29 replies to this topic

[Promit]

I just read this article about Unreal Engine 4:
http://www.wired.com.../05/ff_unreal4/
I'm warning you now, it is extremely melodramatic and generally kind of a shoddy article. If someone has a better summary of the new engine that would be great, although I believe not much is known at this point.

Here is what I do know: I'm bored. Partly because the Wired writer is a bit of a tool, and partly because I'm just not seeing what is exciting about all this. Sweeney and the rest of the crew at Epic do some excellent technical work, though I'll avoid commenting on their game design or art direction. But I don't particularly care for the screenshots being shared. Maybe it looks better in motion, but I'm not seeing anything new and different.

It might just be an allergic reaction to the hype, here. Unreal's strength was always about packaging up the bleeding edge with heavy tooling and making it accessible to developers who couldn't afford to do the engineering work. That deserves real credit, and is probably most responsible for bringing the industry as a whole to a new level of excellence. What I don't see is any evidence that Epic have really managed to advance the bleeding edge of real-time graphics.

One of my high profile friends in the industry suggested that this death march to the top of quality graphics might be completely unsustainable, as art budgets become improbably gigantic. Much like desktop computing, we might be reaching the point where graphics has simply peaked for your standard consumer and other things begin to take priority -- price, convenience, features, etc. That would be similar to the forces that brought netbooks and tablets to the market in strength. It's the same force that pushed laptops past desktops over the course of the last decade. I don't think we're there yet in games; the consoles will give us one more solid generation of leap forward. But I do think the day is coming where sheer compute power is not what drives mainstream consoles.

Bit of a rant there, but I'm curious to hear all of your thoughts on the matter.

Edited by Promit, 17 May 2012 - 02:18 PM.

[MJP]

I'm not sure that we should be using those screenshots as a means of judging "next-gen" graphical quality. They really just strike me as images from a tech-demo that was showing off some specific new-features (probably their new dynamic GI system, or whatever it is they're using).

[InvalidPointer]

I'm sort of curious about the GI system on display there-- I was going to make a thread on the subject but saw this one. Going to guess some sort of fast voxelization and LPVs?

[Frenetic Pony]

Yeah, it's just 14 people making a tech demo to show stuff off for prospective buyers. Plenty of polys and some brdf jiggery, GPGPU particle simulations, etc. The dynamic global illumination looks interesting. I mean, I expect it's a low res voxel grid ala light propagation volumes. But it also appears to include specular and dynamic occlusion, which is certainly a welcome advance.

Edited by Frenetic Pony, 17 May 2012 - 03:55 PM.

[jameszhao00]

I wanna see a demo that shows off dynamic skeletal animation. Something similar in quality to the video below, but reacts properly to collisions

http://youtu.be/DuEzU6Szi5c?t=1m

When I first saw the video above it seriously looked like CG (not because of lighting/particle effects, but because of animation quality)

Edited by jameszhao00, 17 May 2012 - 04:27 PM.

[Frenetic Pony]

One of my high profile friends in the industry suggested that this death march to the top of quality graphics might be completely unsustainable, as art budgets become improbably gigantic. Much like desktop computing, we might be reaching the point where graphics has simply peaked for your standard consumer and other things begin to take priority -- price, convenience, features, etc. That would be similar to the forces that brought netbooks and tablets to the market in strength. It's the same force that pushed laptops past desktops over the course of the last decade. I don't think we're there yet in games; the consoles will give us one more solid generation of leap forward. But I do think the day is coming where sheer compute power is not what drives mainstream consoles.

Quite a bit of a popular misnomer there, heck even CEO's seem to believe this. But seriously, what the heck are art departments going to spend extra money on? More polys and higher rez textures? Those are free! Global illumination? Your programmers are already paid, they need to be doing something. 1080p, higher res shadow maps, more lights? I don't know, maybe when profiling the performance I'll change a few numbers in a config file, tell the artists they can go nuts.

Making games look significantly better could easily be done for relatively close to zero increased costs in terms of art, and upgrading your engine is also another thing developers already budget for. Unless developers go absolutely nuts (and I don't see why they would) and start modeling dynamic sweat systems so correctly simulated sweat stains appear on clothes or something then there's little too worry about in terms of art costs increasing. Now gameplay is a different story. Much more power means more stuff for programmers to do, means possibly you want more programmers to do more stuff. But that's hugely variable on the project your doing. Besides, with digital distribution devs and publishers should hopefully be earning a larger profit margin on each sale AND cutting down on used games as well.

Edited by Frenetic Pony, 17 May 2012 - 06:00 PM.

[Promit]

I wanna see a demo that shows off dynamic skeletal animation. Something similar in quality to the video below, but reacts properly to collisions

http://youtu.be/DuEzU6Szi5c?t=1m

When I first saw the video above it seriously looked like CG (not because of lighting/particle effects, but because of animation quality)

I don't see anything dynamic in the skeletons here. it looks like the same old animation stuff; lots of prebaked animations being selected somewhat intelligently. They're not even blending between all of them.

[jameszhao00]

I wanna see a demo that shows off dynamic skeletal animation. Something similar in quality to the video below, but reacts properly to collisions

http://youtu.be/DuEzU6Szi5c?t=1m

When I first saw the video above it seriously looked like CG (not because of lighting/particle effects, but because of animation quality)

I don't see anything dynamic in the skeletons here. it looks like the same old animation stuff; lots of prebaked animations being selected somewhat intelligently. They're not even blending between all of them.

Yea that Dante's video is all prebaked stuff (I wish it had dynamic animations that reacts properly to collisions). In that particular scene excellent character animations is the major factor in 'this looks spectacular'. As a hobby graphics programmer I fawn over the latest and greatest lighting effect, but as a consumer I just want better animations, original art direction and original gameplay.

[Krohm]

I don't particularly care for the screenshots being shared.

Seconded. The initial screens for UE3 gave me quite of excitement and as far as I've understood, they actually were meant to be real content for ... BulletStorm?
Those do not seem too impressive to me. I won't comment on the technical features but I would have selected them with more care considering the public consumption.

[taby]

It sounds to me like they have fire and smoke and entirely dynamic material-based lighting? If they can combine all three effects into one, similar to what guys like Jensen and Fedkiw could do only 10 years ago (http://physbam.stanf...ford2002-02.pdf , Fig. 9 and 10), but in realtime and on currently available hardware, then colour me fully impressed (not that UE3 is anything short of awesome already). And realtime, realistic optics? The possibility alone of realtime, dynamics light caustics slamming through dynamic, partially self-illuminated inhomogeneous media just makes me want to do a happy dance. This was precisely why a Softimage license cost the price of a car 15 years ago.

Perhaps I'm just falling victim to wishful thinking / excessive extrapolation and inaccurate reporting. Even without this wishful thinking related to fluid dynamics and lighting though, I agree that the vast majority of those Wired screencaps do not match the potential that is generally described in the verbage. I mean, some of the screencaps are totally amazing though, like that one with the knight walking through the busted up castle a la Infinity Blade on steroids, OMFG drool, swoon. So yeah, when are the "official" demo reel screencaps / videos coming out? Are they just teasing for now? I'm pretty sure I saw some video related to the illuminated bowling ball rolling around on its own inside a fairly dark circus tent a number of months ago, and yeah it was frickin intense, so why wasn't that included (or was it from a different studio, and I just be confusing stuffs here)?

Edited by taby, 18 May 2012 - 01:47 AM.

[David Neubelt]

Making games look significantly better could easily be done for relatively close to zero increased costs in terms of art developers already budget for.

How do you figure that? Increasing polycount and texture resolution and the amount of maps allowed per material means that it takes longer to make the same amount of content. If it takes longer to make the same amount of content then it costs more money on content to make.

In my experience, the biggest increase in cost for development has been the increase in the complexity of art content.

-= Dave

[n3Xus]

How do you figure that? Increasing polycount and texture resolution and the amount of maps allowed per material means that it takes longer to make the same amount of content. If it takes longer to make the same amount of content then it costs more money on content to make.

Most character art is already created in high-res and then baked to some low-res model. Not sure about the world meshes/textures.

[jameszhao00]

Found this actually readable article (with a 'feature list'). Seems weird that the article doesn't seem to have sources though.

http://gameindustry....-First-Look.htm

I'm most excited about the move away from unrealscript and to a Kismet/dynamically compiled C++ hybrid (Unity style).

Edited by jameszhao00, 23 May 2012 - 10:07 AM.

[はとぶ]

puff!... I'm not sure about how great moving away from UnrealScript would be, while it does has a performance cost on the game, it also has several advantages on the production/development side. It presents a clean way to manage the whole networking of the game and the language was designed to work on a state-machine manner out of the box... no to mention that you can use the "class" class to store a class-type on it, or all the properties to interact with the editor.

moving to pure C++ and leaving all this language designed features a side would be (IMHO) a great disadvantage... however, I do believe that Unreal should use more extensively the C++ part of the engine to do most of the game logic on it, and manage on the UnrealScript side only a subset of what it does now, focusing it more on the content side, and not that much on the architecture side (pretty much like CryEngine does now with C++ and LUA).

I think that a drawback of current Unreal's UnrealScript is that you have to write the whole logic of the game on the UnrealScript side, and just very few specific things on the C++ part of the engine.

Cheers!

[Frenetic Pony]

How is moving to pure C++ going to make it easier for development versus a nice, more noob friendly scripting language??? Maybe there's something big they aren't telling us, after all they said that you can "build an entire mod" using Kismet. So... is that somehow more like their new scripting language, and you only use that and C++?

[taby]

Oh yeah, realtime dynamic Lambent blood that actually casts light. I'm also looking forward to that, and makes me wonder if the faked Lambent blood in Gears 3 inspired the material-based lighting in UE4, or the other way around.

Edited by taby, 26 May 2012 - 03:48 PM.

[NickUdell]

How is moving to pure C++ going to make it easier for development versus a nice, more noob friendly scripting language??? Maybe there's something big they aren't telling us, after all they said that you can "build an entire mod" using Kismet. So... is that somehow more like their new scripting language, and you only use that and C++?

Kismet is a graphic programming language. You link components such as conditions, events, etc together like a flow chart.

[Frenetic Pony]

So here is the demo: https://www.youtube....d&v=OZmRt8gCsC0

That looks way better than the screenshots had led me to believe, double so if that's if all that lighting is in realtime with no precalculation as I understood that haphazard excuse for an article to be claiming up above. My biggest question would be how, if they had no precalculation/artist finaggling, they got the interior's ambient so dark compared to the exterior stuff. I suppose Notch managed it with Minecraft of all things, but I don't think I've ever seen how.

And watching it a second time, I notice there's color bleed from object not directly in light, and it doesn't seem like there's any especially noticeable secondary light bleeding through walls either, so secondary occlusion of some kind seems evident (or just trickery and small radii). Whatever it is Sweeney is doing here, it's something I'd love to know about.

Edited by Frenetic Pony, 08 June 2012 - 02:23 AM.

[Eternal]

Video with dev commentary: http://www.youtube.com/watch?feature=player_embedded&v=MOvfn1p92_8

After seeing the video and hearing "voxel lighting" I expect their lighting tech to be http://research.nvidia.com/publication/interactive-indirect-illumination-using-voxel-cone-tracing or something very similar.

[Yours3!f]

Ue4 uses a voxelied light system. I heard it in one of the dev videos. And i think they're just doing a deferred shading pass with the added lpvs.
What I found stunning was the particle system. If they combine this with the new ui and workflow then games will improve significantly as the baking pass can be excluded now. This means that devs can focus more on the game play side.