🎉 Celebrating 25 Years of GameDev.net! 🎉

Not many can claim 25 years on the Internet! Join us in celebrating this milestone. Learn more about our history, and thank you for being a part of our community!

Is this type of RTS possible? (polygon count)

Started by
47 comments, last by Newgamemodder 3 years, 2 months ago

First link is about mesh shaders, which is a new replacement for vertex-, tessellation- and geometry shaders. Allows also things like more efficient occlusion culling. If used, it should help with rendering perf. quite a bit.

Second link just says new HW is generally faster than old HW, it seems.

Both links focus entirely in graphics, but there are other factors as well. Path finding or physics might have more impact than graphics, for example.

You can not expect a simple binary answer here. As said, see what your game can do and plan with that in mind.

Advertisement

JoeJ said:

Ultimate Battle Simulator 2 can do 1 million characters on screen, they claim.

Drone Swarm can do 32k.

Those are likely instanced however and may have LOD to boot, therefor at least memory wouldn't be a problem.

Yeah. But i want each to be customizable with pink or cyan hair or bald, cool leather jackets with spikes in all colors and kinds of boots, beards, and i want a slider for smooth gender transition as well! XD

ok so how does one summarize this?

A. It's impossible Hardware or software wise (no matter if you have the strongest gaming GPU available)

B. It's possible depending on the game engine

C. Something completely different?

It's a mix like so: A/3 + B/3 + C/3.

But trying to be serious and helpful, 2800 units does not sound unrealistic to me. If somebody wants to make such game, he surely can.

And i did correct by multiplying polygon count on screen with 60?

So unit wise it's ok, but not polygon wise? (including physics and everything handles by gpu)

JoeJ said:
slider for smooth gender transition as well!

Just… don't… or…. I will really have to implement such feature. That seems like a valid selling point these days!

I remember one of such tech demos back in 2008 or so. It was called “March of the Froblins”, here is youtube video for it - this one is with details for implementation (not just the demo itself):

I'm sure you can find the actual demo or video from it around the internet (just google for “March of the Froblins”. They had about 3000 agents running in realtime on high-end hardware at the time.

They also did some more logic than just “render the characters”, which was processed entirely on GPU if I'm not mistaken.

Unique color and features for agents included (@joej but I doubt there was fluid transition between genders, especially as demo shows froblins - I don't know details of those species).

My current blog on programming, linux and stuff - http://gameprogrammerdiary.blogspot.com

Vilem Otte said:
I doubt there was fluid transition between genders, especially as demo shows froblins - I don't know details of those species

Cute! I love them all, independent of gender!

(Though, back then i always complained about games not making eyes focusing some point in world space. Would add more life than fancy new GPU features.)

Newgamemodder said:
So unit wise it's ok, but not polygon wise? (including physics and everything handles by gpu)

If unit number X is the target, they will scale detail to whatever remains possible. Number of units / NPCs is harder to scale than graphics, so that's surely the more important limit than polygon count. But having no RTS experience i can only guess myself.

JoeJ said:
If unit number X is the target, they will scale detail to whatever remains possible. Number of units / NPCs is harder to scale than graphics, so that's surely the more important limit than polygon count. But having no RTS experience i can only guess myself.

So, in terms of RTS (now I'm hardly an expert - but I did internal attempts on both RTS and MOBA as proof of concepts, apart from interesting demos one can hardly call it a complete project - it also wasn't in demonstrable state).

As long as we stayed in realm of single-player nothing really was a problem on hardware since like 2006 or so. The big problem started appearing elsewhere, and I'm sure you know where - it is in multiplayer. Unless you want 1-on-1 you should stick with client-server architecture (and even for 1-on-1, cheating or such will be a huge problem, without server in place it is very hard, sometimes even impossible to detect such), and that is where problems begin…

Now, if you want to have 3000 active units in the game, if each of them is individual just keeping units in sync at 30 ticks per second will need quite a lot of data (position, orientation, velocity, angular velocity and some additional state variables, etc.), let's say we fit that all in 48 bytes of data. This still means you need 3000 * 48 * 30 = 4320000 bytes per second bandwidth (over 4 MB/s transfer rate). While one may object that we don't need to transfer inactive objects, keep in mind that this is RTS - it is more than likely that ALL of the units are actively moving, fighting or doing something at once. You will also need to transfer other, more static objects (buildings and their state changes, resources, etc. - but those are significantly smaller in terms of bandwidth - also those are somewhat easier to optimize - unlike dynamic units), but we can ignore them for now.

Sure, you can decrease ticks. Sure you can compress data based on some assumptions (to some extent). Etc. … So yes, you can optimize - but for each optimization you will need to sacrifice something (being less generic, additional conditions, limited map size, etc.). Each of the optimizations will be a trade-off.

Now let's say you optimize this down to fit in 512 kB/s requirements, optimize things to reasonable level … keep in mind that you're still running on a server. That also means you (or the one hosting games) have to be running the server.

How many server-side game instances are you able to run on single server (you will be limited by performance of the server and total bandwidth on the server)? Having both high means high running costs.

Most of the multiplayer focused real time strategies do limit actors in some way, examples:

Warcraft 3 - Introduced upkeep, limiting “food” to limit number of actors in game
Starcraft 2 - Limited supply for units
Age of Empires 2 - Max. supply per each player in game
Spellforce 3 - Limited supply for units

And I could go on.

My current blog on programming, linux and stuff - http://gameprogrammerdiary.blogspot.com

Well multiplayer (well online multiplayer) wont be a problem for me ? The mod I'm making is personal so I'll only play against AI (i've counted 2v2 in my polygon calculation). Essentially everything except dead bodies that stay on the map for some minutes are counted.

EDIT: Confused over the 4 MB/s part, do you mean the bus size would have to be that?, a modern GPU can run at 936 GB/s. Or are those 4 MB/s an online specific thing?

What i don't understand is if the rumors of the PS5 being able to run (theoretically) 8 billion polygons per second can't a PC run more/better?

This topic is closed to new replies.

Advertisement