The game does support multithreading, in fact it uses over 16 threads for things like sound and music playback, etc, as well as the AI thread on the host. The AI thread can be extremely intensive, in some cases using a full core in order for the AI to make good decisions in cases where there are 80,000+ ships in the game at once, etc.
However, you are quite right that the core simulation is single-threaded. AI War, like all multiplayer RTS games, is based on a synchronized execution process. For more information about that from the developers of the original Age of Empires game, this is a really classic article that most game designers/programmers in the genre are familiar with:
http://www.gamasutra.com/view/feature/3094/1500_archers_on_a_288_network_.phpTaking that sort of approach, which is needed for any sort of reasonable performance over the Internet, the issue of multithreading in the main simulation becomes absolutely nightmarish because of the need for everything client machine to deterministically give the same result of the simulation. In this sort of game, all operations are cumulative (the order of which ship shoots first, and when exactly what happens in the simulation has a subtle effect on the outcome in the short-term, and major effects in the long-term.
There are also a lot of issues with sharing and locking of in-memory assets in today's programming languages. C# is better about that sort of thing than C++ in my opinion, but it is not ideal to the degree of some of these newer experimental programming languages (which are not yet suitable for game development for a variety of reasons). These sorts of things may be sorted out in time, but for the moment they are a challenge for highly realtime, highly-interlocking applications. For tools such as 3D rendering packages (Vue, for example), they are able to break up one larger task into smaller tasks and complete those independently, simply joining the tiny result at the end. That's a beautiful example of when multithreading works, and in some ways it is the premise behind what multithreading is used in AI War.
But for a single simulation that is order-of-operations-dependent, and which must give deterministic results, the short answer is that I don't believe that is a problem that any game developer has solved yet, AAA or otherwise. Those FPS games and such that use multithreading either just offload AI and other secondary logic, as AI War does, or they use an entirely different network model that is only suitable for games with smaller numbers of active entities (most action games use that sort of model, most strategy games have no choice but to use our model). With an action game you get players jumping from place to place during lag, which is evidence of a desync that is being corrected, but in an RTS title it doesn't work the same way because there is so much more going on in a simulation sense.
As to the question of 64bit, that really only helps insofar as giving you access to more memory -- using 64bit executables has actually proven to be
slower in the past when testing AI War, so it runs as 32bit only. The reason for this, near as I can tell, is that allocating so much bigger integers causes more allocation/deallocation overhead. The performance boosts that are often touted with 64bit architectures are very real, don't get me wrong, but those are all well below the programming level for the most part -- motherboard bus pathways and general memory/CPU architecture is far better, etc. You get the benefit of those things even on a 32bit application. The other, more OS-specific speed boosts of a 64bit architecture are also still in place even with 32bit applications.
Performance is something I'm always tuning, and is something that gets progressively better over time. At the moment I don't have any expectations of increasing it more significantly anytime soon, but you never know what I might come across. The latest 2.001 prerelease versions should perform better than the vanilla 2.0 version, by the way; so if performance is a concern, you could try that out (you don't have to buy the expansion to do so).
The performance that other players have seen varies based on their hardware (clock speeds are often misleading, as CPU/motherboard architecture can make a
huge difference in performance -- there are threads with players talking about that elsewhere on these forums), and the scenario as a whole that they are playing, as well as how many ships are active at the time. If you have 70,000+ ships in the game, but are only directly controlling 3k at once, that is still going to lead to some lag on a machine with your specs. On the other hand, one player has reported using a 3.5Ghz quad core to control a few thousand ships on a map with 201,000 ships total on it. That's well above normal usage for the game, and it runs quite poorly on my own 2.4GHz quad (Q6600). But I've not had significant lag with the numbers you're talking about, except sometimes if there are a lot of collisions going on at once with ships coming out of a wormhole or something. Still, compared to other, smaller games with less-intensive AI (SupCom comes to mind), the performance of AI War is much better in the main.
Anyway, rest assured that continuous improvement of performance is something I always have an eye out for, but I'm not planning any major rearchitectings of the game at present, as it's quite well architected in general at the moment and beats the pants off of pretty much every competitor when it comes to the scale/performance of the simulation, even if it does drop speed during some of the more intensive battles on some hardware.
Thanks for your support, and you interest in the game!
