|
|
|
|
|
by SeanBoocock
3420 days ago
|
|
This is increasingly untrue. Modern consoles like the Xbox One and PS4 guarantee availability of six hardware threads with provisional access to a seventh. This combined with things like VR putting a premium on high FPS has motivated more parallel architectures. Most game engines architectures have supported limited parallelism in the form of dedicated game simulation and render threads, as well as often having asynchronous processing of audio, network, and IO. More modern engines have task based architectures that allow for parallelism of the game simulation. This is often implemented as a fork-join model around sync points spread throughout the simulation update: pre-physics and post-physics update periods for instance. While it is true that some game logic relies on knowledge of global state and complicated dependencies between entities in the simulation, games can and do find meaningful reductions in CPU wall time. |
|
I wish this was increasingly untrue.
Running For Honor just now a new title built for consoles one CPU core is at 100% the rest are <10%
;)