I am not a video game person, before recently I have not played a video game in over 10 years. Since I have been getting more into Eevee I wanted to see what the video game engines are capable off. So my gf recently bought me a PS4 with Spiderman (10/10 my rating) and Kingdom Hearts 3 (6.5/10 my rating). After playing through these games I have some questions.
What are the most computational heavy parts of a game from greatest to least?
Simulations
Polycount
Textures
etc…
The reason for this post is. I remember on PS1 when the games had a cut scene (FFVII) it would be a nicely rendered scene that looked awesome. Now games use the same in game footage and it looks awful (most of the time). Why don’t games render out beautiful animations for the cut scene’s instead of using the lower quality in game footage like they used to?
Offhand, I might suggest that the decision not to use “cut scenes” likely has to do with the tremendous advances of “raw computing power” in the last (10) years: designers don’t have to do this sort of thing anymore. Real-time graphics capability is now "so damned good" (witness: Eevee itself) that they feel that they can get away with rendering the scenes directly – and their customers don’t seem to be complaining. “Moore’s Law” shows no sign of stopping or even slowing down.
That being said, I’m not sure that it’s actually possible to give a meaningful answer to your question: every game (that actually “stresses modern-day hardware at all” … many don’t) is of course going to stress the hardware in different ways.
As far as “what video game engines are capable of,” I strongly suggest that you spend some time now looking at armory3D, which in my humble opinion probably represents the most-significant latest advance, and which is fully integrated into(!) Blender as a plug-in. “Enjoy! And, prepare to be amazed.”
Well, just about EVERYTHING effects the performance of a video game. It’s really about keeping things balanced, sure heavy graphics can really cause a hurdle in the performance side of things, but poorly programmed systems and simulations also cause great effect.
Games like Minecraft have really simple graphics, yet take quite a bit of processing power do to the large amount of data the game generates and has to keep in memory.
Pre-rendered scenes are definitely not gone, but they’re kept for the most important cutscenes in games. One reason they don’t pre-render all cutscenes is because some games (most games these days) have customizability options. It would break the immersion if your character is wearing something in one scene and not in another. Having the ability to change what a cutscene looks like based on certain choices the player makes is pretty important these days.
But the reason non rendered scenes often look bad is because developers have to make sacrifices with the quality. Games are very hard to make and they have to decide where to put their time, energy and resources into it.
I’m no game developer, but I’ve done my own studying on game dev for a couple of years now. (cause I do want to make games in my own time) So those are just my thoughts.
This trend of using in-game engine for cutscenes was helped in large part by Metal Gear Solid on Playstation. With so many cutscenes the game could not use prerendered fmv because it would take too much cd-rom space (the game was 2 cd-roms already). So part of it is to cut down storage requirement since fmv hogs lots of storage data.
Plus, "we have so freakin’ much more raw computer horsepower" available to us now, even in our phones, than we did when e.g. the PlayStation was first architected.
“Horsepower has its privileges.” …
We can do “cut scenes” using real-time rendering now, along with the rest of the game.
more like “parts where we can’t cheat the processor and add a premade version of the work into memory”.
Simulations (physics, basically) are always a heavy load, graphical representation of it even more so.
thing is: you can cheat these three in many games. Physics algorithms are super simplified versions of more realistic calculations (that’s why sometimes objects feel “weightless” in a game) and many factors are summarily skipped or adapted (air friction adds very little in almost any game, and even when needed, it’s normally replaced by a specific algorithm to the specific object that has gameplay changes based on it).
When you don’t cheat the math, you replace it instead. Many animations of “natural” things can be keyframed and many objects that should react to environment stimuli can be rigged.
Also, many lighting and particle effects completely ignore anything other than themselves. (not long ago things like fire and water movement completely ignored walls. you’d see waves propagating from your steps and going through a rock wall)
As such, baking effects and rigging objects to fake what should be a simulation creates gigantic games in terms of storage, but very light on the processor. just point to memory and follow the baked map or animation ^^
A second area that is harder to cheat is AI. sometimes by lack of optimization, sometimes because it evaluates too many variables at once. Grand strategy games suffer heavily with it
