Unreal Engine 3.0 at GDC 2005

OpenGL OpenGL: Advanced Coding
1

I’m sorry that this is off-topic, but most of you graphics freaks will be inspired by this. :smiley:

http://www.watch.impress.co.jp/game/docs/20050314/epic1.zip (27.8MB)
http://www.watch.impress.co.jp/game/docs/20050314/epic2.zip (32.0MB)
http://www.watch.impress.co.jp/game/docs/20050314/epic3.zip (28.9MB)

These three videos were taken at Tim Sweeney’s lecture
named “Game Technology and Content Creation for the Next Generation”, GDC 2005.

Most of the technologies such as Virtual Displacement Mapping were already demonstrated in previous year,
but you can see rigid body simulation using Novodex and the new HDR Cube Light Mapping…etc.

2

Liked the bright ball in the fridge.
Laughed when one of the socks dangling from the ceiling flew straight through the wall, while the guy doing the demo quickly turned and ran away, trying to pretend it hadn’t happened.
Disappointed that there wasn’t more outdoor environments.
I’m not as impressed as I was when I saw the HL2 videos, but obviously the tech here is better (soft shadows), just not as much of a leap as HL2. The stuff in the HL2 videos was stuff that up until then we had only seen separately in small teapot demos, but never seen it all brought together in a game environment…I suppose that’s what gave it it’s impact. How long are we going to have to wait before we get that impact effect again, I wonder?

3

HL 2 never really impressed me that much. I hardly see it as a leap. It’s ok but that it’s not that great. My impressions until close to its release was “that looks kind of crappy.” There are a few things in HL 2 that look pretty good but overall it’s not that great. The game itself is good, I’ve already played it through two times and just started my third time through. :smiley: The Doom 3 engine was a leap from things at that time. After seeing quite a bit of Doom 3 by the time I saw HL 2, it’s no wonder HL 2 seemed like yesterday’s tech to me.

UE3 however has impressed me a whole lot. The lighting and shadows are done a whole lot better than any other engine. The large detailed city scene made city 17 look like wolf3d. The water looks pretty neat from what I can tell from my ****ty low res video. The video showing the physics where all those gears and balls are rolling around causing all kinds of things to happen was pretty cool. I liked the rocket car. :slight_smile: Oh and yeah when that “sock” went through the wall…haha, good stuff. That’s probably the first thing they fixed when GDC was over. :smiley:

Overall UE3 kicks the snot out of Source and Doom 3 engine in all areas. Valve really needs to get to work with upgrading Source b/c really Source is nothing but old tech from HL 1 days with a few shaders and physics thrown in here and there with really hacky hardly working dynamic lights (see flashlight). Like Carmack said, it’s the best last generation engine (which means the gen before doom 3). Oh and don’t get me started on the load times (and frequency of loading) in HL 2. Oh my god…

-SirKnight

4

I’ll be excited to see how many of those nasty creatures the ue3 will be able to simultaneously throw at the player. Doom3 apparently is limited to a couple of enemies at a time (I guess due to heavy use of normalmapping and expensive shadow-calculations) - whereas Source seem to easily handle 10-20 enemies.

Both d3 and hl2 had quite nasty loadtimes, but if the ue3-hype holds true, this is being managed by loads of clever streaming. I guess another important point is the tools made available for each engine - oh, and the fact that ue3 is supposedly cross-platform.

\hornet

5

About simultaneous number of creatures in Doom 3, I guess it is mostly limited because of the pace and atmosphere (and CPU power for shadows and AI, not GPU).
Playing Last Man Standing mod showed tens of monsters in some maps, without problems. (amd 64 3200+ here)

6

I think many people have an annoying tendency to confuse content with code.

The reason that the Unreal Engine 3 demo looks so good is not because Tim Sweeney is some kind of code wizard (even though he’s a very skilled software architect, certainly much more so than John Carmack from a pure computer science point of view) but because epics artists have created some kickass models and textures.

The same thing hold’s true for Doom 3 as well. Since Quake 1 id’s games has become less and less technically impressive, even though of course they look better and better.

The thing that’s most impressive with the new unreal technology is their content creation tools.

/A.B.

7

is not because Tim Sweeney is some kind of code wizard
oh i differ graphics wise both he and carmack are (certainly in the top10 best games programmeurs) i base this on available evidence and not on personal opinion.

8

Originally posted by zed:
[quote]is not because Tim Sweeney is some kind of code wizard
oh i differ graphics wise both he and carmack are (certainly in the top10 best games programmeurs) i base this on available evidence and not on personal opinion.[/QUOTE]I beg to differ, this might have been true about 10 years ago when there weren’t anything that could compete with the quake 1 engine.

Today however is a different story, here’s a short list of games with 3D engines with similar capabilities to the Doom 3 engine:

  • FarCry
  • Riddick
  • PainKiller
  • Halflife2
  • …

and a lot more are in the works (just counting fps:es).

It’s not hard to create a 3D engine on a modern PC, the hard thing is generating content. There are a number of Quake III level viewers out there written by amateurs which proves that it doesn’t take some kind of genius. I guess we’ll soon see a lot of Doom 3 viewers as well.

Exactly what in Doom 3 or Unreal is it that you find so impressive that can be attributed to Carmack or Sweeney?

If you want an example of really impressive engine programming i suggest you take a look at Jak & Dexter or Gran Turismo 4 for the PS2, there you have a couple of guys who plays in a different league.

/A.B.

9

Originally posted by brinck:
[b]Today however is a different story, here’s a short list of games with 3D engines with similar capabilities to the Doom 3 engine:

  • FarCry
  • Riddick
  • PainKiller
  • Halflife2
  • …
    [/b]
    Speaking only about the 3D rendering part : all the engines you listed do not include real time lighting/shadowing. Even if D3’s one seems simple (hard edge shadows, no soft shadows), it’s more advanced than the “previous gen” engines you listed (static lighting aka improved lightmaps, simple shadows). Just a basic engine with a few pixel shader introduced (water).

The HF2 case may be even a lot worse regarding shadows : casting shadows in wrong direction (toward lights :eek: ), casting shadows THROUGH walls ( :eek: again)…

To make it short : old style engine with really nice graphics.

10

Originally posted by brinck:
I guess we’ll soon see a lot of Doom 3 viewers as well.
Already the case since a few weeks after the “Alpha” leakage.

11

this all concerns the grpahics engines (not ai,physics,modding etc)

It’s not hard to create a 3D engine on a modern PC, the hard thing is generating content. There are a number of Quake III level viewers out there written by amateurs which proves that it doesn’t take some kind of genius. I guess we’ll soon see a lot of Doom 3 viewers as well.
as u know the difference between a demo (shortcuts/hacks golore) and a polished game is immense

Exactly what in Doom 3 or Unreal is it that you find so impressive that can be attributed to Carmack or Sweeney?
i believe both do the main programming of the graphics engine for their companies, at the time of release these games (maybe not the last unreal game) have been at the cutting edge graphics wise, if making these engines is so easy why arent we seeing heaps of similar quality engines at the same time? as tfpsly mentioned with FarCry Riddick PainKiller Halflife2 (perhaps roddick excepted) i believe these games could be made from a quake3/UT mod and look exactly the same (perhaps theyll play a bit slower, since theyre not as specalised) quake3/UT is 5 years old.

personally i dont idolise carmack or sweeney, but i can see they are in the top echelon (perhaps they get more praise than they should but all industries are like that). what pisses me off is to see them getting called average programmeurs when anyone with half a brain can see its noty true

(wrong answers)
*carmack cause of his fame can do whatever he wants

12

I am pretty uninpressed by recent developments of the game industry.
When FPs where not there, having a Q3 degree of shading (limited vertex, fragment and blending) required heavy wizardry. Just now, I look at say, the shadows and I say: “Oh, it’s just a PSM” or “I can also do that by doing THIS or THAT in a fragment program”.
Most advancement in recent game development is basically natural evolution. Put 100k tri on a model while few time ago you had to have 5k tri scenes…

13

Sure, it’s natural evolution of the same old thing.
Even Carmack said it in an old .plan that Quake3 engine is pretty good and his game engine doesn’t require a rewrite for the next generation.

At this point in time, I’d be more impressed seeing human AI (if I can chat with it) and more natural looking humans.

I’m sure eventually, there will be another few booms in 3D graphics.

14

UEngine 3.0 videos are truely amazing

“amazing things willbe here soon”
Hey, he was right…

15

Originally posted by V-man:
[b]At this point in time, I’d be more impressed seeing human AI (if I can chat with it) and more natural looking humans.

I’m sure eventually, there will be another few booms in 3D graphics.[/b]
I agree with you.
Also wanted to write down something which can be misunderstood from my previous post.
While the increment in complexity is actually, at least for me, less “interesting” than previous generations, I don’t mean to say it’s easy to get this. Those programmers are smart, I don’t want to discuss their skills.

16

Everyone here who has been involved in programming for some time has one time or another made something great and also made serious mistakes. So you can’t say so and so is better than someone else and vice versa. Such is life.

Now, on the topic of rendering. Right now you can choose to go in one of two ways. You either use old gen. tech ie. lightmapping and do large outdoor/indoor scenes and put all the cycles in more detailed models and AI, etc. or you can have nifty lighting/shadowing but small areas. You can also mix them up like far cry did with some hacks but you won’t have unified lighting/shadowing then. So you can’t blame companies for choosing one way or another. Each one chooses based on what style of game they want to do ie. far cry vs. doom3. What really matters here is not gfx, algorithms, AI, physics, etc. but “how fun the game was and how much it sold”. That’s basically it and if you think it’s about something else then you don’t understand the business of game making. Sure you can moan about how you can do some fancy gfx algo but you are no longer playing at the same level with these companies because their goals are so different from yours. It’s not that they’re not capable but they’re restricted by something else that you probably won’t experience.

17

So you can’t say so and so is better than someone else and vice versa. Such is life.
can u say that El Guerrouj is a better middle distance runner than me?

Now, on the topic of rendering. Right now you can choose to go in one of two ways. You either use old gen. tech ie. lightmapping and do large outdoor/indoor scenes and put all the cycles in more detailed models and AI, etc. or you can have nifty lighting/shadowing but small areas.
my engine deals with unfied light and hi polygon counts (since the year couple of years, polygon count has practically little framerate impact (unless u do someit crazy like choose stencil shadows))

“how fun the game was and how much it sold”
whilst i agree with u here, as u know this forum deals with openG(RAPHICS)L thus discussions about unreal3’s graphics quality are more relevant than unreals gameplay

18

There is no doubt UE3 is an excellent and impressive engine. But categorizing Sweeney, as much as i respect him, as a programming god is IMO going a step too far. He’s been leading UE3 development, right, but do you really believe he would have achieved this alone ? Let’s not forget all the programmers working with him which you probably never heard of. Let’s not forget all the artists that made these incredible scenes, because with programmer’s art, i’m sure UE3’s engine would immediately be less impressive. Overall i’d say these videos are 75% high quality art and 25% programming features.

Y.

19

This is likely really OT, but I hope moderators cut me some slack based on previous posts in this thread.

Originally posted by Ysaneya:
There is no doubt UE3 is an excellent and impressive engine. But categorizing Sweeney, as much as i respect him, as a programming god is IMO going a step too far.
I agree 100%. As far as “programming $DEITY” goes, at least in the gaming arena, I see no more people than the number of fingers on one hand. Carmack is one of them, for obvious reasons. I also think some people from Looking Glass, and maybe even Bethesda, are worth mentioning (but as I don’t know their names, I leave it at this). Going back, I could mention DMA, but that would put us back to the 2D era (but just for kicks, while in 2D mode, think “parallax scroll”).

They were the ones creating the “leaps” I think worth mentioning, and now we are (after “standing on the shoulders of giants”) at a level where just about anyone (obvious simplification) can create a 3D engine for just about any 3D game.

This is obviously an overly broad simplification, but once an effect has been discovered and implemented, it’s mostly just about putting it in context - something level designers and artists are much better suited to do. We have become more of database designers, and data shufflers. Has anyone ever considered a creator of an especially effecient database a “programming $DEITY”? :slight_smile:

What do I think is impressive about the UT3 engine then? To be honest, I don’t think anything by itself is to gaze upon. I think what is worthy attention is what the engine can do in combination with the tools. If the design tools weren’t there, or sucked, the engine might be able to produce real-time physics at the atom level, and visuals making a close-up of a tree make you think you looked at a photo - but without the tools to edit that media it would quite possibly be an engine able to render a Q1 level, but at 40-500 times more CPU, memory and gfx card requirements…

I mentioned this last part, since I myself know how fun it can be going overboard with effects in research and proof-of-concept stuff, that would force all potential users to also buy the “latest and greatest” gfx cards for $500 when if the programmer (me) had stopped and given some thought to what they did, the game would have run perfectly fine with less visual bells on a gfx card much older (which, I’m told, most user except the most hard-core gamers have, which translates to reaching a much large audience).

Personally, on one of my computers at home, I just upgraded to an ATI 9250 (I really need to get in touch with ATI re. their drivers…), and use so “outdated” hardware for the sole reason to be forced to use code-paths allowing lower-end hardware to work decent too (not to mention, many potential users would consider a 92xx card high-end…).

20

Zed, I was mainly talking about programmers using their brains not their physical attributes. Unless you have a brain defect your brain has the same amount of chance at being just as good as Tim’s or JC’s or mine are :slight_smile: People don’t realize this but JC has been making games for very long time before Keen or Wolf came about they made roleplaying games and before that wrote business sw so they had long time to improve their brains by thinking about things and practicing it. That is the real secret to their success and you can reach it as well if you excersize your brain. You can’t say the same thing about your physical attributes which are goverened by heredity and genetics. If you could change your genetics as easily as you can your brain or mind then you can be faster or better than that fellow you mentioned. Plus, let’s not forget that everyone copies from each other and that is how things get better. There are times when you can invent something on your own but it will always be inferior to what a group of people can do if they think together and I believe this is how we are suppose to get better. We can also leverage our weaknesses with other’s strength.

About gfx, I was mainly speaking about global vs. local illumination so it isn’t possible to unify them today. You can fake global with local and you need a 6800 or faster card for that. So for slower cards, stencil shadows is the only sound way today. Better yet, static lighting and shadowing ie. lightmaps which cards can do fast but lightmaps take up lot of memory so there is a tradeoff there. Sorry if I wasn’t clear what I meant in my post.