Showbudget

[m8nkey]

From the wiki "For those familiar with the Half-Life 1 engine, you can think out it as an advanced replacement of the r_speeds meter".

I'm new to source and don't really understand how this is more advanced than the good old fashion e, w_polly and r_speeds.

Sure you can see what's chewing up system resources but there's no set values for what is acceptable. Measurements of time spent in different engine systems isn't the same as a hard value for what it's actually doing in each of those systems.

I'm pretty sure my maps should be right as I'm used to working under limitations of hl1.

[Skjalg]

Yeah I've also complained some about this.. It seems its only function is to really see what stands out (in my case its always the brushwork :/), and try to optimize that stuff..

[fonfa]

the showbudget shows you more accurately what's affecting your map's performance, since it's split up in many categories. you can get rid of some props to make up for a more complex 3d skybox, for example

[Defrag]

What fonfa said. You don't really need to know what is 'acceptable' for brush and model polycounts because it's all based on the context / makeup of a scene.

In HL1, you had wpoly and epoly because that's all there was to it, really. You had brush rendering and entity rendering.

Source is far more complicated. If you had wpoly and epoly then what would it tell you? It wouldn't give you any idea about whether your scene was bogged down by transformations, pixel shading / fillrate, animation, detail props, static props, regular models etc. It's all about balance. I don't really care if my world rendering is hitting 5ms if the rest of my scene is taking a negligible amount of time, just the same as I don't care about my prop rendering if the world geometry's simplicity makes up for it. The only complaint I can make about showbudget is that it forces you to rely on your own system setup to extrapolate how your map will run for other people. However, even if it did give some kind of readout (polys, shader complexity, whatever), it probably wouldn't even help you much. Graphics cards supported by HL2 are hugely varied. There's no real way you could gage whether scene x will run acceptably on graphics card y other than by estimating or extrapolating from your own system. It's not like HL1 where T&L was done on the CPU and the whole engine would start quivering after you hit 1000 wpolys HL2 allows a much bigger polycount (for both wpolys and models) and modern graphics cards burn through high poly models anyway.

I.e. what I'm trying to say is that it's no longer about sticking to absolute limits for a couple of areas -- it's all about the makeup of a level in its entirety.

I make my maps with a target framerate in mind and, if performance dips to a point where I'm not satisfied, I look at showbudget and figure out what I need to optimise.

[fonfa]

Just compare your map's performance to stock maps.

[m8nkey]

The only complaint I can make about showbudget is that it forces you to rely on your own system setup to extrapolate how your map will run for other people.

That's what I was thinking, seems like the results are very system specific. Thanks for the advice, just wanted to make sure I didn't do anything fundamentally wrong on my current map.

[Bluestrike]

Valve just tests on lower systems as well but thats alot harder for us to do.

If you need a fixed number, this is what I did,

I went in my highest settings (hight end pc) and checked the dod:s maps for the lowest fps, and on my own maps I stay above it (100 fps in my case) or try to anyway.