I made this effort post in another forum. I am posting it here for bookmarking
POSTERI had been noodling about the “clients within close proximity” aspect as well. I’ve not been able yet to understand how that would degrade a given program’s performance due (I hope) mainly to my ignorance of software (game engine specifically) functionality in this case.
But from a macro viewpoint, I would assume that if having multiple clients in close proximity does degrade performance in an MMO, then it should follow that increasing the game space available should spread out those clients and, therefor, improve performance (albeit artificially)?
Could the performance improvements that I am quite sure I have observed recently be just due to the increased size of the “play area” recently introduce in SC and therefor be the result of the decreased “density” of other clients?
If the 50 client max per virtual server instance is still correct, and the area of play has increased, then my proximity to other clients should be reduced, right? And from experience I can definitely say I do encounter fewer PCs in SC now (while the number of NPS has increased).
Sorry if I ramble on! This is all very interesting to me.
DSMARTNONE of the above. And it’s far too technical to explain to a layman. Here is a brief overview which I hope helps.
Having 8 clients within proximity inside a box, is less process (graphics, ai, networking etc) intensive than having 50 clients within the same proximity.
Having the same number of clients inside a larger box, doesn’t change anything.
A client inside a larger box doesn’t see any increase|decrease in performance because of the increased play area because it all depends on what is inside the box - and within view of the client - as well as how many clients are within proximity
CryEngine and Lumberyard are “level” based games. Meaning the world is created inside an editor - just like all games which aren’t procedurally generated. The larger the level, the more assets in it - the more the performance decrease when clients are inside. This is why even before moons were introduced in 3.0, performance which sucked before, became worse. Then they had to do something about it via OCS + NBC before they could increase the world size again in 3.3.
Since they couldn’t do it in time, they split 3.3 into two parts. First releasing 3.3 with the WIP versions of OCS+NBC, then 3.3.5 which had Hurston/Lorville with improved versions. It didn’t do anything to improve performance during a standard “gameplay” session. All it did was improve performance for a group of clients within an area that’s not heavily populated. It’s why some people are claiming 60 fps, then going to Lorville and getting 20 fps or less - with few clients.
SC is constructed like a series of interconnected boxes. Think of the PU start location Port Olisar (Crusader) hub, Levski (Delamar moon), and Lorville (Hurston) as 3 separate but connected boxes - all created inside an editor. It’s how they were able to disable Area 18 (shopping which was previously accessible from a menu like Arena Commander, Star Marine) and move all the shops to Port Olisar. It’s also how both AC and SM can be played separately, as they are just separate level boxes.
Each box has a different level of complexity and content which determines the performance requirements.
Welcome to
Stanton key locations currently implemented. The number at the end of each line is my approximation of the level of processing required for 16 (standard number) clients within proximity (clients can see and interact with each other) of each other.
- Space: (empty) - 4
- Space: (areas with hubs such as security hub Kera orbits Cellin, Covalex orbits Daymar, various rest stops) - 7
- Hub: Port Olisar - 6
- Moon (surface) - 4
- Planetoid (surface) - 4
- Planet (surface): 5
- Hub: GrimHEX (Yella) - 6
- Hub: Levski (Delamar) - 6
- City: Lorville (Hurston) - 20
So if you were getting an average of 30 fps in 3.2, then getting 50 fps in 3.3, it has to do with the OCS+NBC implementation. However, it’s sporadic because all those do is - in some cases - prevent the loading of assets inside boxes (and within a certain range of the current box) that the player can neither see, nor located in. But, as most have now seen, it breaks a LOT of things. Heck, even when in the same box, you can see clients and ships being removed prematurely, even if they are in the same box you’re in, within proximity etc. And when clients are all within the same box or within visual interaction distance, guess what happens? If you guessed that performance tanks again because now everything within the same box is processed, then you would be correct.
And THAT is how we end up with these fps spikes which go from bad to worse during “gameplay” interactions between clients. It’s also why you would be hard-pressed to find gameplay related videos which aren’t totally shit. Citizens and streamers instead focus on taking screen shots and nav videos because the game’s visual fidelity is mostly what sells it outside the backer base.
Performance in space is better than inside hubs and on the planets because there’s hardly anything in space. Also, performance on the moon/planet terrain is better because most of the assets (rocks, fauna etc) are procedurally generated like in all games. Meaning, an artist didn’t sit down and place every rock, or tree manually in the editor, as that would be an impossible amount of work.
The reason that Lorville is such a performance hog is because it is the largest box, and has the most assets. And they attempted to reduce the performance demands by preventing players from flying too close and within the city by adding no-fly-zones. Otherwise, anyone getting 20 fps right now flying outside those zones, will be getting about 5 fps or less flying within the city. So imagine what happens if several clients are in that city. And THAT is why, even after touting it for over a year, they had to take it out of 3.3 in order to implement performance improvements via OCS+NBC. When they brought it into 3.3.5, they added those zones once they discovered that even OCS+NBC won’t help because 1) the engine is shit and can’t handle that sort of thing 2) the visual fidelity of the assets is just to great a burden - and CryEngine/Lumberyard are notorious performance hogs in EVERY aspect
And what they did with Lorville, is precisely what they are going to have to do with
Microtech (the next planet) which they have also promoted as a city on a planet. For a hearty lol, take a look at the
concept art and descriptions for these locations, which they were touting as far back as 2013.
Further reading (my articles on OCS + NBC and why they didn’t yield the expected results)
10/27/2018
http://dereksmart.com/forums/reply/6723/10/15/2018
http://dereksmart.com/forums/reply/6682/09/22/2018
http://dereksmart.com/forums/reply/6639/09/08/2018
http://dereksmart.com/forums/reply/6615/02/20/2018
http://dereksmart.com/forums/reply/6239/12/11/2017
http://dereksmart.com/forums/reply/6078/
