Scalability of ND's Uncharted engine *spawn

[Laa-Yosh]

"Yes, Uncharted 2 looks good, but Naughty Dog does a lot of cheating, and their game design allows them to do that".

Which I think is unfair to the developer.

Why would that be unfair? Actually they'd probably agree with this statement.
I sure am ready and willing to admit that we cheat the hell out of our cinematics wherever we can...

It doesn't matter how they achieve their look, or what compromises they make to get there, at all. It's completely irrelevant.

That is not true when you're making comparisons and statements like 'the same engine can be used for completely different kinds of games'.

To somehow shrugg off their achievement as if to say "anyone could have done it under the same circumsatances" is silly.

We're not saying that, nor are we trying do belittle their achievement.

Seriously, we're going in circles for pages now.

 

[Laa-Yosh]

I'm not suggesting their engine could make an open world game with the exact same quality as Uncharted 2, visually, I'm suggesting that the engine would almost certainly be capable of some "open worl" environments, i.e. area's you can traverse in any direction.

If that would be true, then by that very definition ND wouldn't even be close to make the maximum use of the PS3's resources. Can't you understand the connection between these issues?

 

[function]

I would place UC2 as "others".
On rails would be like Sin and Punishment, RE: DC, Panzer Dragoon.
Sand box being GTA, Fallout, Farcry
Others being every others not above.

It's funny you should mention Panzer Dragoon. The (none RPG) games give you the freedom to move within a narrow corridor that follow linear, branching paths. The game controls your pace of progression, which guarantees time to load in the next section. You can't go back or race on ahead at will. You can look and shoot in any direction.

On the Saturn, Team Andromeda did two on-rails shooters and followed up with an RPG. The RPG had a very impressive engine (for the time) and was probably at the pinnacle of Saturn development, but despite this the free roaming sections lacked the intensity and visual splendour of the more tightly controlled battle and town sections, and indeed of the preceding on-rails shoot-em-up. Team members have talked about how development split off from that of the second (shooter) game, presumeably because they needed different things from their engines.

The Xbox game looked superb, and was a great shoot-em-up. I have no issue in thinking that if the game had been a free roaming flight adventure covering tens of square miles that there would have been compromises, and don't get upset that people think it only looks so good because it's "on-rails", because it's quite obviously true.

 

[betan]

If that would be true, then by that very definition ND wouldn't even be close to make the maximum use of the PS3's resources. Can't you understand the connection between these issues?

Don't know about him, I don't understand the correlation but I do suspect you have a very different definition for "maximum resource usage"?

 

[Laa-Yosh]

I'll try to phrase it as straightforward as I can...

IF you are willing to accept certain limitations
(the player can only travel between areas in a linear fashion, he can't see previous/upcoming areas, he can only backtrack a limited amount)

THEN you can create a streaming system that's far more optimized
(uses less resources, can read data from storage a lot faster)

SO you can use more memory for your textures and other stuff.
There are some other advantages, like less AI agents to spend CPU time on, smaller areas to put dynamic shadows on, and so on, but these don't have such a dramatic effect on a game's general looks.


The conclusion is that in order to make the best use of the PS3, the UC2 engine has to make some compromises - compromises that in no way detract from the game's quality.

 

[function]

If that would be true, then by that very definition ND wouldn't even be close to make the maximum use of the PS3's resources. Can't you understand the connection between these issues?

I've always assumed that games like GTA IV do things like calculate the optimal order to stream things in (based on perhaps player velocity and path, optimal seeking of the DVD laser head, etc etc) and have software modules dedicated to things like behaviour of none-present, persistent NPCs. You know, stuff like that.

Perhaps the UC2 engine can do all these kinds of things. I don't know. But if there was something like this that you didn't need to do using your custom engine, why would you develop it, polish it and put it in?

 

[betan]

I'll try to phrase it as straightforward as I can...

IF you are willing to accept certain limitations
(the player can only travel between areas in a linear fashion, he can't see previous/upcoming areas, he can only backtrack a limited amount)

THEN you can create a streaming system that's far more optimized
(uses less resources, can read data from storage a lot faster)

SO you can use more memory for your textures and other stuff.


The conclusion is that in order to make the best use of the PS3, the UC2 engine has to make some compromises - compromises that in no way detract from the game's quality.

Your conclusion is way wrong. The fact that streaming for non-linear games is costlier does not mean resources are not being used efficiently.

Even with the memory hit, it's not like they will keep unused memory space.
A buffer is a buffer, and it's used efficiently as any other memory.

 

[Laa-Yosh]

Er, you misunderstand me a bit... I mean that if you were to accept said limitations, and still would not optimize for it, then you would be wasting resources.
Like, implement the streaming system and persistent world support of GTA4 - in Uncharted 2.

Edit: function gets what I mean, though

 

[betan]

Er, you misunderstand me a bit... I mean that if you were to accept said limitations, and still would not optimize for it, then you would be wasting resources.
Like, implement the streaming system and persistent world support of GTA4 - in Uncharted 2.

Edit: function gets what I mean, though

OK, but I'm still trying to connect that to your response to tha_con.

why would ND not optimize for non-linear streaming if they were to use their engine, for example in next Jak (which has been confirmed I believe)?

 

[Laa-Yosh]

But that's for a later project, we're talking about Uncharted 2 and it's engine.

If THAT engine was built to support non-linear streaming then it would be wasting resources under THAT game, because it doesn't make any use of it.

 

[Shifty Geezer]

why would ND not optimize for non-linear streaming if they were to use their engine, for example in next Jak (which has been confirmed I believe)?

The it is the same engine? I think Richard's post expalins things perfectly, and pins down the argument. You cannot take U2 with it's scenery and extend that scenery to make a larger, open environment. In order to get a larger, open environemnt, you'd need to scale back assets and change the streaming model. You'll have a far smaller working space to fit what's visible, wiht the rest of RAM occupied by stuff waiting to be seen. This is a limitation of the streaming hardware. If we had nanosecond seek times and GB/s, we could stream in far more effectively. But we don't, so we have to choose what we load to view at any given moment, and have to load less to fit in more content.

Now if U2 uses huge amounts of procedural content, dynamic lighting, and such created on the fly, then it would scale well. However that's not really happening this gen.

 

[obonicus]

The spirit of the thread, as I understand scalability of an engine, is "if I were to license this engine and had 0 hours on my budget for engine-programming could I make an Oblivion-type game with the same level of visual fidelity of Uncharted 2?". This isn't an academic exercise: many engine licensing decisions are strongly based on the (lack of) human/material resources to modify the chosen engine.

I think saying 0 hours is going too far in the other direction, though.

That is why you rarely get an engine license that deviates a lot from the engine's debut title either in genre or gameplay, unless that engine was specifically made to be as generic as possible thereby sacrificing some other aspect.

What about UE3? Gears (and Gears 2) is very very much like Uncharted in how it's always shunting the player forward, it even has the infamous 'hand on earpiece' loading moments. A lot of comments I've seen here (on this forum, not this thread) indicate that it was not made to be as generic as possible, but rather that it was designed around the games it was used for. And still we have a pretty broad range of game types on UE3 (though a lot of those with 'bigger' worlds don't look so good). We even have what is often described as an open world game with very high graphic detail (Batman) albeit with a vastly limited scope (and with plenty of gating). Of course, now we might be back to trying to define what an open world game is, which isn't my point.

 

[Laa-Yosh]

Gears is a derivation of UE3; you don't license the Gears engine.
Streaming might have several possible implementations in the engine, too; or the licensee might be able to rewrite it as required.

The point is that those JRPGs are not using the Gears engine, just as any upcoming Jak game wouldn't use UC2's engine without any modification either.

 

[betan]

But that's for a later project, we're talking about Uncharted 2 and it's engine.

If THAT engine was built to support non-linear streaming then it would be wasting resources under THAT game, because it doesn't make any use of it.

Laa-Yosh, OK now your point is clear. The thing that bugs me about all this dance is though we don't know how far they developed it, what is implemented (and maybe disabled) in the engine specifically used in U2 thus we have no idea how much effort is necessary for them to "scale" it enginewise.

 

[Shifty Geezer]

Standard economics would have me think that they didn't create an engine they had no intention of using. If there are a few secret switches that enable 'sandbox mode' with reduced assets, we'll never know, but it'd also have to have been a conscious decision from the off, and not apply to the U2 build of that engine which couldn't do what it does otherwise!

 

[obonicus]

We're arguing semantics at this point. As per Laa-Yosh's last post and Richard's '0 hours' post, if we're going to make the distinction that Gears doesn't use UE3, but rather the Gears engine, that the criteria for 'suitability' is to work completely out of the box, then no games ever share an engine.

Therefore any broader argument about an engine's suitability for anything but the game it was designed for is nonsensical.

It seems to me, though, that this discussion has gone from trying to narrowly-define what an open-world game is to trying to narrowly-define what a game engine comprises, all in an effort to deny even the possibility (outside of the broadest, practically-from-scratch sense) that ND's U2 engine could be used for an open-world game.

That we're talking 'with modifications' should be utterly obvious from the very post. That no one's talking about a 'full rewrite' should be just as obvious. But those aren't the only two possibilities; there must be a reasonable middle-ground between these two extremes that'd better indicate 'suitability'. Otherwise, again, there's just no point in having this discussion at all.

That there already is an example of a smaller-scale open-world game built on top of a game engine designed for a U2-like linear shooter (actually, it'd be more accurate to say Gears-like linear shooter) indicates, to me, that this middle-ground can exist. Otherwise Rocksteady probably wouldn't say that Batman:AA is a UE3 game.

 

[Laa-Yosh]

No, this discussion is about this: some people's apparent idea of platform superiority.

 

[obonicus]

No, this discussion is about this: some people's apparent idea of platform superiority.

That's an ad hominem, and has no place in this discussion. If you can't refute my arguments you probably shouldn't post at all. To keep things simple, I'll quote your earlier post and reiterate my point:

Gears is a derivation of UE3; you don't license the Gears engine.

Do you realize the distinction you're trying to draw here? You are saying that Gears of War does not use the UE3 engine. In a literal sense this is true. But if we're suddenly going to be literal, then it's also true for every game that had more than 0 hours budgeted for engine customization. Taken to the literal sense, everyone's been wasting their time in this thread arguing about something that is trivially true.

And all of this because a few people in this thread have been laboring against the strawman that ND U2 engine could be used for an open-world game without modifications. And then have established the false dichotomy that either the engine must work without modifications or might as well be rewritten from scratch.

I'm just pointing out this sequence of fallacies because people were drawing conclusions that simply made no sense.

Yes, ND cheats via game design to make U2 look the way it does. Yes, ND would have to make compromises to make a true open-world game. Could they make GTA and still be recognizable? I'd imagine not. Can they make Batman:AA? I can't see why not. And the last one is, as I said, often considered to be an open-world game (certainly it's as open as AC).

Edit: To make my point about Batman:AA and Gears perfectly clear: Throughout this thread we've been told that ND cheats to get graphics the way they are, by constantly gating the player, by making certain things completely unreachable. I'm not arguing with that. This was used as proof, then, that ND's engine could not be used for an open-world game, as this gating would not be possible. However, Gears of War (and its sequel) was also a game that heavily used gating for the same purpose. However, Gears' engine, UE3 was used for a small-scale open-world game, in Batman:AA. This was also done through clever game design and tons of gating. Therefore I think that the proof against U2 becomes invalid. The argument then becomes that Gears and Batman do not share the same engine. Which, as I pointed out above, is problematic.

 

[ihamoitc2005]

This is a limitation of the streaming hardware. If we had nanosecond seek times and GB/s, we could stream in far more effectively. But we don't, so we have to choose what we load to view at any given moment, and have to load less to fit in more content.

Wouldn't hard-drive install improve streaming over Blu-Ray/DVD? Faster data transfer and seek times.

 

[Neb]

Wouldn't hard-drive install improve streaming over Blu-Ray/DVD? Faster data transfer and seek times.

CPU load might be higher.