Impressive games for their time.

Some friends of mine ported BGDA from PS2->GC. Technically, the game a extremely impressive. A few fun facts:

1) The level content was tightly streamed. In debug mode you could pull the camera out a bit and see the geo pop in an out just outside of the standard camera view. Because of that...
2) The levels were uniquely textured. Pre-megatexture on PS2. The game was textured using 3DSMAX's standard material. The artists could do whatever they wanted in MAX and it would bake down to a single texture layer. Textures were streamed. So, no repetition. No tiling necessary.
3) They had a dungeon editor in MAX where you could just draw the layout and it would pick geo tiles with the right walls and corners.
4) The polygon count was crazy high. The barmaid had some redic poly count. She had a full body modeled under her cloths.
5) The whole game was coded in less than 10 CPP files. Each one was 10s of thousands of lines long.
6) Everything in the game was hard-coded in C. The menus, the dialogs. If's and char[]s. All hard coded.
7) The water was a bitch to get running well on the GameCube. The same guys went on to make The Conduit.

 

Fun fact about The Conduit:

It was the subject of the Has Anyone Really Been Far Even as Decided to Use Even Go Want to do Look More Like? meme.

 

Truly impressive stuff
Out of curiosity was Baldur's Gate one of the games rumored to have used some sort of bump mapping on the PS2?

 

GTA3, MGS2, Half-life, FarCry, Crysis, Gears1 and The Division.

 

Shows ps2 had its advantages over gc. DF mentioned the ps2 version looked best, assume the water and super sampling. Happen to know how the xbox version fared?

 

There are some people on this board who thought it might have been. But, I don't recall anything like that being mentioned. The normal mapping in MAX was baked down to a static texture. The water had crazy high poly counts with per-vertex normals.

Near the end of the PS2 lifecycle, the Conduit guys and I figured out how to get "per-pixel lighting" (technically) on the PS2. Basically, we had the VU compute lighting for a 256-normal palette each frame. The object would then use a palettized lightmap that updated in real time. Object space normals only. No skinning. Just amazed that it worked at all.

We also figured out how to get cubemaps (mostly) working on the PS2. Any guesses how?

Unfortunately, neither of those techniques ever shipped in a game

 

Did you run into trouble with the xbox port? Was the port not able to retain its ssaa and water like on ps2?

 

A couple of games I remember being blown away by at the time, thinking: "Wow! so this is what games will look like once rest of the devs learn how to code this beast of a console"
PSOne:
Disruptor - Textures were so much better than what I'd seen on PS1,
WipEout - First 3D polygon console game I've owned and played, I'd never played anything like that before.
Toshinden - Thought it looked much better than Tekken, didn't play as well though.

PS2:
Baldur's Gate Dark Alliance - Lighting, effects and modelling were amazing
Jak and Daxter - "Pixar quality" animation and characters. Huge streaming, seamless world were first for me.

...after that, haven't been so easily impressed

Edit: Oh, and MGS2 on PS2, especially the physics got me excited.

 

Killzone 2 in 2009, truly one of a kind in the industry at the time in terms of rendering, post processing, particles, AI and heavy gun play that really made it ahead of its time.

 

Resident Evil Remake on the Cube had some amazing graphics utilizing the static camera angles to the full extent.

 

The Conduit did have water surface simulation in a few parts of the game. Check 6:40

 

How did GT3 and GT4 achieve the reflections? I thought they were using cubemaps. The cars reflected properly the environment at any position.

 

snowblind did the Xbox port in house

 

Any reason for that to be an inferior port?

 

That is definitely the same technique.

 

Probably some variation of https://docs.microsoft.com/en-us/windows/desktop/direct3d9/spherical-environment-mapping

 

So you computed the normals of a 256 color palette? What kind of granularity can you even get with that? Isn't that like 16 different angles in two axis?

 

Start with a full-detail normal map and palettize that like a regular rgb texture. Instead of a uniform sampling of directions, you get the 256 most important ones. Do the lighting on those and replace the palette with the lit results.

 

It liberately employed specular maps at least (no idea whether they've used seperate maps, but you get the idea). That was actually pretty rare on the PS2 as well. For most PS2 games, effects such as metallic sheen was usually achieved via painted texture maps.
Cannot commend them enough for their heavy investment into the area of breast physics also God, I loved that game. Sequel was pretty good as well and so were the Champions of Norrath games. I miss Diablo clones with bigger skill components. You know, games in which getting damaged was actually down to proper collision detection and not dice rolls.

I was also really impressed by other PS2 games such as Primal and Ghosthunter. Those haven't aged quite as well unfortunately. Gameplay is a bit wonky in both games, and Jen's non-animated hands in Primal certainly stick out.