a game or anything. Those numbers all point to it being able to pump more polygons than the GC, but we know that real game performance isn't like that at all.

 

[Shifty Geezer]

What your asking for cannot be provided as a meaningful set of metrics or feature lists. The only comparative data is peak capabilities. What developers actually get from the machines can only be seen, and so measured and compared, in the actual games that were released.

Does PS2 support normal mapping? Look through PS2's library and see if there are any normal-mapped games. Does it support AA? Again, see what games are out there with antialiasing. Everything else is a 'best case' feature set but which doesn't tell you how that relates to the final game output. eg. PS2's huge triangle draw that you say doesn't look any better than GC - the PS2 works by drawing the same objects lots of times, each time with a different effect. Thus where GC or XB might render a 5,000 poly model in one pass with 2 colour, 1 ambient light, and 1 specular map, PS2 would render the same 5,000 poly model four times in four passes to do the same, effectively rendering 20,000 polygons.

Thus the far higher polygon rate is true and valid for PS2, but what that equates to on screen is totally different. You cannot understand a console from paper specs alone. You can't appreciate the comparative performance either. You need to refer to real-world use to see how developers flexed the systems to make those paper specs actually do something. eg. On average PS2 games (counting 100 titles measured) render ~ 1,000,000 triangles per second with four textures per pixel, and XB renders (counting 74 titles measured) ~1.5 million with four texture per pixel. Good luck finding that sort of info though!

 

[swaaye]

Yeah it really does come down to just looking at what the best games on the systems managed to pull off. Xbox is the only console from that era to pull off normal mapping, I believe, for example. Cube is close to Xbox overall, despite notably inferior specs (it's extremely efficient). PS2 is decidedly behind those two but some devs did some amazing things even with its relatively simplistic GPU.

It's interesting that Dreamcast is actually part of that generation. At least according to the Wikipedia folk. I'd always kinda thought of Dreamcast as its own little generation.

 

[Shifty Geezer]

You also saw PS2's fillrate and BW shine in particles and screen effects. At the time of its design, the EE probably wasn't a bad choice. PS2 was very flexible. The major losses were IQ, with serious texture shimmer on far too many titles. Though for people coming from PS1 and it's zig-zag textures, this wasn't anything to complain about!

 

[DeadlyNinja]

Yeah it really does come down to just looking at what the best games on the systems managed to pull off.

Yeah, that's why I specifically said to talk about their best games for examples of the very best effects.

Xbox is the only console from that era to pull off normal mapping, I believe, for example.

Oh yeah, what's the deal with the Xbox port of HL2. Didn't the game have most of its bump mapping removed from the Xbox port?

It's interesting that Dreamcast is actually part of that generation.

Oops, I forgot about the DC...

 

[swaaye]

Oh yeah, what's the deal with the Xbox port of HL2. Didn't the game have most of its bump mapping removed from the Xbox port?

Probably to save RAM. Extra textures use more RAM.

FarCry, Doom3, Riddick, and Halo2 are the best examples of why I say Xbox was the most powerful machine of that generation. Each does things that the other consoles never did, I believe. The shadowing and lighting, and normal mapping.

 

[Neb]

What your asking for cannot be provided as a meaningful set of metrics or feature lists. The only comparative data is peak capabilities. What developers actually get from the machines can only be seen, and so measured and compared, in the actual games that were released.

Does PS2 support normal mapping? Look through PS2's library and see if there are any normal-mapped games. Does it support AA? Again, see what games are out there with antialiasing. Everything else is a 'best case' feature set but which doesn't tell you how that relates to the final game output. eg. PS2's huge triangle draw that you say doesn't look any better than GC - the PS2 works by drawing the same objects lots of times, each time with a different effect. Thus where GC or XB might render a 5,000 poly model in one pass with 2 colour, 1 ambient light, and 1 specular map, PS2 would render the same 5,000 poly model four times in four passes to do the same, effectively rendering 20,000 polygons.

A good reason why models and graphics overall was and is lower quality/detailed on the PS2 games compared to xbox games. That the PS2 had to redraw each frame several times to achieve desired result meant also to have to cut down on detail. Also interesting is that many numbers stated for models of last-gen games where including redraw times of the model, 10000 being in real 5000 or less etc (probably now to by some).

Do as much as possible without the need to redraw the frame multiple times, make up with artwork and lighting/colors.

 

[Neb]

Probably to save RAM. Extra textures use more RAM.

FarCry, Doom3, Riddick, and Halo2 are the best examples of why I say Xbox was the most powerful machine of that generation. Each does things that the other consoles never did, I believe. The shadowing and lighting, and normal mapping.

Without doubt and the technical features of games such as PGR2, GR2, Halo2 (mapping and cut-scenes) etc.

 

[Neb]

OK, now that's a problematic question for starters because you haven't defined what "real polygon performance" is... As for your pointless benchmark numbers, those aren't pointless benchmarks, those are simply specs that the hardware is capable of and they're far from pointless (how else are you going to gauge the performance of your code against the hardware's physical capabilities?).

They are still pretty pointless since it only is peak on paper/special case benchmark. Real-world perfomance tells a whole different story and is what really mathers not 'hype material' numbers.

 

[jlippo]

They are still pretty pointless since it only is peak on paper/special case benchmark. Real-world perfomance tells a whole different story and is what really mathers not 'hype material' numbers.

Those specs did tell you quite a lot about the architecture as ps2 did actually meet those numbers.
They didn't tell that games will have 75 or 66 million polygons/second performance in games, they told that they have 'gpu' and cpu which can do certain operations within these limits.

Realworld numbers from games really depend very much on the code and how someone has tacled the problems and not so much on the best case which the hardware can archieve.
If I would code something with Ps2 and my code would be one hundred of the speed of code sample/specs, I would know to blame myself.

 

[_phil_]

That the PS2 had to redraw each frame several times to achieve desired result meant also to have to cut down on detail.

you think we don't draw things many time ,now ? can you tell what needed to be redraw with and without geometry resubmited ?

Also interesting is that many numbers stated for models of last-gen games where including redraw times of the model, 10000 being in real 5000 or less etc (probably now to by some).

I'm impressed at how much you know ,or at least at so much you think you know,
Your natural ability to distort or select reality to fit your pretty obvious preference.

 

[Fafalada]

A good reason why models and graphics overall was and is lower quality/detailed on the PS2 games compared to xbox games. That the PS2 had to redraw each frame several times to achieve desired result meant also to have to cut down on detail.

Multipass is not something that was in any way exclusive to PS2, not to mention the performance hit on multiple layers on some other hardware was often comparable regardless of whether it was single or multi pass.
XBox could draw more is because the GPU subsystem was faster overall, period + it had at least 2x more usable graphics memory.

Also interesting is that many numbers stated for models of last-gen games where including redraw times of the model, 10000 being in real 5000 or less etc (probably now to by some).

Those were exceptions to the rule though. If you look at the polygon counting thread, the PS2 model numbers stated there are real.

 

[Neb]

you think we don't draw things many time ,now ? can you tell what needed to be redraw with and without geometry resubmited ?

I have not said it applies for all games but looking at how games for both platforms the PS2 version morethan often had cut down detail on many fronts. Now on the PS3 if you mean then shure you are, certainly more than on the PS2.

I'm impressed at how much you know ,or at least at so much you think you know,
Your natural ability to distort or select reality to fit your pretty obvious preference.

So you mean that all devs state the poly numbers for characters/scenes without counting redraws, yes?

 

[Neb]

Multipass is not something that was in any way exclusive to PS2, not to mention the performance hit on multiple layers on some other hardware was often comparable regardless of whether it was single or multi pass.
XBox could draw more is because the GPU subsystem was faster overall, period + it had at least 2x more usable graphics memory.

Of course multipass where used on all platforms with varying results.

Those were exceptions to the rule though. If you look at the polygon counting thread, the PS2 model numbers stated there are real.

I should have said some instead of many games obviously!

 

[fearsomepirate]

1. Yes, the PS2 can do bump mapping. The question should be "What's the cost?" Matrix: Path of Neo had normal maps in a few places. Jak 3 had bump mapping somewhere as well.

2. Aliasing on Gamecube and PS2 was typically handled by use of an interlacing filter, which would blur the image slightly compared to true FSAA but reduce the shimmering considerably. A fair number of PS2 games halved the vertical resolution in order to save space and/or fillrate, which is why some games had really awful shimmer. Xbox had hardware FSAA, which was used in quite a few games, such as PGR2 and Forza Motorsport. A few Gamecube and a few PS2 games also used true FSAA as well.

3. PS2's main limitation in visual quality was the size of its eDRAM and lack of native texture compression, not really fill rate or vertex power. IIRC, it didn't have the ability to texture or draw to the screen from main RAM, but I could be wrong about the latter. Because of this, a relatively large number of PS2 games were limited to 8-bit or even 4-bit textures, plus other things associated with drawing the screen (field rendering, halving the vertical resolution, etc). With S3 texture compression, Gamecube could put as many 24-bit textures in the same space that the PS2 would use for just 4-bit textures, and Xbox had a similar advantage.

Unlike both Xbox and PS2, Gamecube was unable to draw in 32-bit RGBA mode, as it could only store 24-bit frame buffers, so one was limited to either 8:8:8 RGB and doing any transparency effects another way (sometimes difficult, but produced similar color quality to both PS2 and Xbox), or using 6:6:6:6 RGBA, which is noticeably degraded over PS2 and Xbox. The Prince of Persia games and Resident Evil 4 are prime examples of the latter.