so basically it will be like an Onlive/Giaiki-esque system, which you can set up yourself (and with a more even split of who does what)?
I have no idea of what you are talking about, but I can say: PA wont work like Gakai or Onlive. Those really stream a video to the client, PA will be sending game data that your game client will display as a running game. Thats a huge difference from a technical standpoint.
Yeah, the Server makes all/most of the Calculations, but the Clients still have to render everything. Mike
Mike are you 100% sure that the client will not have to perform any kind of simulation ? As far as i understood, if a projectile is entering the client field of vision then the server will send the initial position of the projectile, as well as the interpolated position and the function to simulate the projectile motion. To be honest i'm missing details about how all this is going to work. I hardly understand how all this could work without having the client performing a minimum of simulation, because that would that the server would have to broadcast in real time all units position entering each client field of vision. From a network bandwidth perspective, it looks like quite unfeasible without struggling the network.
While the client is probably doing some kind of interpolation to smooth things down it is not doing any real simulation. Simulation != interpolation. The interpolation would quickly become very inaccurate and therefore is corrected by the server who is actually simulating the whole thinge. ... at least that's what I would think of it.
This is correct, interpolation only exists to smooth out lag and reduce the need for additional bandwidth on updates. This Client/Server model doesn't seem any different from your average FPS dedicated server model, if that makes it easier for people to understand.
Except that neutrino said, that they will not need all of the lag hiding techniques used in an FPS. We'll see on how much they can cut down and still have the game feel responsive and hide lag.
I really wonder how much bandwidth will be required to play heavy games. While i do not see this as an issue if playing on a hosted server, it might not be as much scalable while hosting a game on my own PC (Core i7+32gigs or mem) as when playing on a hosted server, simply because my upload bandwidth is limited to ~500ko/s. Hopefully Neutrino can provide an average expectation, but it migh be a little bit too soon i guess. Have a basic question in the meantime : Let's consider we play on a hosted server with enough bandwidth so that all clients can download all the information from the server in the expected time, is the client upload bandwidth still a botteneck ?
Since the client will only need to upload his own commands it should not use more upload than a 1v1 in supcom.
The client is indeed a pure interpolation with no simulation going on. There are some exceptions to this because there are some things that only the client does (particles and animation for example).
I cannot wait to lay my hands on this game. Am so excited about the dedicated Servers, 20vs20 games, better cheat protection, you can setup your own little server or even rent one like many people/clans do for Battlefield and so on... Great!
In terms of the Uber run servers (as well as the Uber run Galactic war) do you have any plans regarding where they are going to be located? I for example am located in New Zealand, and frequently get left in the cold in terms of nearby servers. Servers located in Australia are normally okay but some games take a fair while before they set up oceanic servers for us folks at the bottom of the world. I guess it is less critical in an RTS vs an FPS, and of course it is likely that there will be community run servers but I am still curious about the ones Uber are going to run
By implication can I infer then that the CPU load may be less than SupCom? And PA is going to tax the graphics card comparatively more than the CPU?
Well there will be less sim work for sure so we can put more CPU into rendering. There is still plenty of work to do. We will definitely be pushing the graphics card pretty hard with all of the virtual texturing, deferred rendering and instancing stuff going on. We still want to be able to run a server for a single player game on the same machine so we can't go too crazy.
I am quite happy to hear this still holds some of the design focus. I do love and will play a lot of competitive multiplayer, but in all likelihood I am going to also run quite a few games from my machine to service my brother and a few friends locally. In addition to this I’ll probably run a fair few games locally for myself against AI when I feel like a less competitive game than I would find online.
The beautiful thing about computers is that they're deterministic systems - give two systems identical input, and they should produce the same output. The server won't have to send things like position updates for ballistics and other pure vector calculations because given a starting position and vector the server and client will both plot the same trajectory for ballistics, direct fire (lasers/beams), flight models etc. The server will of course confirm the hit/miss/damage because it's authoritative, but most of the interpolated minutia can be handled by the client just fine. Where many games run into problems is that they're not running a sim on the back end, they're "rolling the dice" so to speak, so if the client misses one or more sequence updates it wont necessarily be in sync with the server for doing those dice rolls, and you'll end up with warping and other junk that makes players go "Oh come on!"
