Multiplayer Recommendation, Serious thought needs to be put into this.

This is a recommendation for a way of making multiplayer, not a gamer just asking for a feature. ( I know you guys haven't gotten a chance to add it yet, and that's good.)

I have a very high end PC(the CPU is at 4.2 GHz with 8 cores), yet, with just 6 cars on a typical map, I only get 20 FPS. Yes, multithreading could help if it isn't done already(preferably, the game will do most of the physics on the GPU.) However, in my estimation, it will be very difficult to retain the same gameplay and greatly increase the number of vehicles that can be used at one time.

Therefore I have this recommendation. Add multiplayer, and have each client responsible for their own car's physics. Then have updates on the state of the car's body be sent only occasionally. (preferably only when there is significant change.) (So, there is a computer for each car.)
This will allow the game to be run with many cars racing all at the same time, so you could have NASCAR like races, which is a great thing to do with this game. (I don't care if you prefer F1, the point is that it's high speed with cool crashes and lots of cars.)

The way I see it, this is going to be the only way to allow people to ever have a large number of these cars in game at any one time and the only way that this game will really be able to take off in popularity. (not saying it's not reasonably popular now) This won't work in split screen or with AI's. This will only work with online multiplayer.

Now, considering that you all have gotten this far in the game's development, I wouldn't be surprised if you have already thought of this. No matter what, I believe this way of making the multiplayer, or something similar, MUST be implemented for this game to really be worth playing. That is why I have posted this.

 

I am pretty sure Dirt Showdown had a great online rampage experience. but that was becoz of the PSN, maybe we should start a BeamNG network.

 

Starting you're own gaming network with servers and all that junk is by no means cheap. A lot of big companies can't afford that kind of stuff, let alone a young indie game company.

 

The latency problem can be solved by transmitting the updated state of the cars slowly so that only a little more than the usual amount of information is transmitted each update. Yes, this leads to updates on the cars state being delayed.

You may be able to make this less noticed by having several crashed states for the cars, one or more for each approximate angle that the car can get hit on. Then load up that pre-damaged state when a collision happens and the client is still waiting for an exact update of what the car looks like. (Like, load a version of the car that is damaged in the front when there is a head on collision.)

 

dude, the point of soft-body physics, is that the damage is calculated, AS IT HAPPENS, instead of having pre-damaged models

 

I really hope for LAN! That would be the best thing ever!

 

its true that... i buy lfs on 2006 s2... and on this day... on 2013 the cars only fly with lag xD but have nice system online running... but no crashing cars...

 

Flatout 2 and Ultimate Carnage worked pretty well if there was less than 200 latency. Cars behaved reasonably well. Again, however, not exactly sure how to do that with the soft-body sim physics. Those were pre-modeled damage

 

well, i imagine they would at least add lan multiplayer with crashable vehicles, mabey not full online mp for a while though

 

It seems you fellows missed the point of my last post. The pre-calculated damage models will only be displayed for a few refreshes while the proper damage data is being transmitted. It does not completely defeat the point of the game, it allows it to actually function as a game and not just a novel crash simulator.

 

and aprently you miss our point, the point of this engine, is to NOT, have pre-calculated damage models, EVER, all calculated at real time, PERIOD

 

Judging by your pc's stats, you're one who would benefit from this kind of feature. This system doesn't change the way the game plays or how damage is calculated. It only delays the display of the damage on cars other than your own. (which is what will happen to some small degree no matter what in any multiplayer version of the game)

Making any large piece of software is an engineering project. In engineering, you have to do whatever must be done to get the task you set out to do done. You can't just design on principle.

Even the game as it is compromises in some areas. Look at the collisions. They aren't perfect, and don't think that it's not possible to make them perfect. It's just impractical to do so.

Anyway, The singleplayer won't be touched by this system, and the multiplayer won't have crash physics that are at all different. The display of the damage will simply be delayed.

So, chill the f*** out. It won't corrupt gameplay or the end result of the physics.

Any precalculated damage will only be a way to fool the human eye into thinking that the damage calculations are done instantly.

 

Your idea has already been done in fact. Burnout Paradise uses that kind of system to deal with damage and crashes. Almost that is, instead of the client calculating it's own physics then sending it, it only sends position data, and each client connected calculates physics for each car. but like gabester said, the latency is the problem. Playing Paradise online is brutal. There are delay times between action and reaction is MINUTES sometimes, and since the calculated physics for each person is different, the damage is also different screen to screen.

Older burnouts were different, since cars had set crashed and uncrashed models, the engine had a huge amount less to deal with. It just had to deal with player movements. Then swap the undamaged to damaged models out.

So bringing it back to Drive, we can't switch out models. Soft bodies are very different then solid, needs tons of processing. So there's way to much, and it wouldn't be much to inconsistent, to calculate client side, and latency is a killer.

I agree, I think the best we can do with today's tech, is LAN.

 

All clients should have the same models since each client sends their own model to each other client. Yes, there will be a delay, and at times the models may not line up perfectly since collision calculations might be done with old models, but that doesn't mean that the system is worth abandoning. If it gives the potential to have something like 30 cars racing all at the same time, it's worth including.

Yes, I think there should be a LAN system that has more complete updates, but I don't think that is a substitute for having online multiplayer.

 

There are two options I can think of. Having two separate multiplayer mods.
1. Hosts PC runs everything meaning collision work fine and are enabled but there is some 'input lag' for the connected players depending on their pings. Limited to maybe 4-5 cars.
2. Everyone runs their own vehicles, interplayer collisions are disabled but there is no 'input lag'.

1 is good for a few people goofing around while if you want to race, etc which you don't want input lag for you'd use 2. That's the best solution imho.

 

Or, you can have a 3rd mode that is like 2 but with collisions. (just not necessarily accurate ones)

 

ummm, my i5 has no problem running beamng, idk what your talking about, if my intel hd 2500, thats also fine for now, i get a solid 25 fps on gridmap, which is fine for now, as there's plenty of stuff to do on it