Unreal 3 Engine
by dsfsd · in General Discussion · 05/16/2004 (5:14 pm) · 32 replies
Anyone watch the video from E3(On fileplanet)?
The idea of how they use cubemaps for shadows I found very cool.
I think they said it puts out 1,000,000 polygons on screen, built from 100,000,000 poly art sources. Puts a lot of emphasis on normal mapping. It must take some insane hardware to run even at 1,000,000 polygons, but I'm not tech genius so who knows.
I can't even imagine how much they are charging to use the engine.
The idea of how they use cubemaps for shadows I found very cool.
I think they said it puts out 1,000,000 polygons on screen, built from 100,000,000 poly art sources. Puts a lot of emphasis on normal mapping. It must take some insane hardware to run even at 1,000,000 polygons, but I'm not tech genius so who knows.
I can't even imagine how much they are charging to use the engine.
About the author
#3
Independent games aren't about the latest technology. Independent movies aren't about filming on 70mm with the hottest stars and huge sets.
Substance. Be different. Make something cool.
-J
05/16/2004 (6:20 pm)
There is no doubting it's incredible. Independent games aren't about the latest technology. Independent movies aren't about filming on 70mm with the hottest stars and huge sets.
Substance. Be different. Make something cool.
-J
#4
05/17/2004 (2:57 pm)
Hey, TSE will be pretty close to Unreal 3 tech-wise, we just don't have a few million dollars to throw at the art dept to make such a pretty demo ;)
#5
You should see the textures they use.. LARGE and LOTS of them :)
But it does look nice, even up close.
But hey, if you cant buy it, or afford it, its not an issue is it? I could wish for photorealistic realtime global illumination at 200k * 200k res all I wanted, but its NOT realistic to achieve, so why even think about it?
05/17/2004 (3:03 pm)
Tim: it doesnt run that fast either. Theyre targetting possibly 3-5 years from now for a machine that can actually RUN it well.You should see the textures they use.. LARGE and LOTS of them :)
But it does look nice, even up close.
But hey, if you cant buy it, or afford it, its not an issue is it? I could wish for photorealistic realtime global illumination at 200k * 200k res all I wanted, but its NOT realistic to achieve, so why even think about it?
#6
05/17/2004 (3:11 pm)
Because the harder the goal is to achieve, the harder you'll work to achieve the goal.
#7
I was disheartened by the extreme detail of the E3 demo for a couple hours... but think about it, Southpark used construction paper...It's hard being an indie... the most important things, GET YOUR GAME DONE AND MAKE IT FUN.
Glitz is nice and all... it's the substance that counts.
05/17/2004 (4:59 pm)
Unreal technology is and always has been as much about the content pipeline as the runtime. The artists that did the demo had the tech and the tools... and a nice fat paycheck as well. I was disheartened by the extreme detail of the E3 demo for a couple hours... but think about it, Southpark used construction paper...It's hard being an indie... the most important things, GET YOUR GAME DONE AND MAKE IT FUN.
Glitz is nice and all... it's the substance that counts.
#8
So how does this relate to Indie Developers? Well that age old attage that people keep trying to cram down your throat of 'concentrate on game play first and foremost, then on the icing' still applies regardless if you have a few dollars or a few million dollars. You simply cannot replace game play with eye candy and expect it to make for a better game.
That's just my two cents on the subject, like it or hate it.
Logan
05/17/2004 (5:33 pm)
Indeed Josh. I was watching the Unreal 3 demo that was running at the nVidia booth from a distance and I had to honestly say that being able to take a 'few steps back' I didn't see a single thing in the preview that made the game better. Sure it had glitzy graphics and crud like that, but from what I saw it never did a single thing to make it a better game play experiance. Plus since the game is primarily about twitch gameplay and fast paced action you honestly don't have a chance to enjoy all the work that was done. So it really seemed to me that all this work that they had done was a moot point because in the end it didn't result in anything... well except for marketing fodder.So how does this relate to Indie Developers? Well that age old attage that people keep trying to cram down your throat of 'concentrate on game play first and foremost, then on the icing' still applies regardless if you have a few dollars or a few million dollars. You simply cannot replace game play with eye candy and expect it to make for a better game.
That's just my two cents on the subject, like it or hate it.
Logan
#9
So true, I watched it a second time, and I think I'm more impressed with the artwork then the renderings programming technically type stuff. I was like, "Hey I saw that same feature in the TSE bootleg video".
The only thing that struck me was the cubemapped shadows idea where they use two cubemaps (One blurry & one sharp) and fade between the two, I thought that was pretty cool, I had never even thought of using cubemaps like that and it produces a very realistic effect.
Pretty artwork though! :)
05/17/2004 (5:38 pm)
Quote:But hey, if you cant buy it, or afford it, its not an issue is it? I could wish for photorealistic realtime global illumination at 200k * 200k res all I wanted, but its NOT realistic to achieve, so why even think about it?
So true, I watched it a second time, and I think I'm more impressed with the artwork then the renderings programming technically type stuff. I was like, "Hey I saw that same feature in the TSE bootleg video".
The only thing that struck me was the cubemapped shadows idea where they use two cubemaps (One blurry & one sharp) and fade between the two, I thought that was pretty cool, I had never even thought of using cubemaps like that and it produces a very realistic effect.
Pretty artwork though! :)
#10
But I agree. The artwork in Unreal is just incredible, and it totally sells the demo. Would that GG had the budget for that sort of thing. :)
05/17/2004 (10:15 pm)
The problem with the cubemap trick is that you really only get to use it once on modern hardware - any more and the overhead of rendering the scene from 7 different perspectives (or 13 or 19) kills performance, even on really high end hardware like the latest-gen ATI and nVidia hardware.But I agree. The artwork in Unreal is just incredible, and it totally sells the demo. Would that GG had the budget for that sort of thing. :)
#11
If Im lucky, and he gets his name in the credits, never know, might be able to get a little, "look at this pretty stuff" demo done.
I'll see, first I need TSE. *wink*
05/17/2004 (11:26 pm)
You know Ben, I could get someone to do one or two models...If Im lucky, and he gets his name in the credits, never know, might be able to get a little, "look at this pretty stuff" demo done.
I'll see, first I need TSE. *wink*
#12
05/17/2004 (11:53 pm)
Anyone have a clue on how they made those beutiful scenes? BSP techniques? 
#13
Notice that there are no characters running about. Having the characters at the same level of detail on todays hardware will give you about 6 or seven really detailsed characters before the scene grinds to an unceremonious halt :o}
still does look cool though......
05/18/2004 (1:54 am)
:o) Those beautiful scenes are normal maps that's kind of it. The architecture is still low polly (go read their u3 engine blurb about it). They use pixel shaders and hy dynamic lighting i.e. 64bit textures and some post scene processing.Notice that there are no characters running about. Having the characters at the same level of detail on todays hardware will give you about 6 or seven really detailsed characters before the scene grinds to an unceremonious halt :o}
still does look cool though......
#14
05/18/2004 (4:07 am)
Yeah, I mean.. look at the statues outside! :) 
#15
You need to be rendering ~1000 polygons per call, else the CPU limits the GPU. When TSE gets here (and assuming it hasn't replaced its BSP renderer) I gonna look at replacing it. I decided to wait until TSE is here, as its going to be a big job if I have to do it...
And its worth noting the low-poly in Unreal terms is 100K+ polygons in the environment. THe actual environments before normal map generation are much higher (multi-million).
05/18/2004 (4:33 am)
No modern GPU based renderer use BSP techniques, its completely the wrong approach. Any thing that subdivides space on a per polygon level is the wrong approach for GPU work. Unreal 2 (the last engine codenamed Warfare) was when they really dropped BSP support (it was advised not to use it anymore for performance reasons), in Unreal 2 terms hardware brush support was the only way to achieve good performance (which uses an object level octree).You need to be rendering ~1000 polygons per call, else the CPU limits the GPU. When TSE gets here (and assuming it hasn't replaced its BSP renderer) I gonna look at replacing it. I decided to wait until TSE is here, as its going to be a big job if I have to do it...
And its worth noting the low-poly in Unreal terms is 100K+ polygons in the environment. THe actual environments before normal map generation are much higher (multi-million).
#16
05/18/2004 (10:30 am)
Torque doesn't use a BSP based renderer, only BSP based collisions. Everything is blasted from a vertex buff for interiors (if your card can do it).
#17
I still want a 6800 Ultra though.
05/18/2004 (10:41 am)
I read in an interview that this engine's targetted towards the average PC in 2006, so.... yeah. ;)I still want a 6800 Ultra though.
#18
Unreal 3 is WAY out of my timeframe/price range. So other than a "good for them", I can't say much... besides, the real heroes are id Software. They release their damned code for fools like me to hack on. Rotten Epic!
05/18/2004 (10:44 am)
Our game is based on QuakeII from 1997, with a greatly improved rendering system and toolchest. I can attest to BSP based VISIBILITY still being very, very handy. For graphics, networking, pathfinding, etc. When I have time, I do need to take the splitting code out... and optimize the vertex buffering. Unreal 3 is WAY out of my timeframe/price range. So other than a "good for them", I can't say much... besides, the real heroes are id Software. They release their damned code for fools like me to hack on. Rotten Epic!
#19
Of course if you hurry up and let us buy TSE I can get writing some cool shaders in my 'spare' time :-) First up is to convert the articles I wrote for the ShaderX books. Is there going to be a torque shader resource section? Would be really good to have a dedicated resource section where we can add various shaders for other to use. Given the excellent community here, I'd imagine there would quickly develop an extensive shader library.
On the BSP issue, even for non rendering uses in many cases its slower than lots of other compareable algorithms. Of course as its already there no point replacing it but the cache misses and pointer walking really hurt on today's CPU. Just one of those algorithm that hardware has taken a dislike to.
05/18/2004 (1:25 pm)
Glad to hear Torque doesn't use a BSP renderer :-). Heard a few people refer to DIF being BSP based, so assumed otherwise. Haven't actaully looked at the rendering side of Torque much yet (I brought torque to make games not renderers (thats a large part of the day job)).Of course if you hurry up and let us buy TSE I can get writing some cool shaders in my 'spare' time :-) First up is to convert the articles I wrote for the ShaderX books. Is there going to be a torque shader resource section? Would be really good to have a dedicated resource section where we can add various shaders for other to use. Given the excellent community here, I'd imagine there would quickly develop an extensive shader library.
On the BSP issue, even for non rendering uses in many cases its slower than lots of other compareable algorithms. Of course as its already there no point replacing it but the cache misses and pointer walking really hurt on today's CPU. Just one of those algorithm that hardware has taken a dislike to.
#20
But really --- except for maybe an introductory "attract mode" (remember the castle in the first Unreal?), who's really going to be looking at that? Yes, I do occasionally go through Unreal Tournament solo, without any bots, just exploring the level and enjoying the eye-candy. But when you are playing the game, it all fades into the background - as it should.
So is it WORTH it to hire an army of level designers to spend weeks at a time on a single room / courtyard? Are they going to spend years modeling the Sistine Chapel? Can content creation possibly keep up, even with the most advanced tools? If so, how?
05/19/2004 (9:46 am)
You know, I'm not so concerned about the hardware... I think it's AWESOME that top-o'-the-line computers two years from now will be able to render those kinds of scenes smoothly.But really --- except for maybe an introductory "attract mode" (remember the castle in the first Unreal?), who's really going to be looking at that? Yes, I do occasionally go through Unreal Tournament solo, without any bots, just exploring the level and enjoying the eye-candy. But when you are playing the game, it all fades into the background - as it should.
So is it WORTH it to hire an army of level designers to spend weeks at a time on a single room / courtyard? Are they going to spend years modeling the Sistine Chapel? Can content creation possibly keep up, even with the most advanced tools? If so, how?
Torque Owner Mike Stoddart