The Big Forza 2 Thread *

[one]

I was hoping One could clarify that point a little more since it was taken from a Japanese interiew.

It seems to me this sort of dynamic system would be useful for a Demo at GDC to keep the framerate smooth, but not something they would include in the final game.

It's already clear, but from this link
http://www.watch.impress.co.jp/game/docs/20070326/fm2.htm
+ Rendered in 720p. The AA level changes between off/2xAA/4xAA to maintain 60fps
+ Motion blur is generated but not apparent since the designer thinks it's not necessary when it's 60fps
+ Average poly count per scene is not disclosed. A car model is on par with PGR3 (80000-10000) or lower including interior models
+ Shader code is HLSL, it has specular maps, ambient occlusion by spherical harmonics, variant shadow map.
+ The first CPU core does gameplay (physics, AI, network, UI), the second core assists graphics rendering, the third one does audio. Secondary threads of each core do decompression.

Also damage, physics and AI are discussed, no car flipping/burning due to license is mentioned there. Damage is done by having 2 models for each car, an original one and a completely broken one. So the broken state is precomputed. How to transform into the broken state is however realtime, so it's like a morphing filter.

 

[scooby_dooby]

It's already clear, but from this link
http://www.watch.impress.co.jp/game/docs/20070326/fm2.htm
+ Rendered in 720p. The AA level changes between off/2xAA/4xAA to maintain 60fps

Does he state that dynamic AA system is for the final game? Are you sure he's not referring to the Demo?

It just seems very strange a developer would release a game that had dynamic AA from 0-4x, wouldn't that cause some strange visuals?

btw - thx for the translation

 

[22psi]

It's already clear, but from this link
http://www.watch.impress.co.jp/game/docs/20070326/fm2.htm
+ Rendered in 720p. The AA level changes between off/2xAA/4xAA to maintain 60fps
+ Motion blur is generated but not apparent since the designer thinks it's not necessary when it's 60fps
+ Average poly count per scene is not disclosed. A car model is on par with PGR3 (80000-10000) or lower including interior models
+ Shader code is HLSL, it has specular maps, ambient occlusion by spherical harmonics, variant shadow map.
+ The first CPU core does gameplay (physics, AI, network, UI), the second core assists graphics rendering, the third one does audio. Secondary threads of each core do decompression.

Also damage, physics and AI are discussed, no car flipping/burning due to license is mentioned there. Damage is done by having 2 models for each car, an original one and a completely broken one. So the broken state is precomputed. How to transform into the broken state is however realtime, so it's like a morphing filter.

Awesome recap, thank you very much!

 

[mrcorbo]

Damage is done by having 2 models for each car, an original one and a completely broken one. So the broken state is precomputed. How to transform into the broken state is however realtime, so it's like a morphing filter.

Well, that's interesting. So they are doing the damage in real-time. I like the concept they are using, then. Their implementation seems a little rough, though. Maybe not enough granularity? Or maybe they need to calculate and render a few (more?) intermediate states in between the moment of impact and the final resulting damage. The object being to not make it look like the damage just pops onto the car, which is how it looks currently IMO.

Thanks for the link and the translation.

 

[Tap In]

Thanks for the translation, ONE. :smile:

 

[JardeL]

+ The first CPU core does gameplay (physics, AI, network, UI), the second core assists graphics rendering, the third one does audio. Secondary threads of each core do decompression.

Why dont they go this way ; AI + Audio + Network + UI for one core and allocate whole one core for Physics ?.. Doesn't it a waste of processing power to use one core for Audio ?..

BTW, thanks for the translation ...

 

[Cheezdoodles]

Why dont they go this way ; AI + Audio + Network + UI for one core and allocate whole one core for Physics ?.. Doesn't it a waste of processing power to use one core for Audio ?..

BTW, thanks for the translation ...

Obviously, 1 core seems to be overkill for Forza 2's physics, which is impressive because it updates itself 360 times in a second.

 

[infinity4]

original xbox had Forza Motorsport physics engine refreshing @ 240 times per second

is 240 => 360 big improvement???

 

[Neb]

original xbox had Forza Motorsport physics engine refreshing @ 240 times per second

is 240 => 360 big improvement???

It's not only about the refreshrate for the physics engine but also how many objects/parts that has to have physics applied.

 

[Nesh]

original xbox had Forza Motorsport physics engine refreshing @ 240 times per second

is 240 => 360 big improvement???

Well perhaps in-game that could be translated into a much more noticable improvement than what this number suggests.

 

[scooby_dooby]

original xbox had Forza Motorsport physics engine refreshing @ 240 times per second

is 240 => 360 big improvement???

Source? Forza one was 60/s AFAIK

 

[Bigus Dickus]

Skid marks are your benchmark for a true racing simulation? And here I was thinking it was all about physics and handling...

Are you just trying to start an argument just to argue? Fine. Why don't you start by learning to pay a little closer attention to what you have read.

I didn't say "this is the one singular benchmark that separates a true sim and an arcade racer." I said it is the "kind of stuff" which clearly indicates there are many factors that go into differentiating an arcade from a sim.

And why do I think a pretty little texture decal is significant at all? Well it isn't because it looks pretty, that's for sure. It's because real racing leaves these marks, and they serve as markers on the road by which you gage your line. When the slate is completely clean it is more difficult to determine within inches if you are on the best line or not. When you can judge your position relative to these marks your judgment becomes much more precise.

Practicing on a track isn't just developing your memory of the track, but also laying down a reference by which you can judge your line. Thus, the marks have a relevance to your performance, not just how the game looks. And that's the "kind of stuff" that matters in a sim - the things that make your experience more like a real experience. You know, how the car interacts with the track? Like tire marks? And perhaps bumping a wall? Or do you think damaged cars are only to look pretty too?

 

[rusty]

what is all this skid mark talk about?


I'm sure I read Forza 2 has skid marks... and iirc they get laid down as you make them and are persistant across races. Or have I just dreamt that?

edit - wrong end of stick, I managed to find the post by big dickus that started the discussion and I see that the argument wasn't about forza lacking them.

 

[phat]

...

In real life, I don't leave much in the way of skid marks with each lap. The visible "racing line" on a track is usually not laid down because of skid marks, but gradual rubber and oil deposits over many, many laps. In short, I think you're overstating the "real life" relevance of skid marks, at least in the manner they are implemented in most racing games.

Anyway, in my books, a game that has good physics and handling but no skid marks or damage is much preferable to a game that has skid marks and damage but poor physics and handling. (That doesn't mean I think Forza has poor physics and handling--I don't know the game well enough to say much about these aspects of it.)

 

[sir doris]

In real life, I don't leave much in the way of skid marks with each lap. The visible "racing line" on a track is usually not laid down because of skid marks, but gradual rubber and oil deposits over many, many laps. In short, I think you're overstating the "real life" relevance of skid marks, at least in the manner they are implemented in most racing games.

This is true, in actual fact if there are skid marks they are most likely to be off the racing line as the result of the driver messing up (braking too late or getting off line where there is less grip).

 

[Qroach]

Source? Forza one was 60/s AFAIK

Forza 1 was 30 fps.

 

[Neb]

Forza 1 was 30 fps.

He means the physics engines refresh speed.

 

[scooby_dooby]

Practicing on a track isn't just developing your memory of the track, but also laying down a reference by which you can judge your line. Thus, the marks have a relevance to your performance, not just how the game looks. And that's the "kind of stuff" that matters in a sim - the things that make your experience more like a real experience. You know, how the car interacts with the track? Like tire marks? And perhaps bumping a wall? Or do you think damaged cars are only to look pretty too?

This is the truth. For the complainers, just play the game before arguing endlessly about stuff you have no clue about.

Anyways, early pit pass video today with 720p gameplay footage, bout time!

 

[rbushner]

Source? Forza one was 60/s AFAIK

Not his source, but here is some about the Physics of Forza 1

In Forza 1 we updated the physics anywhere between 180 and 360 times a second. That allowed us to have a very quick response and create a lot of torque in the car. Now, if you’re running the graphics at 60 frames per second you’re using up a lot of processing power so you have to put a dampener on those physics, and that’s when things start to feel a little sluggish. We didn’t want that, so we decided to keep the physics running high and pull back on the graphics.

http://www.oxm.co.uk/features/racing/under_the_hood_forza_motorsport_2_interview

At some point physics calculations are going to run into diminishing returns, Do we even know the competitor rates (GT, GTR, LFS)?

I'm pretty excited about this one! Where is our demo?!?!?

 

[Fu3lFr3nzy]

Where is our demo?!?!?

Probably 30th

Or dont tell me, it got delayed since the final game did /sarcasm