Chris Roberts · Game Director, Star Citizen and Squadron 42

Well, I think the most immediate one is the object container streaming stuff, because it's the one that's going to allow us to expand the world in a manageable fashion. So we're working really hard on that for 3.3. It's a lot of work, because it requires every single piece of game code to operate in a thread-safe and batch-safe manner, which is different from how the engine was built previously. So there's a lot of legacy code that we have to rework, and all our new code is being built with this in mind, but again, not all of that was necessarily built with it properly, and we've got to test it and make sure it's all working. So the undertaking to go to object container streaming is as massive as, more massive than, moving to the large world [and 64-bit precision]. It's a huge effort on the engineering side. But once that's in, that's going to open up a bunch of stuff.

And then there's a few features beyond that, like the physics I mentioned earlier, because that's one of the last holdouts that isn't properly batched the same way. It's still multi-threaded, but it's an asynchronous model versus a phased model. So an asynchronous model is you're doing several things at once. So maybe you're updating the entities, then you're updating the physics at the same time, and perhaps even rendering at the same time. And then you have problems syncing between them, because the entity drives the render and the physics drives the entity. So the best is what's called a phased batch model, which is you update the physics, and you use every single core you can, and, say you have 10,000 physics objects, you distribute them to all these different cores and they do that simulation, usually in little islands, because any objects that would be interacting with each other would be on one core. And so you blast through that as much as possible, use all the computing power to resolve the physics, then you do the update of the entities the same way, and then you render the same way. So we're moving to that model. But that does require refactoring, and physics is one of the last things. And when we do that, the physics can scale much better, because the problem we have on our servers, which have a lot more cores than our clients, is that physics only operates on four cores maximum. So even if you have a server with 32 cores, you can't scale. And when we can scale the physics, that's when we can really increase the player count on any individual server. And then the final part, obviously, is the server meshing. So I think what will happen is with the object container streaming and then moving to the batch model of physics, we'll be able to get the player count on any one server to at least 200 or so. And then the server meshing will allow us to go quite a bit beyond that in terms of people in areas or instances in the same place. And then that world becomes truly massive, and everyone can exist in the same area and venture around and live their virtual life.

Jared Huckaby · Host, Community

Just gave Ricky a heart attack.

Chris Roberts · Game Director, Star Citizen and Squadron 42

That's on our plan. That's our project plan. I'm just telling you the bits of tech that are going to enable this seamless universe online.