Killzone 2 technology discussion thread (renamed)

Status
Not open for further replies.
Why would they suddenly need fairies if it wasn't in the design document? I realise that plans change and all, but you shouldn't chop off your foot on the offchance that you might lose one of your shoes.

Yes, plans change and yes, you should chop off your foot on that chance when it comes to constructing software. We are moving to a broader area here, but the general idea is that things change all time, requirements change, design documents change, and you can't fight it: embrace change and do whatever you can to keep things simple and flexible so you can quickly adapt when and if it's needed. If you don't do it, the project will inevitably suffer.

My personal idea (and I repeat it's my personal idea, because not everyone agrees with me on this) is that a more flexible and simple design is generally better than a less flexible one. Simplicity first, flexibility after it.

I would guess that the developers have made their decisions based on a lot of thought and planning, and with the artists in mind. Let's judge their decision after the game is out?

I absolutely believe they did. I'm saying that from my limited perspective and on the basis of my limited information and my experience on this particular subject, I don't agree with this trade off, because I find it too limiting for the artists.
 
My personal idea (and I repeat it's my personal idea, because not everyone agrees with me on this) is that a more flexible and simple design is generally better than a less flexible one. Simplicity first, flexibility after it.


I'm not following your logic here. Flexible & simple > less flexible (and more simple)? But simplicity > flexibility??

And doesn't a deferred renderer precisely show the trade off of flexibility for simplicity?



To me, the very definition of an optimized system is a focused system. Focus means trading in flexibility for some other bonuses. You try to be a jack of all trades, you end up being a master of none.
 
Artists should be provided with some powerful and extremely simple tools to use (we shouldn't harm their creativity giving them some overfeatured piece of unnecessary complicated crap to work with), but we shouldn't aim to give them everything they ask for or all the flexibility they want, in fact a few restrictions and limits are exceptional ways to give a game that particular special look that your game and your only game has,and it also stimulates artists creativity as they manage to find novel ways to use existing tools and technology to do something they were not designed for (snow shader used to shade water ftw!)
 
I'm not following your logic here. Flexible & simple > less flexible (and more simple)? But simplicity > flexibility??

And doesn't a deferred renderer precisely show the trade off of flexibility for simplicity?

I don't think that in the current generation, to achieve the same goal, a deferred renderer is simpler than a forward renderer to implement.
Even for the simple reason that you still need to render some parts of the seen with a forward renderer (transparency), so you actually need to support both.
My understanding is that in this situation, they chose a deferred approach beacuse they wanted many lights in the scene, which is a perfectly understandable requirement and they sacrificed many other features for this.

For me, in general, given a certain set of requirements, a simple design is always preferable to a more complicate one and a flexible design is more prefarable than a inflexible one. Often these two forces are pushing in different directions so you usually have to strike some kind of balance, and I almost always tend towards simplicity anyway.
Which means that in the curent generation, I would avoid a DR, because, in my opinion, it's too complicate to implement given a certain goal and too inflexible to achieve that goal. If the goal is many lights in the scene, I would still go for a forward renderer, but this is my personal preference given my experience on the subject.

On the other hand, if you consider a DX10-like GPU, I would probably prefer a fully deferred approach: I'm actually writing a DR engine (again!) on DX10 this time for my own pleasure in my spare time to try out a couple of ideas and indeed it is lovely.

To me, the very definition of an optimized system is a focused system. Focus means trading in flexibility for some other bonuses. You try to be a jack of all trades, you end up being a master of none.

I agree with you on this. I'm not advocating implementing every possible feature for the sake of it, actually the contrary. I'm advocating keeping things simple and as flexible as possible (because requirements will change!) , without being tempted by the dark side of coding :D

Artists should be provided with some powerful and extremely simple tools to use (we shouldn't harm their creativity giving them some overfeatured piece of unnecessary complicated crap to work with), but we shouldn't aim to give them everything they ask for or all the flexibility they want, in fact a few restrictions and limits are exceptional ways to give a game that particular special look that your game and your only game has,and it also stimulates artists creativity as they manage to find novel ways to use existing tools and technology to do something they were not designed for (snow shader used to shade water ftw!)

Well, I can't say no to an artist, I really can't. I always answer yes, I can't help it. Unless the hardware can't do what they ask, and in that case it becomes really painful for me to say no. That's why I was the only coder at the Art of Crysis presentation. I like them. I really do :)

So, if an artist asks something and I can do it and have time to do it, I do it. I'm not a big fan of the concept that limits stimulate creativity. I don't believe in this. I think that simple and flexible tools stimulate creativity.
 
On the other hand, if you consider a DX10-like GPU, I would probably prefer a fully deferred approach: I'm actually writing a DR engine (again!) on DX10 this time for my own pleasure in my spare time to try out a couple of ideas and indeed it is lovely.

Thanks Fran for your posts. I hold similar view as you but I have a question (I am not a graphics guy). In your statement above, what aspect of RSX holds you back in DR implementation ? I am under the impression that the issue of DX9 vs DX10 is irrelevant in PS3 (You essentially operate on the metal or some thin layer on top, I think).

In your view, are the SPUs suitable for DR design and implementaiton (e.g., How do they help to bridge any design or implementation issues) ? Based on your experiences and imagination, what other directions can they take moving forward (beyond the lighting advantages you mentioned) ?

I am assuming you're not under Sony NDA.
 
Last edited by a moderator:
Well, I can't say no to an artist, I really can't. I always answer yes, I can't help it. Unless the hardware can't do what they ask, and in that case it becomes really painful for me to say no.

Do you only have extremely cute and hot artists at your office or something :p ?
 
2 slides from E3 (and Develop) from Impress Watch

The second one seems to describe what PS Edge does.

3de388.jpg

3de422.jpg
 
Fran said:
So, if an artist asks something and I can do it and have time to do it, I do it. I'm not a big fan of the concept that limits stimulate creativity.
It's not so much a stimulation as it is a fundamental part of creative process. Art is in great part defined by the very limitations of its expressive medium.
Eg. poetry is what it is because of limits of the written/spoken word. And I am completely convinced that some things can ONLY be expressed through poetry, and no other medium would be able to convey them remotely as well, even directly tapping into artists brain (thus removing 'all' limitations).

In my own meandering experience, every time when I allowed artists almost completely free reign, results never justified the means - we'd get marginal improvements at exponential increase in resource costs (which in respective situations wasn't a problem, but the point is cost/effect ratio was bad).
Over time the results tend to improve as they get better at utilizing the increased freedom, but that just means getting used to the higher ceiling of limitations.
The problem I'm getting at is that keeping "ceiling" in a constant state of flux runs a danger of artists never getting the feel of efficiently working within constraints (which will always exist in realtime no matter what magic we try as programmers).
All IMO of course.

I don't believe in this. I think that simple and flexible tools stimulate creativity.
I'm completely in agreement about tools, but flexible and simple tools do not preclude limitations to the expressive medium, as I pointed out above.

Btw, I second other people's questions, what DO the artists look like at your studio? And if they are like Panajev describes, are you guys hiring? :oops:
 
I find this philosophy of 'limitations make better game art' quite a bit silly, personally.
If poly count X won't let me model five articulated fingers, then the resulting blocky hand won't look better, nor would an 8-sided wheel on a car, would it? If our texture artists can't use a specular map, then the metal surfaces won't look different enough from non-reflective surfaces either. Would Poliphony Digital's cars look better without HDR reflection maps? How about a Gears without normal maps?

If someone can make good looking stuff from less, then it's usually despite the limitations and not because of them. And the one unable to use the higher potential properly is probably just someone not talented/experienced/disciplined enough... Our texture crew has worked on all our cinematics, obviously, and they've done character textures for two AA games already (one is mentioned here on B3D quite often), so there are examples.

Then again, I'm no coder, just an artist... (and I kinda hate the english language for not making the distinction between someone who makes stuff for a living and someone who's making fine art) It's just that I feel Fran's opinion is kinda oppressed here.
 
Thanks Fran for your posts. I hold similar view as you but I have a question (I am not a graphics guy). In your statement above, what aspect of RSX holds you back in DR implementation ? I am under the impression that the issue of DX9 vs DX10 is irrelevant in PS3 (You essentially operate on the metal or some thin layer on top, I think).

I'm afraid I can't answer this question since I don't know the RSX, I work on the 360.

Do you only have extremely cute and hot artists at your office or something :p ?

We have a couple of pretty ones at last, especially a cute italian :D


In my own meandering experience, every time when I allowed artists almost completely free reign, results never justified the means - we'd get marginal improvements at exponential increase in resource costs (which in respective situations wasn't a problem, but the point is cost/effect ratio was bad).
Over time the results tend to improve as they get better at utilizing the increased freedom, but that just means getting used to the higher ceiling of limitations.

I agree on this: at some point there is a wall they have to hit and they have to hear that "no", because implementing that last effect in the last level for that single mesh is probably too expensive and I can use those resources in a better way.
But in the general case, I don't think you can obtain a better experience with a very limited PSX than you can get with a PS3, as much as the first one is stimulating creativity by its limits.
I like your definition of art though, it's very interesting: any reference I can read about that?

Btw, I second other people's questions, what DO the artists look like at your studio? And if they are like Panajev describes, are you guys hiring? :oops:

Sorry, that artist is (almost) taken :D
 
I find this philosophy of 'limitations make better game art' quite a bit silly, personally.
If poly count X won't let me model five articulated fingers, then the resulting blocky hand won't look better, nor would an 8-sided wheel on a car, would it? If our texture artists can't use a specular map, then the metal surfaces won't look different enough from non-reflective surfaces either. Would Poliphony Digital's cars look better without HDR reflection maps? How about a Gears without normal maps?
But those examples are working within the artistic limits of the engine. GeOW doesn't have 100 dynamic lights, does it? They chose one artistic feature over another. What good would Okami gain with specular lightmaps? The choice in art style defined the requirements of the engine, and features beyond that artistic choice are redundant. Like selecting a range of colours in your palette, you select a range of features that you optimize for the hardware, those features helping to define the look of your game.


If you added all the possible features, the end result is quite likely to look a mess, and even if it was handled tastefully, the investment would exceed the returns. If you add the option for specular maps just in case the artist wants to add one map to one creature, you're adding all that workload. Feature creep is a curse! Setting on the design decisions early on, if they can remain static, allows the engineers to create optimal solutions.
 
Feature creep is a curse! Setting on the design decisions early on, if they can remain static, allows the engineers to create optimal solutions.

Feature creep is indeed a curse!
And early design decisions is as well a curse. Requirements won't remain static and the design has to adapt. Big Design Up Front just doesn't work.
 
Status
Not open for further replies.
Back
Top