*renamed* Lighting and shadows in games

The problem is this distance is not a well defined quantity for more complex geometries. Even worse is when you have an object behind another, for example a smaller pole lying in the shadow the the bigger pole. You need a sharp shadow there, but the shadow map rendered from the light source can't even see the small pole. Then, as the light source moves, suddenly the small pole appears and you render a sharp shadow now.

This "next step" is extremely complicated and needs a completely different approach. The only realtime solution I know of without these artifacts is the penumbra wedge stuff from Assarsson and Akenine-Moller. It's stencil shadow based and really expensive, though. Not worth the cost over regular shadow maps, IMO, but damn are they pretty.

Here's a screenshot from my implementation of that algorithm:
http://forum.beyond3d.com/attachment.php?attachmentid=286&d=1187727472

The approximation they use for the aforementioned distance only works for simple cases. I did something very similar to this with VSM, as they are great for fast, variable width shadows after you strategically mipmap them or use summed area tables - much faster than 64 sample PCF.

 

That's all pretty interesting but computationally expensive I'm sure. And on the Enlighten Radiosity system, decent looks on 1/3 of an SPU is a very bold claim. I really want to see it myself work in a manner that's acceptable by today's standards.

 

It's weird that they express it in terms of SPU usage. That means it must be a per-vertex process.

The other thing I noticed is that as they moved the lights and objects around, they didn't have any occlusion effects on the wall. It's similar to what we saw in a demo by a forum member:
http://forum.beyond3d.com/showthread.php?t=44941&highlight=radiosity
This is also a per-vertex process executed by the CPU.

It's great that objects can scatter light around, but having them block this light too is even more important, IMO. Otherwise you still need static lightmaps to darken areas around objects near walls or in corners, and when you move them or change the light it'll look unrealistic, just like current lighting technology.

 

I think the MGS4 shadows is an art direction choice.
It makes it look sharp and a lot cleaner.

 

after playing the MGO beta, I dont think its a art direction choice. Like clearly a technical limitation with the low res shadow mapping and the KP team is just not as good as say Naughty Dog. Like making the shadows jaggy and pixelated at time make you game looks better and more artistic? I can see that GTA4 is using shdow dithering to imitate soft shadow edge, definitely not MGS4.

 

I honestly do think that KP really doesn't have a proper grasp of the PS3's hardware (RSX in particular). The game looks good however the environments themselves look alright. And if the game really does render in 576 lines I'll be very annoyed by such a fact. At least the game though is Metal Gear Solid, hence I can forgive all those things

But it is a bit unresting don't you think? A grade-A dev like KP having a hard time with the PS3?

 

the mgo and newer mgs4 shots are 1024X768 and temporal AA. Not true 720p but still better than 576 or halo 3 and COD4. Some can argue that its HD because some older HD plasma have the same native resolution Personally I don't think any developers are having trouble with RSX, but what to do with the SPU. See how Factor5 made one of the best looking gamecube game ever and it was a launch title, even PS3 is too much for them to handle LOL. I just hope when they release a substance for MGS4 they can do some upgrade to the graphic engine.

 

Isn't temporal AA an ATI thing therefore unlikely to be used in Metal Gear Solid.

 

already been used in DMC4 and Lost planet for PS3. Its not exactly the same thing as the ATI temporal, but its a similar techniq and they dont want to confuse ppl to call it something else.

 

Someone else mixing up Haze and MGO >_<

Reminds me of people lumping Lair and Heavenly Sword together

 

no mix up, some just refuse to believe and Haze isn't 1024X768.

http://forum.beyond3d.com/showpost.php?p=1161435&postcount=1059

 

I really had a problem with Halo 3 (or any other game for that matter) regarding its lack of shadows that reacted to the light sources around them. This issue was further worsened by the fact that shadows would frequently disappear from characters and objects once they became a certain distance away from the main character, or if too many objects were present on screen. Another issue was the fact that objects (bridges, buildings, etc that sort of stuff) in the game world did not seem to cast real-time shadows, but used that old pre-baked nonsense from last generation. Anyway, I thought Halo 3 was an overall good looking game, but just found the lack of any real-time shadow system to be a bit of a letdown graphically.

 

Being that Doom 3 preceedes Drakes Uncharted, MGS4, GTA4 and Heavenly Sword (and others) I am amazed that people believe that shadows that they expect will actually be seen in videogames on the PS3 given the current knowledge that game developers have on the platform.

I mean these are all different types of games, rendering in different ways, Doom 3/Quake 4 PC being the most spoiled in terms of not really having to render as many polygons.

Anyways not to change the subject but Nvidia did introduce SuperShadow gimmick with Geforce FX and SuperShadow II with Geforce 6800 as a way to supposedly help speed up shadow rendering in Doom 3 engine PC games and probably others as well, though it has alot of PR hype.

Game devs on consoles have alot to cover with the current gen consoles.

 

Its Ultrashadow actually , but yeah I think you can force it on in the Doom 3 .ini, which is not even on by default, and Splinter Cell Chaos Theory on PC also support it. Barely any performence gain though.

 

Yeah, sorry I had forgotten the proper term as I was working off of memory but yes it was Ultra Shadow.

In my experience with Doom 3 PC, I had a couple of rigs I tried it on.

One was a Intel PIII 1Ghz, 512MB RDRAM and ATI Fire GL 8800 128MB card (ATI R200 circa 2001) that although below the requirements allowed playable framerates with the last driver release in 2005 for that workstation card.

Then I added a cheap AMD Athlon XP 2000+ 1.67Ghz for a fully playable, no slide show dip game.

Finally I played it twice on a AMD Athlon XP 3200+ 2.333Mhz, 1.5GB dual channel and Nvidia GeForce 6800GT 370Mhz core by BFG, that card being Nv40 based featured Ultra Shadow II and the last time I replayed D3 was in 2006 after reformatting and installing last drivers then other than that the game played fine, the shadows I believe were as impressive as they could get.

As far as noticing anything the only thing I can say is that I slowed the game down enabling AA+AF over 8X so I settled on 4X including for the expasion game so the shadows looked impressive as they could but the fact that the game is all about tight corridors, low res textures and low polycounts that are noticable is unnavoidable.

Basically Doom 3's shadows are as basic these days as the rendering used in Virtua Fighter 1.

The only way I can see any improvement is if John Carmack/Id Software would reprogram the game using the RAGE based Id Tech 5, other than that if any other developer felt like making a Doom 3 rip off or clone (since the game was all about shadows) they can go on ahead and do so, I mean isn't EA already having a game that looks similar for the current gen consoles.

Anyways I feel lighting and shadows in the current gen console games will improve, but only after years and only if the game requires them in the art direction, after all Doom 3 was all about art.

 

The new Siren game has great lighting/shadow system, it seems like pretty much everything is able to cast shadows, very comprehensive. It also has this great lens flare effect that comes into effect not only when you look at light sources, but also any hotspots from your flashlight. Very cool.

http://www.vimeo.com/945895

 

I highly doubt it's a lack of knowledge on their part. They've already said they weren't able to achieve what they wanted due to hardware limitations. It seems to me it's more of the PS3 hardware not being what they needed. The PS3 is much easier to develop for than the PS2, but they were still able to achieve very impressive results with it last gen, even compared to the more powerful Xbox.

I think it's more of a case of the Cell hype meeting reality.

 

Actually they retracted that statement because it was a bad translation done by Kotaku. Kojima said he was very happy with the PS3 hardware and was surprise by the "disappointment" statement when the journalists asked him. It was reposted on Kotaku.

http://kotaku.com/5008653/americans-brag-ps3-no-letdown-kotaku-is-wrong

is it really that hard to believe that they are no as good as the best PS3 developer out there? Kojima even said western developers already surpass them especially on the technical side of things

http://kotaku.com/390267/kojima-western-devs-have-surpassed-us

 

It's a common sense optimization - you can crop volumes in software to get equivalent results, but if HW can do it for you, it's nothing to complain about.
It does nothing to alleviate the core bottlenecks of volume rendering though - so yea, the PR was a bit silly.

The look may be basic, but the process is on expensive side relative to alternative methods - and on current console hardware (last gen was a different story). Actually obscenely expensive if you want to improve the look to be competitive with alternative methods.