Official v0.7 Thread

[kox]

The blockyness is a different thing.
At lagging i mean: the train moves smoothly, but the shadows are following it by steps. Lower resolution means bigger steps. That's why i said it looks like the shadows are updated in every second or more frame.

BTW will multilight shadowing going to be implemented in future versions of orts?

I know those "terrain shadows" are not really shadows just non illuminated faces, but still looking very badly, and i didn't seen anything like that in msts or in other games.

[twpol]

BTW will multilight shadowing going to be implemented in future versions of orts?

Yes, I hope to be implementing it at some point. It's a pretty complex feature though.

I know those "terrain shadows" are not really shadows just non illuminated faces, but still looking very badly, and i didn't seen anything like that in msts or in other games.

I hope, with mutlilight shadowing, we can generate real shadows from headlights - and even multiple headlights - which will look a lot better.

[kox]

Great. According to railworks 3 the shadow casting of headlights cause big performance hit(like everything in that messy tsx engine), so it can be disabled per light. But if i play tes v Skyrim, i see a lot of multi light shadow casts, with no noticeable performance hit.

[longiron]

But if i play tes v Skyrim, i see a lot of multi light shadow casts, with no noticeable performance hit.

That comparison is really not fair. You are comparing shadowing 100s professionally developed art assets over 100 meters viewable, each with four LOD, four texture mipmaps and four shadow maps to Open Rails trying to shadow 1000s of community built art assets over 1000s of meters, some having LOD, none with texture mipmaps or shadow maps.

And look at the performance hit for a gaming system with a high end video card (GeForce 560 Ti) Skyrim takes when shadowing gets turned up - Under 30 FPS. Low FPS is the death of a train sim because the illusion of smooth movement is destroyed, but is acceptable for a first person shooter.
FPS.jpg

We are not developing a First Person shooter where 99% of the scenery is within 30 meters of the viewer.

The OR dev team has done an outstanding job balancing fidelity and performance, so the game is usable by almost everyone in the community.

[kox]

I'm not compared it to openrails, because it haven't got multi light shadowing yet i just say there they could do very nice multilight shadowing with low system requirements. I compared skyrim to railworks where the shadows are blocky, flickering and makes the game to run at unplayable frame rates.
BTW that graphs are probably made with instanely high antialiasing(that's the holy grail of those gamers). My gpu which is not new (ati radeon hd3850) perfectly runs skyrim on overall high-ultra mixed setting at 30+ fps on 1280*1024 without Anti aliasing ( i don't like it anywhere, because it have huge performance hit, with very little improvement).

MSTS assets usually don't have texture mipmaps? That's strange, since those can be generated by a single click with various tools.
About shadow maps: as i see OR uses the main mesh for shadowing like other modern engines, so any built in shadow thing is not needed for assets (except ambient occlusion baking like every other games).
Anyway i saw the a lot of MSTS asset makers following the dogma of "only the polygon number that matters in MSTS", and don't know about the drawcalls which is probably causing more trouble and slowdown than the high polygon number. I've seen locomotives for msts that had low polycount but 30+! nodes which is surely cause a big hit to frame rate.

I'm very pleased with the current shadowing quality and performance of OR(my only trouble with it just the shadow darkening strength(i think darker shadows would be more realistic), and the blur radius(i think it's too much), which are probably can be set by a little constant in the shader code), and i'm wondering if the performance can be stay to be good with multilight shadowing.

[ED_4]

Hey, congratulations on 0.7 already. Great work Team OR! Not that far off now from 1.0. (This is perhaps in a year or so more?)

I wonder what we can see and do by that time? Maybe be able to also edit and create activities just like in MSTS, but for OR enhanced routes where MSTS stutters?

In any case keep up the great work Team OR!

[Stuz]

I agree with Edwin. Another great update. To see the load go on the loco is a great enhancement visually. Nice to see all the trees come back in Rogers Pass and Shuswap.All the backroom stuff in physics etc. makes for a great ride!
Regards Stu.

[twpol]

MSTS assets usually don't have texture mipmaps? That's strange, since those can be generated by a single click with various tools.

Yeah, I'd be surprised if this is the case - but I've seen no hard data either way personally.

About shadow maps: as i see OR uses the main mesh for shadowing like other modern engines, so any built in shadow thing is not needed for assets (except ambient occlusion baking like every other games).

This is correct; we also use the same LOD as the one visible on-screen.

Anyway i saw the a lot of MSTS asset makers following the dogma of "only the polygon number that matters in MSTS", and don't know about the drawcalls which is probably causing more trouble and slowdown than the high polygon number. I've seen locomotives for msts that had low polycount but 30+! nodes which is surely cause a big hit to frame rate.

I believe that for most GPUs it is indeed the draw calls that cost more; I have a pretty good graphics card and profiling has shown that the draw calls are a particularly problematic area. It is a tricky one to solve, though, since we need a seperate draw call for each material (texture + options like the shader name) and each instance right now. There's certainly room for improvement within OR, but models can obviously help here too. (Don't take this as serious advice yet, there's still much to investigate.)

[kox]

Draw calls are always problem for current 3d games if that's not a tunnel fps I can see in OR, the render thread is always stick to 98% cpu core usage while the other threads rarely touch it. This is common in games, while i can't see it really though(because i have ati card and amd don't have gpu profiling tools just for directx10+), but i suspect it

It is a tricky one to solve, though, since we need a seperate draw call for each material (texture + options like the shader name) and each instance right now.

How can be solved from OR? I though it can be only solved by telling other how can they optimize the models. And of course by making OR to support bulk object assets (like forest objects, or asset blocks in railworks3).

What about using .net4 or newer version of xna these may improve the performance?

[twpol]

How can be solved from OR? I though it can be only solved by telling other how can they optimize the models. And of course by making OR to support bulk object assets (like forest objects, or asset blocks in railworks3).

Currently, we're splitting things based on the exact primitive data in the shape files, but there are cases where multiple primitives can be combined (if they have the same texture, flags and heirachy). We could also combine the same primitives in multiple shape instances into single draw calls, in some cases.

What about using .net4 or newer version of xna these may improve the performance?

I think we're using some features they've removed from the next XNA, unfortunately, so it'd be pretty hard to test this, but we have done experiments with other APIs. No change yet though.

[releismetrain]

Hi,
Earlier I posted in this thread "One thing is the roads are not showing up,I think they never have ?"

The roads have now appeared probably for the first time since I started using Open Rails.

I changed the World Object Density from 0 to10 and that fixed the problem.

The World Object Density was always set at 0 in my setup.

Regards Geoff.

[OTTODAD]

The World Object Density was always set at 0 in my setup.

It's 10 by default.

So who or what has changed it ?

O t t o

[dick8299]

It is 9 by default in my copy.

[csantucci]

I've run an activity from the German Ruebelandbahn and I want to say that that was a pleasure; thanks to the pop-up of the event messages I could shunt, start the run and bring it to an end enjoying the wonderful route layout. Thank you to the OR-Team!
Leaving aside the still nonworking signalling I only would like to write here some small things I noted:
- on the log file I found error messages like this: Warning: Cannot parse the constant number 1.7m in D:\Programmi\Microsoft Games\Ruebelandbahn\trains\trainset\RLB_Holzroller \RLB_171-003.eng:line 1488
Opening the .eng file I found following related line: HeadOut ( 1.7m 3.4m 7.5m ) . I assume OR does not like the "m" after the numeric value. Maybe if it is not too complicated the parser could be arranged to accept the unit of measure in this case
- moreover on same log file I fond error messages like this: STF Error in D:\Programmi\Microsoft Games\Ruebelandbahn\ROUTES\Ruebelandbahn\World\w-005675+014939.w . Line 7029: Formato della stringa di input non corretto (Incorrect input string format).
Opening the .w file I found following block at the related line:
Signal (
UiD ( 3895 )
FileName ( hlhp1.s )
Position ( 48.6791 465.068 -489.888 )
QDirection ( 0.0171604 -0.594662 0.012699 0.803692 )
VDbId ( 4294967295 )
StaticDetailLevel ( 0 )
SignalSubObj ( 000000ff )
SignalUnits ( 8
SignalUnit ( 0
TrItemId ( 0 173 )
)
SignalUnit ( 1
TrItemId ( 0 174 )
)
SignalUnit ( 2
TrItemId ( 0 175 )
)
SignalUnit ( 3
TrItemId ( 0 176 )
)
SignalUnit ( 4
TrItemId ( 0 177 )
)
SignalUnit ( 5
TrItemId ( 0 208 )
)
SignalUnit ( 6
TrItemId ( 0 209 )
)
SignalUnit ( 7
TrItemId ( 0 210 )
)
)
)
Strangely enough, the line giving the failure is line "StaticDetailLevel ( 0 )"
- I have the impression, but I'm not sure, that in OR electric locos the compressor operates also with pantos down (I heard the related sound), which would be unrealistic
- There is one level crossing in the route (only one) where, even with barriers down and train passing by, cars go on trespassing the barriers.

Carlo

[kox]

Is ShadowMap.fx is the only responsible shader for shadowing in OR? (i wonder because i just read railworks3 shadowing shaders because of the low performance, and it's 5x bigger in character number, and much more complex, with lots of conditional and branched parts)