Are long consists impossible?

Post your problems and installation issues here!

Re: Are long consists impossible?

Unread postby RiscIt » Fri Dec 02, 2011 8:02 am

Nah I saw that. I read that as they are sick of dealing with people complaining about Donner Pass aesthetics (the multitude of scenery glitches and track issues related to superelevation and general accuracy - some spots, like the truckee river bridges south of Verdi are simply ridiculous, IMO - even at the 45 mph speed limit those bridges are dangerous). Framerate problems don't seem to be acknowledged at all, and those crept in back with HSC. Of course they may have been there all along, but I don't remember getting elastic coupler issues with any of the longer train scenarios that came with San Bernardino (then again, even those "long trains" are only 40 cars or so, so it's very possible we've always had this bug but never experienced it before).

There was a time when RSC linked to a graphics tweaking guide 3 times a day on the RSC Facebook page which said nothing more than "turn everything off, and then turn things back on until you can't stand it anymore".... (common sense for most folks, or so I had thought) In any case, if you don't have PhysX acceleration, "turning everything off" didn't help much. Even reverting to the pre-TS2012 mode leaves you with rubberband couplers. I completed the last HSC scenario with everything turned down and at 10 fps on a graphics card which gives me >30 fps with max settings in games like Grand Theft Auto 4 and Crysis 2. Those fireworks at the end were hard to see at all and they weren't that exciting as a slideshow.

So I still don't see them admitting there is a problem. I'm fairly certain they do their development and quality control in an NVidia bubble where the tracks are lined with candycanes, flowers, and PhysX acceleration for all.

Simply stating that the app requires a PhysX accelerator would be enough to take responsibility for the issue (assuming of course it is proven to be the case) but I suspect that would narrow an already niche market to an unacceptably smaller size. It doesn't help that "Versions 186 and newer of the ForceWare drivers disable PhysX hardware acceleration if a GPU from a different manufacturer, such as AMD, is present in the system." (http://en.wikipedia.org/wiki/PhysX#Hard ... celeration).

I don't envy the situation they are in, but not owning the issue is pretty weak, in my opinion. General lack of communication, particularly as they are so community driven, is a real sore spot with me.
RiscIt
 

Re: Are long consists impossible?

Unread postby Kali » Fri Dec 02, 2011 12:37 pm

Rubber couplers is because the coupler setup is bad, not strictly an engine fault - I would be enormously surprised if nVidia users didn't have exactly the same issues with stretchy couplers as the rest of us and two pages of this thread were talking about fixes that work that we can do right now. Moving around with the game paused also puts you in something of a bubble, I don't think it's doing as much asset sorting/culling as it usually does, so I'm not convinced it's all physics either ( otherwise you'd get abysmal performance on a flat green piece of ground with a huge train too ).

I would like to be able to run HSC without the whole game juddering to a halt every now and then when it finds a particularly fun section of scenery... there is definitely vast room for improvement.
Kali
 
Posts: 1600
Joined: Mon Mar 14, 2011 1:00 am
Location: England-by-Sea

Re: Are long consists impossible?

Unread postby RiscIt » Fri Dec 02, 2011 11:25 pm

Rubber couplers is because the coupler setup is bad, not strictly an engine fault

I'm not saying that rubber couplers are a symptom of the problem, I'm suggesting that they probably make the problem worse because calculating spring tension is more involved than calculating absolute limits with static bounds. I'm suggesting that if the coupling methods were changed, the physics calculations may be able to be simplified to run faster. That is, if the SpringCoeffiecient & Damping variables were factored out/ignored completely (rather than just set to values which negate them - they still get calculated in to things even then), then I bet the physics pipeline would get a bit faster. It would be great if they would at least get ignored once the MaxDistance variable is hit on the coupling in question.

Moving around with the game paused also puts you in something of a bubble

Yes you're in the usual bubble of not being able to roam too far from the train you're attached to, but it still lets you roam far enough to need to load new assetts as you go. Another observation I've made which annoys me is that in any decent 3d engine, if something isn't in view, it isn't rendered. That is, if something complex is hard to render, changing the camera view to look away from it removes it from the scene and your framerate improves. The framerate problems I experience in RW3 are present no matter where I look. In theory, I should be able to move the camera a few hundred feet upwards and stare straight up at the sky (so that nothing is in frame but the skybox) and see an improvement but I don't. So either the engine is too dumb to remove things not in view from the scene as it's rendered, or the low framerate is not due to scene content.

otherwise you'd get abysmal performance on a flat green piece of ground with a huge train too

You've prompted me to do some testing. :) I don't have anything conlusive enough to make any definitive statements on, but on the quick route I created with a 100 tank car 10 engine train and nothing but a few miles of track, I get 75-100 fps when it loads, and it immeadiably drops to 30 as soon as I click on the lead engine and it needs to start calculating movement.

...just sayin'... *!!wink!!*
RiscIt
 

Re: Are long consists impossible?

Unread postby Kali » Sat Dec 03, 2011 2:08 am

Hmm, interesting. And yeah, couplers could do with a clamp in the blueprints for sure ( although that is a couple more operations until you hit bounds ). I haven't made my springs hard enough to ignore the spring factor totally - you can see what happens if you do that, there's a screenshot of a flying train!

If I was looking to blame physics I think I'd also be looking at SE and bumpy track ( even when there isn't bumpy track ) given my messing around with 300+ levels of unevenness results in a near-pause when there's a large "pothole". I think loading up RW2 and doing exactly the same tests might be enlightening.
Kali
 
Posts: 1600
Joined: Mon Mar 14, 2011 1:00 am
Location: England-by-Sea

Re: Are long consists impossible?

Unread postby Machinist » Sat Dec 03, 2011 5:44 am

I've recently ran long consists (over 70 up to 100 cars) for VORA mini session using BritKits couplers (it was already mentioned on here by Griphos and Kali) and derailment issues have gone (even with powered-end GEVO's as helpers): the couplers are not stretching in the air out of wagons or overcompressing into the cars, and the gap between front and rear speeds was reduced from up 10-12mph (the disastrous "bungee jump" effect) to average 4-5mph when handling quite heavily the levers (throttle and brakes).
Who doesn't have dog, hunts with cat.
User avatar
Machinist
 
Posts: 1105
Joined: Fri Apr 02, 2010 1:02 am
Location: São Paulo, Brazil

Re: Are long consists impossible?

Unread postby g_nash » Sat Dec 03, 2011 6:12 am

Kali wrote:However there definitely is something related to scenery too, because I can drive light engine around HSC and get slowdowns in the same spots.


always has been like that .. from RS v1.0 to RW3

Kali wrote:
...............I think loading up RW2 and doing exactly the same tests might be enlightening.


have a peek at RS unpatched if you have it

.
g_nash
 
Posts: 462
Joined: Fri Feb 13, 2009 11:43 pm

Re: Are long consists impossible?

Unread postby Kali » Sat Dec 03, 2011 8:28 am

I don't, unfortunately; all I have around is RW2 116 and 117. I do get slowdowns in the same spots on HSC standing still, with one core capped out and my GPU use down by 40% or so... move a little - and by that I mean a couple of feet - and CPU use goes down, GPU up, fps up. I'm not sure if the 8-view bubble isn't actually the pre-computed asset bubble, but we'll never know that unless we go work for RSC...
Kali
 
Posts: 1600
Joined: Mon Mar 14, 2011 1:00 am
Location: England-by-Sea

Re: Are long consists impossible?

Unread postby sabreman61 » Tue Oct 15, 2013 6:48 pm

yes very much so What I did was go to edit a free roam scenario and pick a big rail yard to start in. then go to the consists button on the left and put 1 consist on the track and then another on the track next to it( and so on and so fourth) the key is having 4-6 locomotives in between each consist start with 6 to 8 locomotives in front and at least 2-3 n the back and 2-3 locos in the next consist in front and in back. after you make 3-4 consists on track next to each other slowly start by pulling your first train out of the yard and joining the next consist onto the back of the first 1 and so on and so forth
I built a 280 car 19 Locomotive train its hard to keep on the track gotta go slow but yes it is possible you just gotta build seprate trains and join them together make sure you have locos spread throughout the consist

Good Luck
User avatar
sabreman61
 
Posts: 207
Joined: Tue Oct 15, 2013 1:09 am
Location: Cortland NY

Previous

Return to Problems and Peculiarities

Who is online

Users browsing this forum: No registered users and 4 guests