Exodus Of The Machine Teaser 1

[Coppermantis]

Or it could just be aliens. That would probably be the easier explanation.  I'm just wildly speculating based off of very little evidence at this point.

Or it could be something which, surprisingly, nobody has suggested yet.

Not Humans, not Aliens, so that leaves...The AI? But that doesn't make any sense.

I'll just wait for the Alpha/Beta, I'm so excited.

And AI War is not dead!  We plan another expansion to that later this year.

 

Wow, you guys are being so productive. AVWW2, Shattered Haven shortly after, Now Skyward Collapse and Exodus, with an AI War expansion this year? So many opportunities to throw money at you guys :V

[x4000]

And hopefully, another two games or expansions on top of all that.  We're in top gear these days -- huge code assets behind us, more great game ideas than we can use in another three years at this rate, and a much-expanded team.  When you consider I'm not having to spend half my time on art, for instance, that makes a big difference.  And easily half of our programming time on Valley 1 was engine programming; even on Valley 2, there was a ton of that.  Now that's all done, and we're reaping the fruits of our past labors.  In the main, it all comes down to design and content and new art, etc, rather than so much new technical expansion at every step of the way.  It's like starting with the Unreal 3 engine rather than starting a new FPS game from a blank sheet of paper -- still a ton of work, but not in the same league, heh.

[keith.lamothe]

If they are humans who split off onto their own planet prior to the civil war, though, they must be capable of space flight and thus must have advanced weaponry.

Out of curiosity, have you read much mil-sci-fi?

I'm wondering if it's just the bits I happen to have read, but I've found that the "musts" in that quote there to be rather not the case

[madcow]

And hopefully, another two games or expansions on top of all that.... blah blah blah other stuff I'm not quoting.

I eagerly await my AVWW2 play-as-demonica prequel or kick-Elder's-rear sequel in expansion form.

[x4000]

Hahahaha.

[Coppermantis]

If they are humans who split off onto their own planet prior to the civil war, though, they must be capable of space flight and thus must have advanced weaponry.

Out of curiosity, have you read much mil-sci-fi?

I'm wondering if it's just the bits I happen to have read, but I've found that the "musts" in that quote there to be rather not the case

Generally no. I've read some Sci-fi like the Ender's game saga but other than that, I generally read non-fiction.

Although, upon further reflection, Ender's game alone disproves my theory. So yeah. Forget about that. 

[keith.lamothe]

Although, upon further reflection, Ender's game alone disproves my theory. So yeah. Forget about that. 

I thought it was an interesting conjecture

[Wingflier]

Hey speaking of AI War, you guys have done some fantastic engine optimizations to your newer titles (especially when hosting multiplayer, netcoding, and such), and it would be amazing to see those ported to AI War in the next expansion.

I mentioned this in another thread as well, but AI War begins to lag, even on the best computers (mine is rated 7.7 out of 7.9 on Microsoft's Performance Scale), when less than 1,000 units are on the same planet. From what I understand this is pure processing power because each unit is trying to make its own decisions and with so many units doing that, it completely bogs down the system.

While I understand the intention for this design mechanic, what 1 fighter does, is really not that much different than what 50 fighters do; it won't make that much impact on the game if they all just made their decisions together, instead of trying to do it individually. So what I'm saying is that I feel like we could redesign some of the game mechanics and port some of the engine improvements you guys have made over the years, and bring AI War's performance to a whole new level for the next expansion.

I'm really proud of your guy's work so far, I can't wait to see what else you have in store.

[keith.lamothe]

While I understand the intention for this design mechanic, what 1 fighter does, is really not that much different than what 50 fighters do; it won't make that much impact on the game if they all just made their decisions together, instead of trying to do it individually.

The cap scale option is available to you   Or are you already on ultra-low caps?

As far as engine optimizations, most of the ones that are really relevant to AIW are already included in it.  We'd like to get it (and our new stuff) onto a newer version of Unity but there's still some issues with that and I'm not sure if it'd really solve the problems you get with O(n^2) processing times.

[Wingflier]

Yeah, well I'll probably just shut up now because I really don't know what causes the AI War processing bog-downs.  I just feel like you guys could, in your greater knowledge, probably reduce that to something more handleable by most modern machines in the future. (I do play on low caps btw)

[x4000]

Hey speaking of AI War, you guys have done some fantastic engine optimizations to your newer titles (especially when hosting multiplayer, netcoding, and such), and it would be amazing to see those ported to AI War in the next expansion.

Oh, goodness -- AI War's multiplayer is by far the best we have in many respects.  It's a strategy game multiplayer model, though, so it's not suitable for action games.  Our action game multiplayer is actually substantially less evolved, although it works in many ways.  Still, AI War really couldn't work any other way in multiplayer than it does now.  It's blindingly efficient, to be honest -- no other strategy game has to contend with 1/10th of the data a big AI War game does.  Most don't contend with 1/100th, even.

I mentioned this in another thread as well, but AI War begins to lag, even on the best computers (mine is rated 7.7 out of 7.9 on Microsoft's Performance Scale), when less than 1,000 units are on the same planet. From what I understand this is pure processing power because each unit is trying to make its own decisions and with so many units doing that, it completely bogs down the system.

While I understand the intention for this design mechanic, what 1 fighter does, is really not that much different than what 50 fighters do; it won't make that much impact on the game if they all just made their decisions together, instead of trying to do it individually.

Actually, it would utterly break the entire game in terms of the AI that you love.  The AI is based around essentially flocking mechanics, and if you try to direct a flock as a whole it's no longer a flock.  And all the personality just vanishes.  We've been plugging away at efficiency improvements for years on AI War, and it's easily our most-tuned game ever.  It just has an incredible amount of computations per second to run (millions or hundreds of millions, depending).  None of our other games have remotely the same requirements, nor does any other strategy game I've ever seen.  But each year we've made some strides in AI War's internal efficiency.

So what I'm saying is that I feel like we could redesign some of the game mechanics and port some of the engine improvements you guys have made over the years, and bring AI War's performance to a whole new level for the next expansion.

Well, basically to get what you're after, we'd have to make it not AI War anymore.  That's really not something I ever want to do.  Computers are getting faster all the time anyhow, not that that's really an excuse.  But we have many options where you can play with smaller unit counts or a lower granularity of unit movements and so forth if you want to essentially have an AI War style of game with better performance on existing machines.  Those sorts of things have all been done and there's nothing else that we can really think of to do at this time.

Trying to split short-term targeting data onto its own thread is something we've considered, but it's so much data to transfer back and forth that that quickly becomes untenable.  We know this because we do have a secondary thread for the AI's higher-functions (all the really hard-crunching math, versus stuff that boils down closer to instinct).  And we have to significantly throttle the data transfer to that thread, or else bad things happen.  If we throttled the more instinctual level of data in that fashion, you'd see ships flying past each other without taking shots, or not noticing in a timely fashion when something bad happens or when they should take an immediate action, etc.

Basically, the design parameters that make AI War what it is are extremely constraining when it comes to things of that nature, and while I won't claim we've exhausted every last possibility, we've implemented or eliminated everything we have ever thought of.  I don't want to make AI War not AI War anymore (boy THAT would go over well), so our focus is to improve on what it is (which I think is great, for the record), and then to not ever do anything with those sort of hands-tying constraints again if we can help it.

I'm really proud of your guy's work so far, I can't wait to see what else you have in store.

Thanks for that!

[keith.lamothe]

Yeah, well I'll probably just shut up now because I really don't know what causes the AI War processing bog-downs.  I just feel like you guys could, in your greater knowledge, probably reduce that to something more handleable by most modern machines in the future. (I do play on low caps btw)

It varies but if you have a lot of photon lance stuff that's the slowest performance wise.  Which is why we only give it to a few unit types and those all have low caps.  But if you have a huge FS fleet it can cause bog down because of that.

Other than that it's largely related to situations where you and the enemy both have a lot of ships on the planet, so the targeting checks get pretty involved (those are aggregated, much like you suggest, but it still has a cost).


I do wish Unity let us do line-by-line cpu profiling.  That's the #1 thing I miss from the pre-Unity days, and it was the most effective tool I had for optimizing AIW (thankfully I really used it thoroughly before the switch to Unity, but stuff that's been added since then has not had the benefit of that tool).

[x4000]

While I understand the intention for this design mechanic, what 1 fighter does, is really not that much different than what 50 fighters do; it won't make that much impact on the game if they all just made their decisions together, instead of trying to do it individually.

The cap scale option is available to you   Or are you already on ultra-low caps?

As far as engine optimizations, most of the ones that are really relevant to AIW are already included in it.  We'd like to get it (and our new stuff) onto a newer version of Unity but there's still some issues with that and I'm not sure if it'd really solve the problems you get with O(n^2) processing times.

For the record, with the latest version of Unity there are some things that we could improve performance-wise on the graphical side.  That would help with certain things, to be sure, but when it's a big battle that's like 1-5% of the total costs and slowdowns from all the data we have.  As you said, it's the O(n^2) CPU processing on the unit calculations themselves that is the big thing.

I think that the newer versions of Unity do allow for more direct profiling, though I think that's mostly on the memory side of things.  Still, if they ever allow third-party CPU profiling that is better than what they have now, that would be one way to at least locate hotspots, as you note.

Oh, by the way -- just as a general note, when we did version 4.0 we did a lot of design shifts that were in favor of the CPU.  The smaller numbers of individual AI ships per planets, the barracks and carriers, the giant command posts instead of all those turrets, etc.  All those things helped immensely with the load the 3.x and prior versions had.  The game has really evolved a ton over time, but I just don't see where we'd take it next in terms of performance at the moment.

[Wingflier]

Oh okay. It was just a flaw in my perception then. You guys have done all you can and I'm happy with that. AI War just holds a special place in my heart and I wanted to make sure it gets the love it deserves <3

[x4000]

Oh yeah, we have both have a special place in our hearts for it, too.   And something may occur to us further, sure.  But I felt like it was worth explaining some of what we've learned over the months and months of man-hours we've sunk into improving performance on the game over the last four years, heh.   Not trying to bludgeon you with the info though!