I’ve got a feeling the gt 1030 will play just fine --- Post updated --- Btw your welcome for the info
Ok cool, was worried about heavy bottlenecking but as long as things are stable I ultimately don't care. btw, thank you for the info
I disabled 4 cores of 8086K, running only 2 cores and HT @ 4.3Ghz, ZERO effect to performance, absolutely no change to anything. There is only one exception, Gamergulls's traffic mods, but anything else, absolutely no change at all. Multiple cores are totally overrated, at least in my use per core performance is only thing that really matters 90% of time, rendering and running many vehicles is where it starts to matter. I can run 3-4 vehicles with 2 cores while holding 60fps just fine, so number of cores is just for rendering and running multiple vehicles in this game, with single vehicle even at Italy there is no practical performance gains from more than 4 threads. It is better to have faster core speed and less cores, fast i3 models will have better performance than dual Xeons with this vehicle. I would rather have overclocked i3 K model with 4 threads for 1-2 vehicles than Xeon or Ryzen, because performance will be much better in BeamNG. For other uses multiple cores have of course their place, but to run this kind of high quality detailed vehicle, overclocked i3-8350K would be really hard to beat.
Yes, you are completely correct and in that regard my single core performance is pretty low. I know it is the physics that is pulling me down though because when I pause the physics, my FPS shoot straight back up to 60FPS with no issues. If it were a render problem (like... to many pollys on screen or something) then I should see little to no performance increase when I pause the physics. So yes, you are correct. My computer can run 1 car just as well as I can run 24 cars just because of the way BeamNG uses processing power. That being said though, I expect this thing to run much much better when I am all through and done with it. Right now there are mistakes that will need adjusting and things that need fixing... there are probably nodes that are rubbing against tons of planes on the vehicle that are causing massive amounts of collision calculations that I am just not aware of right now. Plus there are instabilities that pop up from time to time be it duplicate beams that I put in on accident, too short beams, too high spring or damping rates... whatever... you name it... and if there are instabilities (where things vibrate on their own for what seems like no reason) then the computer has to run tons of numbers that would normally (under stable conditions) have answers that don't fluctuate very much. When one of these instabilities happens, it tends to suck up more computing power (been like that ever since RoR even. Not going to claim I know why it does that because every physics calculation should end up with an answer anyways whether there is stress on the beam or not, but that is just a trend that I have noticed) Anyways, long story short, I am paying attention to the performance of this thing constantly... It will get faster with time... but right now... it's not too fast... but that's fine because its a WIP Edit: So if I or Sithhy ever post a video and the FPS is sub par... just take into account that this is not representative of the final product.
As it is vehicle per core until all cores are used up, there is way to split physics load for more cores. Only thing you need to do is set up body to be another vehicle that connects with tow hook nodes, then figure out how to have tow hitches to be connected when part is selected and also how to select body from parts menu when it is different vehicle, it is only half impossible feat to pull off But it really is incredible how much per core and all cores computing power we can have these days, it is possible to build so much more complex things, but still if we could split very complex beam structures to few different pieces, it would help to run even more complexity. I think that using hack like towhitches might allow more complexity in some cases, but realistically for Brutus, I don't know if that route will be possible even for experimenting purposes. There are some challenges. Constant collision on node and coltris at some spot can eat a lot of performance, it can happen really easily and can be hard to detect. With tracks, you probably can't avoid such, which might also contribute a lot to how heavy it is running. Instabilities can hit collision like million times in very short time, which might be why those impact performance. Can be something else too, but active colliding is something I have notice to have impact, so maybe it is something from there that opens door to figure out those performance effects. Can tracks be attached via multiple towhitch kind of connection? (like MrCrash tractor attachments have multiple of those connections) Kinda like powered trailer, but again how to connect it on spawn and other partsmanager issues prevail, it might only be possible to use tracks as predefined config and not selectable via parts menu that way, but it might give big boost on performance as said "trailer" runs on different core. Gear is not relayed to trailer though, throttle can be relayed, so that would be big obstacle with that hack, maybe something is possible to be done to that via custom controller, but these kind of thoughts I have had with giant robot I was planning once, kinda would of been possible to loose an arm as it would of been connected to torso and hydros by tow hitches. Maybe thoughts give you ideas for some projects some day, if not useful for this project, freedom to create structures in this game are often far greater when going outside of box, absolutely no limits, but figuring out how to make it possible takes often several attempts Oh and there is LUA, CPU graphics stuff, UI and such that compete with each other on mostly one thread, then physics of single vehicle, but it gets bit more complex nowdays with 0.15 and newer as graphics stuff can occupy more cores, anyway point being, there is limit how much and complex LUA you can run, LUA might get paused with physics too. Think about ESC and alike, not likely to have effect in your case, but just bringing up how complex it is to actually know if it is physics or something else. In your case very likely the physics, but trouble is it can't be generalized, takes lot of experiments to really find out which it is and still at times it is very hard to tell, new performance profiler tool of the game helps a lot though, have that on and pause physics, see which drops a lot, check task manager which cores drop a lot, change graphics stuff, shadows, reflections, shader and mesh details etc. that helps to figure out things, still not easy, but you can figure out even more new things from there Maybe some day it is possible to split parts of vehicle to other cores, 100 000 beams might be bit much for one core, but 10 cores will not pulls much of sweat with it and just think what monstrosity it is possible to build if it is possible to use all cores for single vehicle!
I like the way you think... I think I even suggested something like this before somewhere... can't remember where though. I am willing to bet that I don't need to do something like this to get the tracks working with higher FPS mostly because I can run the truck with tracks without the body while towing another B1 with the body (but on normal wheels) and my fps is fine (Yeah, I know, that's on 2 cores now, but still), but once I apply the body to the tracked B1 my FPS takes a dive... fairly hard. That being said, I think it is some interaction of some sort somewhere (thats a sentence and a half) that is messing things up. I know I have eliminated a lot from it already by just gapping things better on the JBeam in particular spots. When I first implemented the tracks, they could barely move without exploding and in turn meant that my FPS dive bombed to a mear 14 FPS. Now that I have had some time to work with them and steadily increase their stability and remove a lot of unnecessary details from them, they have soared to over 45 FPS and, if I keep at it, I am willing to bet that the more I keep optimizing them, the more that frame rate will increase until the load on the system basically becomes a none issue again. Also... this was that surprise thing I was talking about on your "Guys come to talk about tires" thread... now that the cat is out of the bag, I guess I can talk about it freely there so that everyone can see what I am on about. Yeah, per core I think my computer can handle about 80 to 100 thousand beams... but once you throw collisions and graphics in, that number plummets to about 5000 or so... that might not be totally accurate numbers, but it's in the ballpark.
Used Xeons are not that expensive, depends which model one goes for of course. Also with around 2000 euros you can get pretty much all the performance what you possible can, getting more expensive hardware is not helping at least with current BeamNG version and anyone that has a job and does decent job at handling money and keeping expenses low can get 2000 euros saved in a year or so. Just 2 Cores, almost 6000 beams each bus, that is 24000 beams with 2 cores only, but 4 vehicles, 4 threads and graphics and stuff. One thread can do 2 buses or 12000 beams or so, graphics etc. take also at least one thread. I think your machine can handle lot more than 100k if spread among all cores, as long as care is taken to avoid graphics becoming bottleneck on CPU. Gotta love IPC and clock speed in same package. Now in theory this should run 12 buses with 6 cores enabled instead of 2 Cores enabled, but each bus adds something to one thread (graphics and who knows what) which then becomes a limit before getting to use all cores, I think same happens for you but just bit sooner.It probably gets better with more improvements to game. Performance in this game is tricky thing, but 4 buses on 2 cores only, that probably stirs some thoughts. FPS will plummet if I crush those buses though as more collisions and more something which could be called instability in structure starts happening. So yeah, with tracks stability is certainly going to help with performance. Sticking is then another thing that is bit hard to solve with tracks. Any node that is at edge of thin structure is like a hook, it can be used to cause incredible amount of destruction, but with tracks running over vehicles can become challenging if track is kinda flat plane kind of thing, but with more 3D shape to each piece of track, nodecount and weight goes up, so it is annoying puzzle for certain. Will be interesting to see what kind of solutions you come up with!
I actually bought my computer off of eBay. Its an old Mac Pro 5.1. I got it for about $200 bucks because according to the listing, the computer would turn on but wouldn't display anything on screen. I just figured it was a dead graphics card, but turned out the guy must had had a bad cable or something because when I got it, I had no problems at all... just fired it up and it ran beautifully. Then about 6 months later I threw a used GTX 1070 at it and a 2Tb SSD and that really pumped up the power quite decently. So for about $700 total I have: 2 Xeon 6 Core processors with Hyperthreading (24 virtual cores) 32 Gb's of Ram (can be upgraded to 128... but I mean... I have no idea why I need that) 2 Tb Sandisk SSD + 2Tb Hard Disk NVidia GTX 1070 So... yeah... I use it a lot for heavy calculations with FEA (things that my work computer just can't handle) but it runs as a pretty good gaming rig as well. Yeah, your probably right... I have no idea how far this thing would go, but it really only starts struggling (usually) when the GTX 1070 maxes out... I have had up to 64 cars on screen at once before, but it just can't handle that many pollies. Pausing the physics did literally nothing to the framerate.
They are specific to the B1... but converting them wouldn't be difficult if someone wanted them on something else. I probably won't do it... but someone probably will anyways. They are attached to 2 nodes just like normal tires, then have just 2 shocks for stabilization that connect to the top of the hub. (to attempt to keep them from flipping over) They also use the same drivetrain designations to drive the wheels off of the stock driveshafts (FL, FR, RL, RR) so there is nothing you need to change powertrain wise to get them to work with a normal vehicle. Again... I probably won't do it... but I am sure someone will.
