Just something I was thinking about this morning but on Clinchfield the boxcars are more detailed than on previous routes as they have the working doors which can be opened. (As shown on the live streams). Now we've all noticed that the frame rates drop when coming across AI trains running in the opposite direction but something I've noticed is that the frame rates drop a lot more when it's a boxcar train coming the other way compared to when it's just a coal train. You can also be in Dante Yard and it be pretty full of coal cars and have no loss of frame rate. To me this is pointing to the main culprit for the poor frame rates as being the new more detailed boxcars and more to the point a whole train of these new more detailed boxcars. At this point I have to ask do we really need boxcars that have opening doors if they serve absolutely no purpose in the game at all AND cause issues with frame rates? I think I'd much rather have the standard boxcars in use and have better / more consistent frame rates each time I come across an AI train.
Opening doors isn't new - the Sand Patch box car has opening doors, the auto rack doors open, the NTP standard van doors can be opened. That isn't the cause of any framerate issues
Ok I wasn't aware of those other cars being as detailed so thanks for the speedy response Matt. In that case then what IS the cause of the more noticeable frame rate drops on Clinchfield when meeting AI trains? As I say you can do a scenario in Dante Yard for example with a lot of coal cars (I'd say touching 100 or more easily) and other loco's all around and no frame rate drops but as soon as an AI train rolls through with 3xSD40s and a rake of boxcars the frame rates drop by around 10-20fps then shoot back up as soon as it's passed.
It’s great that DTG have included such a detail, but every time I have opened a boxcar it is mysteriously empty. If you guys go to the trouble of modelling them, shouldn’t the ones with perishables from the south contain said perishables? I’m just saying that to maintain the immersion they should surely be loaded in at least one direction. Perhaps I’ve just opened the ones going back empty?
DTG provides nice details like opening doors on freight cars, and then get dinged for not loading the cars and suspected frame rate issues.... true example of “darned if you do, darned if you don’t”.
Must admit I didn't know about the SPG boxcars. I'll have to take a look! I would never complain about the little details that bring more reality and immersion to the game.
That's because static consists don't have physics, player ones do, and AI consists I think have basic physics at least. Also Clinchfield does have more cars on its freight trains than most, Sand Patch is the only other comparable one. But even then it's freight cars are longer, so I believe the trains on Sand Patch are longer, but they have less cars than on Clinchfield.
A chunk of framerate is caused by the simple act of moving physics volumes around - it's not a cheap process. The literal length of train doesn't really account for anything, number of axles is really the main differentiator for performance. if you had 2 axle wagons, you could have more for the same framerate. Clinchfield generally has 40-50 car trains except one service which starts with 60. SandPatch has 40-50 car trains also. We chose that number because it seemed to strike a balance between an acceptable length train vs performance. It's something that we want to improve, but it's also one of the most difficult things to improve because it's nestled in how the Bullet Physics library works. Matt.
Ok I’ve just run a scenario and have had to wait for two AI trains to come the other way. AI train 1 = 4xF7s with boxcars and reefers fps dropped by approx 18fps. AI train 2 = 3xF7s with coal cars fps dropped by approx 4fps. Before both AI train approached the fps was at the same level within 1fps. Surely that points to the reefers and/or boxcars on Clinchfield as being the source of the problem?
Depends if there are other trains moving off screen during one of those examples and not in the other, and the make-up of those trains. Let me know what times they are and where, i'll go take a look Matt.
The examples just from today were both in the CRR Extra 3016 South - Southbound Coal scenario. They were the only two passes with AI trains during the run. The first one in the first passing loop that you come to and the second was at the second passing loop you come to. Other examples that I have found so far though are as follows (I have only got as far as the CRR Extra 3016 South scenario up till now)..... Limited Power scenario. When the scenario first starts you are stood waiting for an AI train to pass. This is made up of 3xSD40s with boxcars and reefers and the fps drops by approx 10fps as the AI train approaches and passes. The Setup scenario. When you are making your way back after dropping off the first cut of cars you are passed by an AI train of 3xF7s with coal cars where the fps only drops by 3-4fps but when you are passed about an hour later by another AI train this time made up of 3xSD40s with boxcars and reefers the fps drops by approx 10-12fps whilst the AI train passes. It doesn't seem to matter whether the weather is clear or bad or whether it's at the start or middle of a scenario or run but the AI trains made up of boxcars / reefers on Clinchfield have a much bigger hit on fps than any other trains I come across on any other route.
The ones you mentioned that have bigger framerate drops have 3 SD40's on them, the one that didn't had 3 F7's on them. The former's twice as powerful as the latter, thus they're likely carrying longer trains, it could be that causing framerate drops, not the boxcars. If you're convinced it's the boxcars you should find a coal and boxcar train that are equal in number, and with the same number of units. As is it just sounds like the drops are being caused at least in part by longer consists.
OSD rolling stock has doors that work, the box cars with the side door & also the end doors for what I think might be car transport wagons.
Id say this comparison airs on the side of caution with regard to an exact train length match and also possibly the weather....... Fremont Ascent.... Meet with AI train 3xSD40 with 50 coal cars and a caboose fps drop of approx 5fps..... (In rainy weather) Limited Power.... Meet with AI train 3xSD40 with 40 boxcars/reefers and a caboose fps drop of approx 10fps. (In fine weather) F7 Tuturial.... Meet with AI train 3xSD40 with 40 boxcars/reefers and a caboose fps drop of approx 10fps. (In fine weather) So an AI coal train made up of the same locos but that is 25% longer than a boxcar train and is in rainy weather has approx half the fps loss.
Interesting topic, I wonder if the boxcars really do have a bigger performance impact for some reason. The opening doors shouldn't be an issue - after all the hopper cars also have animated, opening parts. The model of the inside is probably not an issue either, after all it's just a few extra polygons to make the walls, and the hoppers also have their inner walls modelled. So maybe the box cars are simply more detailed on the outside, having more polygons? But the devs surely has methods to check this and see how performance heavy each model is, so I suppose they would balance it Regarding this, just a thought, but do cars need to have all 4 of their axles simulated? I mean it's not like the grip on the rails would change so frequently that all 4 axles need to be calculated separately for 100% correct brake performance and such. After all I don't think we have things like leaves on the rails and wet/dry spots that would make a difference as each axle roll over them, right? So I wonder if things could be optimizied if instead of simulating all 4 axles, they would simulate only 2? Of course that depends how things are built up inside the simulation already. it could be too much work to change it now, just to make the simulation less accurate at the end. Just a thought anyway.
WaveyDavey some excellent feedback there and i've passed it on to the relevant people to take a look - those numbers imply something wonky on the boxcars/reefers that isn't wonky on the coal cars. It is unlikely to be poly's or textures - it'll be something magical and hard to explain like "sleep groups" or something. (bits of vehicles get put to sleep at different times because who cares about brake hoses if you can't see them, for example, so why calculate all their jostling around?) - sometimes we have found some bits get missed out of sleep groups and this can cause performance problems. Not saying that's the issue but it's an example. As I say, i've raised it. Simulating 2 axles instead of 4 - i'm sure I asked that question to the engineers once, and I got looked at funny and told to go away I mean there was a more detailed explanation of why it wouldn't work that I don't really remember now but that was the essence of it Matt.
Thanks Matt much appreciated and thanks for your input and assistance in progressing the possible issue.
Understandable, thanks for the answer and for the insight. Hope you guys can find the issue and fix it easily.
Silly question but may help to rule it out, you haven't possibly created or acquired a livery for the reefers/box cars? Also by extension one for the coal wagons?
