Digital Foundry Article Technical Discussion Archive [2011]

[Neb]

About the 3D:

“James Cameron has seen Crysis 2, and he loved what he saw because his eyes are trained for 3D more than anyone else’s, and that makes me confident that we have a super-high-end 3D experience on all formats.”

http://www.next-gen.biz/news/crytek-james-cameron-loves-crysis-2

Can't complain

Seems then to be really good and convincing 3D considering Cameron should know well about 3D results and tech considering the movie he directed. :smile:

Is the lighting the only real technical stand out here? It seems like it, from the report. It seems DF briefly covered the texture and geometry pop-in and shadow issues at the end. I would be interested in knowing, if any other technical implementations make the technical radar. There were a lot of compliants about laggy controls, but DF said they were swift.

They dont seem to go to technical and last article lines shows why.

 

[Neb]

True, but is it worth the huge IQ drop to have true S3D?

Yeah especially when compromisses is sub-HD resolution and graphics rendering degradations or HD S3D with graphics that looks like a cross between this-gen console and last-gen console games in HD. That said it's not good that vsync is tunred off in S3D for C2 360 demo.

 

[Neb]

Also from article:

Way back in December 2009, our 'Can Consoles Run Crysis'? feature saw us attempt to recreate Crytek's demos within the existing PC game, and guest contributor Nebula rebuilt a PC config based on the visual make-up of the 360 and PS3 footage. His conclusion was that CE3 on console was based on the PC game's "medium" quality setting with some elements dialled back, while others operated on the equivalent of the "very high" setting. Fast forward to 2011 and the ability to see the consoles' actual .cfg file validates the majority of his findings.

Spoiler

But then again, we really are seeing just a very small section of the game and it is clearly way too soon to be coming to any kind of definitive conclusion. The second, secret Crysis 2 demo level already hints at a game capable of so much more than we're seeing thus far in the Skyline map.

 

[Laa-Yosh]

True, but is it worth the huge IQ drop to have true S3D?

It was the same question about 32 bit, and the answer is that yes it is. Or rather, that we'll have IQ better than today, with 3D on top.

See, eventually we'll get more and more performance that can be allocated to various uses, and there'll be several possible ways to find a good compromise for stereo 3D. I'm not saying that it has to be two completely independent fully rendered frames, as that is the no compromise, +100% work approach; but Crytek's +1,5% work clearly lacks in quality. Basically, it's another way to drop IQ.

 

[Laa-Yosh]

Seems then to be really good and convincing 3D considering Cameron should know well about 3D results and tech considering the movie he directed. :smile:

Or maybe he just had a short look at it? I can't see the man having enough free time to get a tour at Crytek...

 

[Neb]

Or maybe he just had a short look at it? I can't see the man having enough free time to get a tour at Crytek...

Possibly but now we are just going into speculation mode. What stands is what he said be it short or long observation time with game in 3D. It clearly was not bad.

 

[makattack]

We have 2D games that made huge IQ drops to have...2D! At least this adds something to the gameplay experience, right? Plus, there are "true" S3D (if something is labeled "true", that means the others are false) games that are 720p. Some devs seem to be able to improve on the original 2D 720p IQ even after going to 3D 720p for the sequel.

It really depends on what the game's displaying *and* doing in terms of simulating physics, ai, modelling a world, etc., doesn't it?

Here's a nice CS academic paper on S3D effects, using an implementation of 3D Asteroids in Java (the original game from the 70's used an unique vector graphics processor).

Oops, forgot link:PDF (via Google Docs) paper on Effects of S3D Asteroids

To produce the stereo pair, viewing features of Java 3D’s scene graph were exploited. Two independent views of the scene were produced (see figure 14 (b)). One view was translated to the left while the other was translated to the right to give the left and right eye perspectives i.e. the stereo pair. The actual amount of translation producing the best stereoscopic effect was determined through trial and error. This turned out to be 3mm each way. One further issue to address was that of the images’ scales. Given that the two images were horizontally interleaved by the display to produce the stereoscopic effect, the individual images’ horizontal scales needed to be halved in order to obtain the correct scale following interleaving.

 

[kagemaru]

Great article. I wish the game offered an offline mode so they could possibly measure the response time of the controls since many people complained about laggy inputs.

Also it may be pointless, but I would curious to have read some theories or thoughts to why the issues are present, especially the geometry pop-in.

Hopefully another, more in-depth analysis is done for the final game. Always appreciate these.

I'm just saying that the compromises to almost every other technique used in the game probably leaves plenty processing power to do HDR with GI. I mean MLB: The Show does HDR (day/night cycles) with GI at 1920x1080/60fps. That's all I'm saying. It's impressive, but can't any team of devs make 1 technique impressive by sacrificing most other techniques? Consoles CAN supposedly, handle 4096x4096 textures, but at what cost?

EVERY game makes compromises, that's the nature of console development. You can see these same compromises in other high end games that still look great: cascading shadows in Assassins Creed 2, texture pop-in in Gears 2, low res particles in KZ2, and so on.

Also, you can question if the trade off is worth it, but I'd bet if they scaled back the lighting to improve the cascading shadows and texture streaming, the game would lose it's graphical identity and in the end look worse. Not sure if that example makes sense from a technical PoV (trading HDR and GI for texture streaming and shadows) but it should make my point clear.

Basically I'd wait for the final code, and info on how consoles perform during SP campaign, before judging anything.

 

[Laa-Yosh]

To be honest I think that most of the audience doesn't notice or care about the GI being real time and they'd be perfectly happy with static light maps. It's also perfectly possible to create a unique look with no dynamic GI; in fact it'd also allow for larger open areas as seen in other engines that prefer less ambitious tech. But that's a different issue.

Pop-in textures and shadows and objects are however very noticeable, even to the completely untrained eye. Crytek may have mad a bad compromise here, advancing technology mostly for the sake of it.

 

[_phil_]

Yeah sure ... Yerli said Cameron loved crysis2 in 3d , full hand of salt and all...
I bet he saw the reall stereo ,not the console trick.

Depth-based 3d can't have wide enough paralax , you'd have too much pixels to reconstruct to patch the holes.
This trick really can't compete with 2 separate renderings at least for experienced people trained to wide paralax true stereo ,it's actually low standards.

 

[Ruskie]

To be honest I think that most of the audience doesn't notice or care about the GI being real time and they'd be perfectly happy with static light maps. It's also perfectly possible to create a unique look with no dynamic GI; in fact it'd also allow for larger open areas as seen in other engines that prefer less ambitious tech. But that's a different issue.

Pop-in textures and shadows and objects are however very noticeable, even to the completely untrained eye. Crytek may have mad a bad compromise here, advancing technology mostly for the sake of it.

How are they suppose to bake it when Crysis 2 is game with dynamic TOD?

 

[Laa-Yosh]

A game like AC or Oblivion might require constantly changing time of day (although part 1 was pretty fine without it) but in a shooter it's just showing off for technology's sake. It's very likely not a continuous world but a set of levels with a path to follow and little backtracking, so the player is unlikely to spend enough time in a single location to actually experience night and day transitions.
They could change time of day between levels the same way any other FPS does and be OK with it. Now they're probably looking forward to get more criticism about the pop-ins compared to what praise they'll get for dynamic GI.

 

[Trejser]

How are they suppose to bake it when Crysis 2 is game with dynamic TOD?

Crysis 1 have real time or baked GI?

 

[Neb]

Crysis 1 have real time or baked GI?

It doesnt have the GI system with indirect color that CE2 has. It does have cheap "GI" like other games and it pretty much everything is realtime and game uses dynamic TOD. AFAIK CE3 GI system is continuation and improvment of the method they didn't finalize for CE2/Crysis.

 

[_phil_]

CE1 didn't have GI.Just an ambient cube.
GI on CE3 is built on top of many pre existing papers, technics and methods.(Just to point out , complex tech never comes out of nowhere by a single individual or cie).

 

[Laa-Yosh]

Yeah, what's important about graphics rendering tech is to spot the papers that have relevant research, and to make a good implementation. Carmack's been very good at both, but he also prefers to focus on just a few things at any time, whereas Crytek is going at it on all fronts.

 

[Tap In]

The Crysis 2 multiplayer demo is a curious beast. There is a palpable sense of technical accomplishment in CryEngine 3's "tentpole" technical elements - the full, real-time global illumination system in particular provides a unique lighting to the game, and the sense of fluidity in the movement combined with what is a very solid frame-rate combines to make an FPS that feels very, very smooth indeed.
...

But then again, we really are seeing just a very small section of the game and it is clearly way too soon to be coming to any kind of definitive conclusion. The second, secret Crysis 2 demo level already hints at a game capable of so much more than we're seeing thus far in the Skyline map.

..... The notion of such an inordinately complex task being carried out in real-time on five-year-old console tech is hugely significant and is a massive technical accomplishment for the Frankfurt-based developer.

Real-time global illumination is one of the key innovations in CryEngine 3

Mitigating the effect in places is Crytek's excellent implementation of object and camera-based motion blur. This may well be a 30FPS shooter, but many people have noted that the overall feeling of the gameplay is similar to Call of Duty. While all the COD titles top out at 60 frames per second, Crysis 2 runs at half the frame-rate but thanks to some swift response from the controller combined with the motion blur effect, the feeling is of a game that stands out from the 30FPS pack.

Right now, it's safe to say that our findings are far from complete. The one level players get to sample is a multiplayer specific stage and doesn't seem to bear much relation to the campaign action we've seen of the game thus far. Crytek has also gone on the record to say that the demo is based on a build of Crysis 2 that is some weeks old, suggesting that a range of smaller changes will be made to the final game.

these are the things that stood out to me most from playing the demo and I never saw Pier 17 nor have we seen what SP has to offer

 

[L. Scofield]

I'm just saying that the compromises to almost every other technique used in the game probably leaves plenty processing power to do HDR with GI. I mean MLB: The Show does HDR (day/night cycles) with GI at 1920x1080/60fps.

Do you have any video of that? I'd like to see how dynamic their GI is.

That's all I'm saying. It's impressive, but can't any team of devs make 1 technique impressive by sacrificing most other techniques?

You keep saying this but, how many sacrifices did Crytek actually made? Slight subHD, temporal AA and draw distance for some objects/effects. Not only that, but how do you assume it was the realtime GI that caused those trade-offs?

On the topic of S3D, James endorses Crytech technique. If he was such a big fan, how come he didn't even choose that technique for Avatar's 3D? The technique Crytech uses is far from new.

Because it doesn't need to run in realtime LOL It's like asking why Pixar doesn't use MLAA

 

[Neb]

Do you have any video of that? I'd like to see how dynamic their GI is.

Saw this.

http://blog.us.playstation.com/2009/02/19/mlb-09-the-show-–-lighting-explained/

You keep saying this but, how many sacrifices did Crytek actually made? Slight subHD, temporal AA and draw distance for some objects/effects. Not only that, but how do you assume it was the realtime GI that caused those trade-offs?

Especially considering GI rendering time on 360...

 

[Ruskie]

Yap...1ms for real time GI is hardly trade of.Same goes for their SSAO,also 1 ms.