Thread by @delaneykingrox

Still getting a lot of questions about why #gamedev optimisation of meshes and textures matters if the game can handle it.

Over optimisation is when you spend a lot of time working on something that doesn't need optimising to do the job. Doing so wastes time you can put to better use.

So why shave a few polys off a model, cut down the texture size or strip out some bones when you can have so many these days?

Firstly, not all games are for top hardware, the latest drivers and the latest engines. What is a meagre savings on one platform is vital on another.

But even on a top of the line hardcore machine, shaving a little off an individual asset won't appear to make a difference. Right?

Well yeah, but you don't have one asset in a game project. You have hundreds, if not thousands.

A 2048x2048 texture usually takes up between four or five and a bit megabytes, depending on compression.

The next texture size up is 4086x4086. That is four times the size, because it is a double the length and the width.

Spread that out over hundreds if not thousands of textures and that single step up in texture size sounds a little more concerning.

Same with models. Reducing a few MB, or even a few KB of data from your meshes can multiply out quickly.

So as a video game artist, learning to optimise is still a huge part of your job, even with modern tools, hardware and engines, and I encourage all artists to instinctively and casually optimise as they go, even if it is a little here and a little there.

Even if you hit target fixed spec, players will enjoy that their downloads will be fast, loading screens short and DLC is ready to play moments after paying for it.

One other note on what I see in modern game art in this regard- often parts of a model will appear blocky and angular from too little polys, such as arms or shoulders. There is absolutely no reason that should ever happen these days, as usually sitting right next to said blocky-

-bit is some small detail modelled in stupid detail. If you are tight up to your trianglr budget, put those fuckers to work where they count.

I just cannot get my head around this weird mismatch. :D

*trianglr, like Grindr but for geometry.

*triangles, obvs.

So do yourself a favour. Aim to use just the resources you need to sell an effect. Clean up a little as you go, and for the love of satan, drink some water.

Nobody likes waiting for loading screens.