Frame rate

News & General Discussion
1

I am not entirely sure where to post this, or even how to title this.

I was looking up poly count and was seeing a wide range of answers, obviously because all games are different.

I wanted to know what the limit for poly count was on my pc before frame rate was heavily impacted.

I modeled a handrail and duplicated it almost 1700 times, added materials and textures, but the models weren’t unwrapped and this is the result.

I have to say it was way higher than i expected. This is a quick and dirty test that could probably be improved.

My pc specs are Core I-3 6100, 8 gb ram, EVGA Geforce GT 740.

I wanted reassurance that i would be ok on models and i think this tells me not to worry about poly count.

If anyone has suggestions to improve the test, or tell me how bad my procedure was please do. I’m always looking for assistance.

2

A game has a lot more going on that just a bunch of vertices sitting still in space.

It’s a complicated subject, and this is only a simple test. It’s good you are thinking about it and testing things out, but I don’t think your results here will give you much predictive ability about anything.

There is a lot of reading to be done both here or at places like the polycount wiki, but I wouldn’t worry too much about this if you are just getting into modeling or game design. It’s kind of like, you don’t need to know how a carburetor works to change the tires on your car, right? But eventually, as you get better and learn more about the basics, you’ll get to the more complicated stuff and then have a base of knowledge to apply.

If you haven’t already, you could follow one of the Unity tutorials to get a simple game up and running, and then bring a lot of your own models or whatever stuff out there you can find for free and fill the scene out as much as you want. And then throw in post process effects, and texture everything, add in sound, etc. That will give a better picture of the many factors that effect your framerate performance. Keep in mind that game performance optimization isn’t something you will master in a week, or six months, or years probably. But if you think you want to build your own models, it makes sense to learn how to build performant models based on the level of complexity you want. For that, I’d start by browsing the polycount wiki and then looking for good game modeling tutorials, for instance from Pluralsight or Lynda.

3

Hey,

It’s a bad test kind of. The reason for this is it falls into one of the areas Unity is optimised for - drawing a lot of the same stuff, with the same material. This should be really fast regardless of draw calls.

Note the relatively low setpass calls? Setpass means “lets stop and set some stuff up on gpu” and it’s bad. So those are only 40-50.That’s great! lower is always good for setpass.

In general you want to reduce draw calls with instancing (for dynamic geo) or merging geo (for static), and reduce setpass by using the same materials where possible.

It’s kind of hard to slow Unity down if you do everything in moderation.

1 Like