Guerrilla Games held a session at the Game Developers Conference titled "Taking Killzone Shadow Fall Image Quality into the Next Generation" during which Lead Engineer Michal Valient gave quite a few relevant numbers and provided more insight about the technology behind the PS4 launch game.
We learn that each level of Killzone: Shadow Fall is 10 to 100 times larger than a Killzone 3 level, while the largest map is 8 km (almost five miles) long.
Physically based rendering was used to achieve high quality with better performance. While initially the studio thought to use global illumination, in the end they went with Image Based Lighting, a technique that involves capturing an omni-directional picture of real world lighting and then projecting it on a half-sphere by light probes all over the level to simulate lighting instead of rendering it with traditional techniques.
In addition to that, there are are 200 different static lights per level, plus another 200 light maps.
The GPGPU (General-P0urpose computing on Graphics Processing Units) feature of the PS4 was also used to calculate reflections via ray-marching/ray-tracing.
Each character was made of about 40,000 triangles, with four times as many vertices and four times as large textures as in Killzone 3. Shaders were made of six to twelve textures per material.
There also are twice as many geometry instances (basically different elements of the graphics that can be reused, for instance different trees, wall sections and so forth) as in Killzone 3, and 10,000 to 25,000 of them were used per level. There's a total of 683,334 building blocks in the while game.
The presentation also included an in-depth view on the game's Temporal Reprojection feature used for multiplayer, but that explanation was pretty much the same that was given on Guerrilla Game's blog a while ago.
One thing is for sure: Killzone: Shadow Fall is a hell of a LEGO build.
[Guest reporting: Matt Zardoni]