Strengths and weaknesses of GameCube relative to its peers *spawn

So you're saying the snow in this image should have been just flat with no bumps?

18.jpg


Normal maps aren't a substitute for good geometry, but they definitely are necessary to add detail because the amount of geometry you actually need is ridiculous, beyond even the latest GPUs.

Except those aren't normal maps, it's environment mapped bump mapping. RS doesn't sacrifice poly counts for normal maps, this is my whole point. And exactly.

Which absolutely, today normal maps are extremely important, i'm just talking about the Xbox here.

You're complaining about artistic choices and compromises which were part of that generation, and you're also being highly subjective is selecting your performance criteria. For you personally, high object detail trumps all, so you attribute 'the best' console to the one that satisfies your perception. Which is fine as a subjective discussion of 'what console do you prefer?' but no good for a technical discussion of the hardware.

Basically, like so may, you already have a view and the discussion isn't about challenging that but expressing it, to the point technical arguments are brushed aside and you stick with gut feelings. Which you are entitled to do, but not in the B3D Tech fora. Please stick to facts and evidence and, ideally (although not required), please come with an open mind ready to accept one's view may not be accurate and there may be other ways to look at things. I don't think I'm wrong to say many of us were once somewhat blinkered fanboys who have since outgrown such limited thinking thanks to B3D. I found this place many years ago trying to found out how good/bad (okay, how totally Awesome and Superior) PS2 actually was, migrating through various lesser forums until I found this one and got real answers that challenged my preconceptions. Then I came to see all the machines as different engineering solutions and came to understand it's really about the software and the developers, and the whole branding thing is bollocks only worth indulging in as a fun pasttime as long as one can keep it from getting serious.

I never said high object detail trumps all, it's a combination of lighting, texture work (which gamecube also excels at) and effects, and yes object detail. And sure I think it's fair to say preferring detailed modeling over blocky modeling is subjective, I can agree with that. Beauty is in the eye of the beholder ;-) No but seriously, i get what you're saying.

I'm sticking to facts and evidence, i'm not bringing up anything theoretical. And my views have changed, as I noted with burnout. And i'm still here ready to learn something else.
 
Except those aren't normal maps, it's environment mapped bump mapping.
It's detail added in a way not possible using geometry. That's why bump mapping of any from was invented, to add more detail to hardware with limited geometry abilities.

RS doesn't sacrifice poly counts for normal maps, this is my whole point.
RS would have looked better if they added normal mapping.

Which absolutely, today normal maps are extremely important, i'm just talking about the Xbox here.
It was even more important back then because polygons were large and flat. Normal maps could bring detail to characters and environments.

I'm sticking to facts and evidence, i'm not bringing up anything theoretical.
But your request for evidence is for subjective evidence of a game on XB or PS2 that 'looks better' that a game on GC. Computer games are all smoke and mirrors and personal opinion, so 'looks better' is never going to be a decent metric. We never know comparing platform exclusives what the game would look like optimised for another machine, and we never know with cross-platform games whether they are good ports or not, so screenshot comparisons really don't convey much real information. The actual data you've been provided is 1) XB could do more polygons than GC. 2) It could add normal maps to those polygons.
 
It's detail added in a way not possible using geometry. That's why bump mapping of any from was invented, to add more detail to hardware with limited geometry abilities.

*RS would have looked better if they added normal mapping.

*Assuming no other graphical sacrifices were made.

*It was even more important back then because polygons were large and flat. Normal maps could bring detail to characters and environments.

But your request for evidence is for subjective evidence of a game on XB or PS2 that 'looks better' that a game on GC. Computer games are all smoke and mirrors and personal opinion, so 'looks better' is never going to be a decent metric. We never know comparing platform exclusives what the game would look like optimised for another machine, and we never know with cross-platform games whether they are good ports or not, so screenshot comparisons really don't convey much real information.

*Sure, but again not when using them means more sharp, spiky objects.

No, this is where you're mistaking me. I'm not saying those games on GC look best because of their polygon count alone. To try and show you i'm not just trying to validate what I think looks best, there are games on Ps2 which I think look better than most GC and Xbox games. I think Klonoa 2 on Ps2 looks better than a lot of what Nintendo produced in that era. If you want, you can continue to pin huge bias on me while pretending that no one else has them. We never really lose our biases, we just tone it down.

I was talking about polygon counts in these games... because I wanted to compare that metric at the time. As I said, it's the total package that matters. Also, stop with telling me to stop with the screenshots. I used it to show one aspect of the game, just as you used it again after me because you thought I was putting all the importance on geometry.

"The actual data you've been provided is 1) XB could do more polygons than GC."

That is not a fact, there is no evidence. I'm getting theoretical performance numbers and "developer sentiments." Besides, you don't think there is a developer out there that would say the GC could do more polygons than Xbox? And even so, it still wouldn't end the conversation in my favor. This is why we have to look at the actual games and what they achieved.
 
Last edited:
"The actual data you've been provided is 1) XB could do more polygons than GC."

That is not a fact, there is no evidence. I'm getting theoretical performance numbers and "developer sentiments." Besides, you don't think there is a developer out there that would say the GC could do more polygons than Xbox? And even so, it still wouldn't end the conversation in my favor. This is why we have to look at the actual games and what they achieved.
So you want to use the evidence of your eyes looking at screenshots instead of the evidence of developers who created games and reports of numbers of polygons from titles? How do your eyes factor in optimisation, culling, particular engine requirements, etc.to separate what the hardware is capable of versus what the software achieved?
 
So you want to use the evidence of your eyes looking at screenshots instead of the evidence of developers who created games and reports of numbers of polygons from titles? How do your eyes factor in optimisation, culling, particular engine requirements, etc.to separate what the hardware is capable of versus what the software achieved?
Uuuugh, no I don't want "evidence in screenshots", and no a developer claiming something isn't proof, unless it actually exists in a game, real time demo ; anything tangible we can actually confirm. The absence of evidence isn't the evidence of absence, but it isn't evidence.

I don't think anything else is going to come out of this conversation.
 
and no a developer claiming something isn't proof, unless it actually exists in a game, real time demo ; anything tangible we can actually confirm.
ERP released multiple games across multiple systems. That constitutes first hand evidence. If you don't want to take that at face value...*shrug*. Not sure how you're going to learn anything new as your asking to if you won't listen to the people that worked on the machines.
 
ERP released multiple games across multiple systems. That constitutes first hand evidence. If you don't want to take that at face value...*shrug*. Not sure how you're going to learn anything new as your asking to if you won't listen to the people that worked on the machines.
I don't doubt what he wrote in that thread. In fact it makes sense when looking at a few poor gamecube conversions. However it seems to be purely from the perspective of a multiplatform developer, not someone who worked to get the most out of each console. *Please note this isn't a way of me saying "lazy dev"

In all fairness then, I admit Re4 Ps2 as an example wouldn't be getting the most out of the platform, though even if it were a ps2 exclusive i'm sure the overall look would need to be cut down. It would likely have some advantage, perhaps particles.
 
Last edited:
And i'll say one last thing as well, with regards to graphics. Normal maps don't fix the lack of geometric complexity on heads, fingers and limbs. So in Doom 3 (also Chaos theory) and riddick (though riddick himself looks fairly rounded) you end up with much more blocky looking models, though they have extra surface detail. Halo combat evolved looks downright primitive compared to a large number of games on all 3 of these consoles because of its low poly counts and stiff animation. I thought about it, and when I think of the most impressive Xbox games to me, none of them have normal mapping, and they run at high frame rates, or at least stable 30fps. Panzer dragoon orta, Conker, Voodoo vince outrun 2 etc. etc.

HTupolev has already explained some of Halo's background, and it's important to note that Doom 3 was designed to run on devices with much lower T&L ability than the dual vertex shader GF4s, so its poly count would have suffered. GF3's had to be able to run the game with only one vertex shader.

I acknowledge the xbox is capable of this type of lighting and "bump mapping", while the others cannot, it's just that I don't think the console has any business using these techniques. The lighting may be more advanced, but it isn't exactly more pleasing to the eye and it results in significant sacrifices to other parts of the game.

The 'sacrifices' are much lower in Xbox specific games. The outstanding Halo 2 used normal mapping very effectively, and if I recall correctly Panzer Orta does and the earth shatteringly brilliant Rallisport Challenge 2 does to (it uses relatively complex pixel shaders for the time, so it wouldn't surprise me if they did the ice track detail using it).

I need to find my Xbox and play Rallisport Challenge 2 again. I wish I hadn't chucked all my CRTs. :( :(

By "GC didn't need normal mapping" I mean GC didn't need normal maps to compete with Xbox on a graphical level.

Well that's subjective, but my personal thought would be it depends on the game. You don't have to plaster normal maps over everything for them to be a worthwhile addition.

Nothing but more theoretical data and no in game results. If I don't see results, I don't care, which is why I brought up if outsiders looked into this conversation ; they likely wouldn't either.

Well, those hypothetical outsiders probably wouldn't be equipped for this conversation so we shouldn't care what they think.

The results aren't theoretical, they're very real and independently verifiable. They are for a synthetic test (that's what 3DMark is) but you do get to see the results on screen. They're also using DX8 for PC, over an ancient AGP bus by the looks of things. And as ERP stated, you can go faster on Xbox than on PC.

Interestingly - and this is the kind of thing the popular socialites here like to try and spot - you'll see that:
- GF4 Ti 4400 (275 mhz) hits 11.5 mpps w/8 lights to the GF3 Ti 500 (240 mhz) 5.6 mpps - more than 2X faster
- GF4 Ti 4400 (275 mhz) hits 46.5 mpps w/1 lights to the GF3 Ti 500 (240 mhz) 26.1 mpps - rather less than 2X faster

Evidence of the kind of Xbox avoidable bottleneck ERP was talking about, even on a mighty AMD Athlon XP 2000? Just an interesting observation ....

Anyway, here's the thing: triangle setup and T&L wouldn't have been a bottleneck in massively exceeding RS poly counts, and triangle setup and T&L were absolutely fast enough to allow a game to hit a peak of 30 mpps. The hardware can do it. For real.

If you're struggling to transform 15 million pps it's because you're using a lot of lights on a lot of the verts or because you're bottlenecked on the CPU or in the pixel shader or TMU or ROP or something. But that could happen on any platform.

If someone says the Xbox was more capable simply because it could use rendering techniques the GC could not, that's true, but in terms of all around detail and actual results (games) both consoles are pretty similar, though I still say gamecube edges it out (except with games like burnout); it certainly has less bottlenecks and was the more thought out piece of hardware, in fact it's probably the most well rounded console period. Though admittedly that's me looking at 5th gen onward.

Well you're perfectly entitled to think GC games look better. But the Xbox could throw more maths behind every pixel and vertex.

GC is impressive, but you're looking in the wrong place to see what's so impressive about it, IMO.

It's good to know gamecube couldn't do anything I thought it could and insights to how the ps2 renders games, that's all good stuff. But without evidence and results, my outlook's not going to change.

You're entitled to your outlook, but the rest of us prefer things that are a little more concrete.

Emotional attachments are hard to shed. I've been there, every bit as much as you I suspect!

After your response (which I will read and consider) i'd like to talk about the cpus next. I admit the my idea of the gekko being superior was *sorta* based around that Factor 5 clock frequency comment, so it's probably not as good as I thought it was. But I also know it has more registers and cache for starters. Anyways, i'd like to learn anything I can about these cpus.

I don't know a lot, but a good place to start might be to compare some die sizes: https://forum.beyond3d.com/threads/console-die-sizes.53343/

... and then search for some benchmarks for the PowerPC 750CXe (closest none console chip iirc), scaling results for clock frequency. It's a while since I looked at any benchmarks, but what I remember being impressive was its performance in relation to its die size.

The GC and Wii have both been hacked, so you may be able to find some benches direct from those systems. There used to be some, so I expect they're still out in the wild somewhere ...


"The actual data you've been provided is 1) XB could do more polygons than GC."

That is not a fact, there is no evidence. I'm getting theoretical performance numbers and "developer sentiments." Besides, you don't think there is a developer out there that would say the GC could do more polygons than Xbox? And even so, it still wouldn't end the conversation in my favor. This is why we have to look at the actual games and what they achieved.

There's evidence that the hardware could in reality transform and present vastly more geometry to the screen than RS used. There's also proof that the GF4 could "do" more polygons than the ATI Radeon 8500, ATIs flagship. It's not unreasonable to think that newer, significantly larger and much more highly clocked GPUs would have greater capability that older, smaller, and much lower clocked chips.

And there's also a statement from a highly experienced multiplatform dev, speaking in hindsight, after the generation, about the hardware. And while you say this is just a "sentiment", this would have been based on profiling actual, optimised software. It's not a guess, it's the result of real code, that's really been tested, for realz.

And no, I don't think there's a developer who would say GC could push more polygons for real. Not a programmer, who optimised the rendering pipeline anyway ... maybe a PR mouthpiece or something or some gobshite with a blog who knows a guy that was a tester or something spurious like that.

But as always, I invite you to try and find them! Try! Find a developer! Or find something showing GC can hit 50 mpps! Or find some developer docs on the T&L unit!

I've actually spent a while here looking for statements from developers, looking for old hardware benches, looking at contemporary T&L units (nothing matches GF4 even when clocked higher). Meanwhile, you dismiss everything and simply state that your entirely unsupported opinion is a truth that must be disproved through an increasingly narrow and moving set of goal posts.

How about you do some work now?
 
I don't doubt what he wrote in that thread. In fact it makes sense when looking at a few poor gamecube conversions. However it seems to be purely from the perspective of a multiplatform developer, not someone who worked to get the most out of each console. *Please note this isn't a way of me saying "lazy dev"

In all fairness then, I admit Re4 Ps2 as an example wouldn't be getting the most out of the platform, though even if it were a ps2 exclusive i'm sure the overall look would need to be cut down. It would likely have some advantage, perhaps particles.

Thing about citing RE4 is that it was built from the ground up for GameCube and then released on a completely different platform in about 10 months. I wonder how God of War 2 would look on a GameCube if ported under the same conditions.

Unlike the GameCube however, PS2 was a complicated machine that required a lot of work to get good results from. In fact we know now from Corysama in the PS2 thread that late in its lifespan some developers figured out how to get the VU0 to trigger an interrupt on the main CPU core which would have allowed more games to use VU0 in parallel. Unfortunately this was a few years too late and development had shifted to next gen consoles.
 
There should actually be a separation between what hardware could do potentially and what it achieved in real world conditions by devs working to deadlines and budgets.
 
Also performance might not be good enough to use a feature, or it might be really difficult for other reasons (requiring some odd input format or something), also sometimes you have to work around hardware deficiencies (like not having registers in pixel shaders).
I just mean that just the hardware being capable of something is different from being usable in real world conditions.
 
Also performance might not be good enough to use a feature, or it might be really difficult for other reasons (requiring some odd input format or something), also sometimes you have to work around hardware deficiencies (like not having registers in pixel shaders).
I just mean that just the hardware being capable of something is different from being usable in real world conditions.

I remember the DC and its normal mapping abilities, that only one game used (Shenmue 2) for one object, in one scene (a scene where you view only that one object). It seemed more like an AM2 developer having some fun with the featureset than anything else ...

SimonF was sure it was fast enough to be used in-game, but I recall a developer wondering if the overhead would have been too high. That's assuming artists were making normal mapped content then, which they weren't ...
 
@ Function Now that you mention it, PD orta does look kinda "bumpy" :p That's such a beautiful game, one that i'm sure gamecube wouldn't be able to replicate completely because smile bit clearly used the Xbox's advantages. I'll have to read more about that game's development.
I don't know a lot, but a good place to start might be to compare some die sizes: https://forum.beyond3d.com/threads/console-die-sizes.53343/

... and then search for some benchmarks for the PowerPC 750CXe (closest none console chip iirc), scaling results for clock frequency. It's a while since I looked at any benchmarks, but what I remember being impressive was its performance in relation to its die size.

The GC and Wii have both been hacked, so you may be able to find some benches direct from those systems. There used to be some, so I expect they're still out in the wild somewhere ...

Yep i've looked at all the die sizes and have them memorized believe it or not, i've also got clock speeds and memory amounts of all consoles from 5th gen on memorized. I will try and find everything I can before I raise the topic.

"Emotional attachments are hard to shed. I've been there, every bit as much as you I suspect!"

I suspect you are closer than you think...

But as always, I invite you to try and find them! Try! Find a developer! Or find something showing GC can hit 50 mpps! Or find some developer docs on the T&L unit!

I've actually spent a while here looking for statements from developers, looking for old hardware benches, looking at contemporary T&L units (nothing matches GF4 even when clocked higher). Meanwhile, you dismiss everything and simply state that your entirely unsupported opinion is a truth that must be disproved through an increasingly narrow and moving set of goal posts.

How about you do some work now?

It's just an opinion supported by software, that's all ;)But yes I can also try to find any developer statements I can. And you know I haven't dismissed everything you said, well maybe initially but I came around on a couple things haha.
 
Last edited:
Thing about citing RE4 is that it was built from the ground up for GameCube and then released on a completely different platform in about 10 months. I wonder how God of War 2 would look on a GameCube if ported under the same conditions.

Unlike the GameCube however, PS2 was a complicated machine that required a lot of work to get good results from. In fact we know now from Corysama in the PS2 thread that late in its lifespan some developers figured out how to get the VU0 to trigger an interrupt on the main CPU core which would have allowed more games to use VU0 in parallel. Unfortunately this was a few years too late and development had shifted to next gen consoles.

Imo sony should have waited a year (or two at the most, though that may be too much), the ps2 was still selling and there were a few great games, and the annual titles were still coming out on it.
 
I have to say about Orta though, there are moments when I do notice a lack of poygonal detail, which would make sense if it used normal maps. In that case I think the trade off was well worth it. All at 60 frames too!
 
@ Function Now that you mention it, PD orta does look kinda "bumpy" :p That's such a beautiful game, one that i'm sure gamecube wouldn't be able to replicate completely because smile bit clearly used the Xbox's advantages. I'll have to read more about that game's development.

I have to say about Orta though, there are moments when I do notice a lack of poygonal detail, which would make sense if it used normal maps. In that case I think the trade off was well worth it.

The places I remember normal maps being used (it's a few years since I last played through) were on bosses such as one form of the final boss, boss at the end of the big battleship level, one of the earlier bosses. Normally on silvery shiney stuff. I suspect it was an early attempt at a metallic-y material and they applied variations to a few things. It was early days.

Orta started off as a DC game and was hurriedly shifted over to Xbox after DC was canned, so it was their first gig with the hardware, which was much more powerful than the DC and far more flexible. Impressive results all things considered. A second generation title would have been epic. I wonder if the normal mapped bosses would have been attempted on the DC...

Yep i've looked at all the die sizes and have them memorized believe it or not, i've also got clock speeds and memory amounts of all consoles from 5th gen on memorized. I will try and find everything I can before I raise the topic.

Nothing wrong with raising a topic and asking questions, just take a moment to consider the answers! :D

It's just an opinion supported by software, that's all ;)But yes I can also try to find any developer statements I can. And you know I haven't dismissed everything you said, well maybe initially but I came around on a couple things haha.

For reference, here's the Xbox game in question:

https://forum.beyond3d.com/threads/boss-calls-it-quits-polygonsheep-pics-added.802/#post-11459

If ERP says he measured up to 30 mpps, then that's about as solid a confirmation as you're going to get about something on the internet. And it's straight from the developer that did it! Remember this is very early 2002. Here's bonus alpha (pre-alpha?) footage:

https://forum.beyond3d.com/posts/11495/

The same ERP that worked out how to reduce CPU overhead on high poly workloads. ERP is a boss. I wish he still posted. :(
 
@ function

Looks pretty good for early development, I hate any time a game with potential isn't made because money.

Yeah, Panzer dragoon orta 2 could've gotten the most out of console. To be honest, I always thought the Xbox might've not been pushed quite as much as the gamecube. I think all 4 6th gen consoles had something left in the tank.
 
Just a small update, I could not find any benchmarks whatsoever of the IBM 750cl (links to benches are dead) which is supposedly broadway in desktop form, however what I am reading is what I originally thought, that the gekko does perform better per clock than the coppermine pentium inside Xbox. However it also seems that my initial thinking that they were on par (or that it was better, actually) was simply untrue, the gekko doesn't seem to overcome the large clock disadvantage. Which makes sense in hindsight and with a clear head.

A good bit of its advantages seem to stem mostly from the system being more balanced, and this is what i'm leaning towards now ; that Xbox WAS more *powerful* (and I do not say *capable* although we do know it does some things better) however it didn't have the equilibrium that all of gamecube's components had.

I will try and make a comprehensive post of my cpu findings soon, in a day or two probably.

---

On another note that developer at Criterion's statements on burnout 3 gave me newfound respect for a certain Wii title : Excite truck. Not only does it have vehicle deformation but it also has landscape changes, all with nice detailing and a large draw distance at 60fps.


Didn't think much of what the game was doing until I learned these things are pretty demanding.
 
Didn't think much of what the game was doing until I learned these things are pretty demanding.

Well, the issue with burnout was that they had all that, plus more cars and tons of particles and physics going on. Excite truck looks pretty good, but the wii runs a lot faster than the gamecube and has a lot more ram too.
 
Well, the issue with burnout was that they had all that, plus more cars and tons of particles and physics going on. Excite truck looks pretty good, *but the wii runs a lot faster than the gamecube and has a lot more ram too.
Physics wise it's comparable (though in ET you can't keep driving with damages, you crash and then recover), but without the scale and deformable environments of excite truck. When you consider the environments it *may* be doing more than burnout with regards to flexible geometry. The particles wouldn't be an issue they're just polys.

*That's the point, it took a big spec increase to GC's architecture for it to do these things, hence Criterion's issues. As a launch title it's almost as if the development team set out to do something gc couldn't.
 
Last edited:
Back
Top