Next generation console graphics

Lately i have been seeing a lot of articles on N4G claiming that either the next generation console graphics will be the same as it’s now, that nothing can beat the halo 4 graphics, that the next generation is not really necessary, and that a game already had next gen graphics, all of which are complete and utter bullshit, but im not going to stop there, instead lets wade trough the knee deep cesspool of video game journalism while wearing the boots of “well duh” and some kind of sanity based breathing/filtration system.
I am going to talk about the PS4, but the XBOX 720 will also apply since it will be in the same general spec range.

Rumors
Last February rumors began circulating that the PS4 where going to get an AMD A8-3850 APU and aRadeon HD7670 GPU, these rumors do seem to originate from one source (probably IGN) and besides the rumors from years ago about the next gen cell chip (the one that is basically 4 cells in one chip) nothing else has been rumored, which tells me it’s crap because of two things
1. total GFLOP is about 1248 if my math is correct, which is 3 times that of the ps3, it should be at least 10.
2. The rumor is the first and only time those chips has been mentioned and nowhere has anyone said it out loud, not even that AMD is doing the chips for the PS4, it sounds like pure speculation.
Add to the fact that Nvidia has basically done all sonys other GPUs and it starts to smell like fanboy fantasy to me.
Sony has always had an at least up to date hardware if not beyond that, so had they said A10-5700 and Radeon HD7970 then it would have made some sense, but now it didn’t and that makes this rumor false, even if sony wants off the shelf parts for cost control.

Projected system specs
So im not going to say exactly what kind of chips the PS4 will have, as i said the A10-5700 and Radeon HD7970 would work, as will pretty much any intel/arm/IBM/Nvidia combo as well as long as it’s about one order of magnitude faster (ie. 10 times) or more, i think the most important difference between the 720 and PS4 will be things like memory architecture, ram and things like that, because you can always trade some processing power for ram and vice versa.
I am going to go out on a limb and say that the absolute minimum is 2GB ram and that’s if your retarded, but it will probably be 4 or 8GB of unified memory, possibly also a small internal SSD for quick cashing.
The PS4 will have 4K resolution (games will still render at 1080P), but the 720 will not and the PS4 will also have a better blu-ray player to take advantage of that fact.
Over all i think the PS4 will have slightly better specs because that’s what sony does.

How does games deal with current system specs
Ok, now for some meat to this article,  console(or pc for that matter) graphics has 4 main areas of processing with fixed budgets, some of which can be traded back and forth between them, but not entirely, these are.
1. Physics
2. Memory
3. Polygons
4. Fragments

Physics dictates where everything is and makes everything more dynamic, even mario bros has physics so for the purpose of this article i mean mass ridgid body physics and particle systems.
Memory stores mainly texture data, so more means better textures, but it can also be used to store framebuffers and things like that.
More polygons means more and/or smoother geometry
Fragments make pixels, more power for fragments makes either more or prettier pixels.

Physics is primarily done on the CPU, but can these days be supplemented by the gpu, however this steals rendering time. on the Ps3 this was also the main job for the SPUs which made the PS3 really fast but hard to code for.
Using techniques like deferred rendering does get rid of a lot of superfluous fragments to render thereby making it faster, but it eats ram for lunch so to maintain texture quality you constantly have to swap textures which in turn eats CPU time.
Polygons and fragments are all done of the GPU traditionally (though the cell chip could take some of that work if needed), so it’s a direct trade-off between more polygons or more/better fragments.

So games generally handles this by analyzing the visuals themselves and adapting things like polycounts and shading methods to suit the situation and make the best choices and trade-offs, and i think games like uncharted 3 is definitely the pinnacle of that, but then they hit a wall in where they can’t make it look better with better coding, when that happens you either make a game with the same visuals but perhaps more spectacle to make it look better (like in halo 4 or Black ops 2) or you could be a bit sneakier.
Remember ICO and shadows of the colossus, great games visually speaking (for the ps2 at least), but they are shining examples of when you start adapting the game to the console or in other words, you make a game that always puts you in environments and situations that enable you to get the best possible outcome graphically speaking, thereby making it easier to make it look better because you don’t have to render a lot of other stuff.
The same thing is happening with games like “the last of us”, “the last guardian” and “Beyond: Two Souls”. They look better because they are better at not rendering stuff, and when they do render stuff the visuals and game is such that it doesn’t take the same amount of processing to achieve high visual fidelity.

What are the important Changes
Moving on to future systems it is somewhat easy to project ahead based on current hardware
1. both new consoles will have DX11 class hardware, that means almost infinitely programmable hardware, tessellation and strong GPGPU capabilities.
2. both will have at least 10 times the power of their predecessors.
3. Virtualization will be a big thing, so think rage but better.
4. ram will be less of an issue, current systems can cope well with maximum density textures, now imagine having 4 to 8 times that to play with.

What this means graphically
Poly count will increase, but imagine Nathan drake for a moment. Yes his mesh could use an update using tessellation, but i don’t think that you would need more than twice the polygons before you start loosing effectiveness of a high polycount, after all he doesn’t occupy that much of the screen, instead most of that is going into the environment, not as much making dirt looking nicer but in larger draw distances and more stuff  on the ground like leaves and stuff.
When it comes to fragment processing i think a portion of the extra processing power will be taken up by lighting, well shadows specifically, but also reflections and good AA, but the other significant portion will be allocated to particle effects like dust and such, because done correctly that stuff is magic.
The ram will also be used partially for higher detailed textures, but as i said before, current gen systems can already sort of max out maximum texture fidelity and virtualization with texture streaming does the rest, so a large portion will be used for other kinds of textures.
Like something i have always thought of doing is a projected 3D array of low res lightprobes (basically a virtual 5D texture), when filled with data and then properly interpolated it should create a pretty neat lighting effect.

The most visually significant thing i can think of though does belong in the realm of physics, remember the stunningly realistic facial expressions in LA noire, now imagine that on everything, imagine a soldier running in slow motion and with every step his entire body, uniform and equipment drops down only to be stopped moments later by the tug of the fabric.
Imagine killing an enemy by shooting of a branch from a tree with a rocker launcher or traping enemies under a tarp by shooting the wire that holds it.
Basically everything will be physically interactive in some way, if it’s not soft body, cloth dynamics, liquid dynamics or breakables it will at least collide properly and fall down without fuss.

What this means for games
The trend we have seen so far with the PS2 to PS3 is that games tend to get bigger and more expansive. Deferred rendering means that you can render a lot more objects at a distance without taking that much of a rendering hit, something games like infamous really exploit, For the next gen draw distance is definitely going to increase along with map size and a general feeling of having a living city full of people.
Take GTA3, you would drive around and see a few cars and the occasional pedestrian, fast forward to GTA4 and you have a living city that has plenty of cars and a lot more pedestrians, but it’s still only about a dozen or two at any given time and although you constantly meet people on the street it’s not hard to move around.
Now project forward to GTA6 (Because GTA5 will not be that dynamically different from GTA4), i think it will have a lot more people and cars (instead of dozens, think thousands), the city in question will feel more alive and behave as it should.
This will not just effect sandbox games like GTA and infamous, but basically every game, it will lead to games that traditionally had more of a linear approach will now have areas moth more freedom of movement.
Like in god of war you could come to a huge cave system with a lava river crossing it, you can see the exit and the various ways of getting to there, and you have to plan ahead more than just navigate the course.

As mentioned previously a lot more of the game world will be physically interactive, this was big with halflife 2, now imagine the fun with next gen physics, It will force a whole new look at what gameplay could be.
Similarly animation and animation AI will be beefed up, dont think an enemy should be allowed to stand, just shoot him in the knee.

Generally speaking when it comes to enemies we will be looking at more not bigger and when it comes to the game map it will be bigger and further, both of them will feel more alive.

Conclusion
Whenever you see an article saying this current gen game is next gen quality, that a certain game cannot be improved upon with next gen hardware or that next gen consoles aren’t really necessary since you can do the dame game on the current generation, just keep one thing in mind, they are so very very wrong.
Take a look at the recent animated movies from disney, pixar and dreamworks, with a few tweaks those should be possible on next gen hardware with virtually no visual difference.
Even a lot of the fully CG stuff in Avatar should be possible.
Trust me on this, you have seen nothing yet


3 Comments

  • By uh20, October 31, 2012 @ 05:01

    all consoles are getting power boosts and fully supported high quality, of course quality wont go up much, but physics, and abundance of quality stuff will grow due to the hardwares heightened limits

    course i dont care, i prefer to not look to kindly at the companies that are proprietizing and adding developer fees
    which basically excludes every platform but android,linux,webos

  • By H3g3m0n, November 7, 2012 @ 00:31

    The reason graphics haven’t been improving is because of consoles. Every major game has to be able to run on the Xbox/PS3 and possibly Wii (in some cases the Smartphone market too). Having a new generation of consoles will allow the graphics to move forward. But I think the real stuff of interest will be the extra ram/cpu. Something like Minecraft just about works on a Xbox but It’s unlikely we would ever see something like that from a AAA company due to things like the time it takes to load chunks and so on and that’s just a small indie game. We could start to see fully voxel based game engines (search for gigavoxel for an example). We could see fully raytraced games. Or at least hybrid ray-tracing. And just a huge increase in game complexity, particularly for sandbox style games.

  • By Peter Wallström, November 7, 2012 @ 01:46

    Well yea, improvements will come more from ram+cpu, or rather the non rendering side of games, give a PS3 4 times the ram and you would see some really nice looking stuff, but it still plays the same.
    Though I’m not sure about gigavoxels or any form of ray-tracing, the first one takes up a lot more ram than you could imagine and the fact that we haven’t really developed either the rendering or animation techniques for it.
    Maybe some kind of hybrid approach would work, but it will probably be more of a curiosity rather than the norm.
    Minecraft works because you have meter cubes, cut that down to half meter cubes and you suddenly increase the memory and processing requirement(but not rendering) by a factor of 8, which would make a non blocky game run too slow even on top of the line hardware.
    Though on the other hand one of the things I’m working on is metaball-textures which could make things like this way faster, but it’s still just a theory.

    Ray-tracing on the other hand still takes up to much processing power and would likely need specialized hardware to run well, especially when faking it for a fraction of the cost produces almost the same result.

Other Links to this Post

RSS feed for comments on this post. TrackBack URI

Leave a comment

*


WordPress Themes