Exponentially increasing 'upkeep' costs?

Currently, if a game of PA last "too long" basically everyone has infinite money and way, way, way more units are built than the game can really handle at more than a slideshow framerate even on high end PC's.

Some past RTS games have had a concept of 'upkeep' - a resource cost that increases as your empire grows by various metrics (unit count, resource income itself). An exponentially increasing upkeep cost would make the resource curve logarithmic instead of linear.

I think PA would benefit from this. The downside is no more swimming in money bins in the late game building anything you want. But the upside is, no more leaving 15 factories running, building units you no longer need, just because you can afford it. Perhaps the AI wouldn't just flood its planet with stacks of overlapping units it has no use for.

Thoughts?

This has been discussed a lot and pretty solidly rejected.

This is supposed to be massive scale.

Also, with an upkeep system, if I lose some economy and have a massive army, my army then shuts down because I don't have the economy to support it.

That's bad.

It's won't help. Most of things that actually make game lag isn't just unit number, but usually number of small behaviors that take most of sim time.

And yeah game is actually running on quite slow servers with really limited resources which motivate developers to optimize CPU/RAM/bandwidth usage. Actually performance improved a lot comparing to what we have in Beta.

Upkeep does not have to be based on army size. It could actually be based purely on resource income. Basically, metal extractors get "less efficient" the more of them you have, for example. Same with power. And there's actually a plausible explanation for this - right now in PA there is zero cost to transport energy or metal from their generation and extraction points to factories and structures that consume them. This upkeep could be that cost...

That is horrible, just as horrible game destroying idea.

I do not approve.

Can you explain why? Why did Warcraft have it? Is it because Warcraft sucks? Because while I didn't like it, that's not a good answer.

There's better ways to deal with large economies. Balancing metal spots, metal extractor income, changing costs of units and buildings, etc.

And I also don't think large economies are a bad thing. This is supposed to be a massive scale game.

Large economies and the large armies that come from them is a good thing.

People need to build more factories and fabricators when the late game economies get rolling.

Warcraft had it because it's an old game and computers could not handle large armies.

We have much more powerful computers now and can handle massive armies.

Anything that limits the size of our armies is not a good thing.

It's true that masses of wiggling units suffering from, let's calling it "pathing anxiety" are a major cause of slowdown, BUT one of the reasons this happens is just way too much stuff is built in the late game.

Don't get me wrong, I would LOVE if the game never got slow then it'd be fine the way it is. But the econ as it is makes it too easy to outbuild the engine's capacity.

Warcraft had it because it was a handicap if your opponent had more units then you.

And goes completely against the point of having a large army in the first place, and does not fit the theme of this game at all.

Warcraft like starcraft is a micro oriented small scale rts game, its priority's are completely different to a game like PA.

Path finding is not bottleneck for PA because it's using flow fields. As far as I understand pathing for 5 units and 5000 units have mostly same CPU cost if all units have same size and type because flow fields aren't created per-unit.

Anyway you can easily press Ctrl+P few times and check what exactly eating simulation time.

The game isn't finished. Mainly, the performance is far from complete.

We shouldn't handicap the game because we're running into performance issues on an incomplete game.

Uber is confident that we'll be able to play with a million units, so I'm not overly concerned with performance and vehemently oppose mechanics that will lower the scale of PA.

Brian is right. Uber's servers are not state-of-the art. In time, when it is necessary, they will have much better servers with CPU's that can handle a hundred thousand or even a million units. This is 2014 after all, and upkeep systems are honestly not for this type of game, or really for any game at this point in time with today's tech.

Logarithmic growth is usually great compared to linear, but i really do believe that is not needed, it takes away from the challenge of having to balance economy.

On the other hand, a game mode with a logarithmic economy wouldn't be so bad...

Man... I cannot wait until I can start hosting and renting out servers.

Sorry I didn't explain well. It's not "path finding" exactly, it's the little adjustments to make units exist in a non-overlapping formation, which they always strive to do.

It's easy to see this in action, actually, it just takes some time. Make a sandbox game, build about 20 vehicle facs, and have them produce infernos constantly. Do not give any move orders to the infernos OR to the factories. As the units build up by the factories, eventually the sim speed will slow. Build about 1000 infernos. Now look at the fields of infernos, all packed together. You'll see that many of them are "wiggling" a little bit. I am pretty sure they are trying to adjust position to not be so close together, in some cases overlapping. The sim speed shown by hitting Control-P a few times at this point in my test was around 5. Next, select all of the infernos and set them to patrol the entire planet. While they are moving, the sim speed will go even slower of course, in my test around 1-2. But once they are all spread out, select them all again and hit S to stop them. Sim speed is now nearly 10!! But when they weren't moving before, all squished together, it was half that.

This type of things also happens when you tell a few hundred units to go to one spot and there are landscape features in the way. Clearly an inordinate amount of sim time spent trying to tweak unit arrangements.

Above when you said "number of small behaviors that take most of sim time", I thought that THIS was what you were talking about.

Yeah, it's exactly what I mean and this is called flocking. There is also number of other small problems that cause lag like some crazy physics engine glitches when limb of bot starting to fly somewhere is space... Or at least I suppose it's one of them.

Your test is totally valid and I did something like that before:
https://forums.uberent.com/threads/just-notice-about-physics-in-simulation-performance.58214/

Unfortunately my test tend to show that there is always small part of groups eating 90% of sim time and when stopped/deleted simulation performance back to normal. So limiting number of units is clearly not right fix for this and this is clearly won't help much. E.g 10 stuck units can eat more CPU time than 1000 moving units...

Call me crazy, but TA was around in that era, and handled pretty large armies (for the time) quite well. Computers could handle large armies just fine if you designed your game engine with such a scale in intent, it's just that Warcraft wasn't designed for such a scale.

That being said, most of the crowd is going to be staunchly against a logarithmic economy, for reasons that ultimately boil down to it being against the style of RTS that PA falls into. TA and SupCom, both of which are what PA is spiritually succeeding, used linear economies without "upkeep costs" or other things to temper massive incomes, so it stands to reason that a game aiming to be far more massive than both in scale would retain the linear, non-diminishing economy that allowed such large incomes in previous games. The game's current choke points in processing probably wouldn't be resolved by implementing this, as I don't believe unit count is a major processing bottleneck and the current Uber-provided server sessions for the game are not exactly running on stellar hardware specs. A number of these bottlenecks are probably being gradually mitigated by the developers, and we'll all hopefully be able to really see what the PA engine can do in the coming months. Ultimately, until we get our hands on the server-side code and the game is very near release, it's very hard to determine what is an engine limitation, and with the amount of confidence projected by the developers that we'll be playing with gargantuan armies and large player counts, unit count probably will probably not be one of those limits.

In short, while that mechanic fits in with small-scale RTS games like Warcraft or Starcraft, it stands quite strongly against the grain of style and scale that PA is and likely wouldn't improve performance. Now, if you want that mechanic implemented anyway (for gameplay reasons or otherwise), there's always the option of server mods to make it happen, but a logarithmic economy will likely not be received as a fitting mechanic for the stock game.

Well, there is also the fact that moving around 1000 ground units across a planet DOES slow the game to sim speed of 1-2. That's where I started thinking it would be better if the game didn't enable building so much so easily. I'd LOVE for it to not be an issue, I'm just not sure that it'll ever be gone.

THAT IS THE POINT OF THE GAME!

The problem with this game's economy is that Uber kinda dug themselves into a hole that gets deeper the more you try to fill it.

The game is on such an enormous scale that it's very easy to just get +1000 metal, ideally you want to spend it, but you won't know what to spend it on.

Okay, easy fix, right? Just make everything take more metal and energy per second.
But wait, then the beginning of the game will be even slower because we'll need more start up eco before we can get a factory.

Okay, so add in exponential costs!
But hold on, now we won't have enough eco because everything will be costing and we'll have too much using too much.

Okay, then use less!
But then our armies will be pathetically small and we'll end up having surplus again because we're not using everything.

Okay, then uh... Make t2 mexes bring in less metal?
But we'll still have entire moons covered in metal extractors, there's no way that won't overflow the pool.

Like I said, gets deeper the more you try to fill it.