Graphic stuttering

[amigawolf]

In reality, if you have a decent PC with decent video card and plenty of memory and Windows 7, you should not see much stuttering in RailWorks, even with all options on maximum.

Mike

Phenom 940 Quad Core, 8 gig of ram, gtx 280, Samsung F3 1tb HD (probably the fastest 7200rpm drive you can get right now) windows 7 x64 & I've always had loads of stuttering on railworks, was just the same with rail simulator too, for me this game has never been anywhere near smooth & I'd say my system should easily be able to handle railworks!

[Wikkus]

Phenom 940 Quad Core, 8 gig of ram, gtx 280, Samsung F3 1tb HD (probably the fastest 7200rpm drive you can get right now) windows 7 x64 & I've always had loads of stuttering on railworks, was just the same with rail simulator too, for me this game has never been anywhere near smooth & I'd say my system should easily be able to handle railworks!

Nice rig. But what resolution are you running at? How have you got that drive configured? How fragmented is it? What do you use to defrag?

I run less gear and mine's generally silky smooth on the default routes with occasional hesitation (drop to maybe 25fps...) at 1920 x 1280 ...

Rik.

[ightenhill]

Theres something iffy witht he way the game handles AA and AF and its how your Graphics cards driver interacts with the game and sorts it out that stops the rubber banding .. You either cause a conflict and have them fighting each other or have it working as smooth as silk..

[paulz6]

I find the stuttering corresponds with high rates of working memory deltas displayed in the process statistics in task manager.
This is possibly a result of a large memory structure de-allocation/re-allocation operation involved by the loading of tiles, in which case those with faster CPU's and memory subsystems will perceive a less notable stutter.
I doubt the stutter is caused by a synchronised disk read block, which is probably performed on a helper thread, otherwise the stutter would be far worse.

[bremen]

I'm back

One thing is the game stuck for a moment to load objects and textures and this is caused by the speed of the HD.. the fragmentation... the amount of ram... the general "quality" of the setup of the OS.. drivers.. the quality of the loader inside the game (Gothic3 it a state of art in this camp... the stutter is everwhere but it has great shaders. ) etc.. etc... and one thing is looking to a single objects and the movement is like a rubber band as mentioned by Trainguy76 in his post

If using view number two, which follows the engine/loco, the surrounding scenery looks to be "jumping/snapping/studdering".
Much like a rubber-band.

Very unscientific approach. Performance not very good with the shadows enabled ... performance much better with the shadows disabled.

WHY is better whitout the shadows?
Because the shader for the shadow is compiled to run with ps2...
VertexShader = compile vs_2_0 VS_TransformAndTexture();
.. and the rest of shaders run with 1.1...

A great mix.. so the videocard needs to switch from one shader model to the other.. loosing speed in the mean time.

I'll never do an object for my route without shadows because i live in a world with shadows and i want see them in the sim.

[paulz6]

I don't see a lot of I/O activity in the process statistics to support any speed of disk access theories. I suppose anyone who as mounted RW on a solid state drive could help provide some insight.

[Acorncomputer]

I have a solid state drive on my notebook but the speed of access (perhaps 20 times faster than spinning disc) does not make much difference to the stuttering although the small amount there is does not bother me.

[vlado960]

WHY is better whitout the shadows?
Because the shader for the shadow is compiled to run with ps2...
VertexShader = compile vs_2_0 VS_TransformAndTexture();
.. and the rest of shaders run with 1.1...

A great mix.. so the videocard needs to switch from one shader model to the other.. loosing speed in the mean time.

I'll never do an object for my route without shadows because i live in a world with shadows and i want see them in the sim.

Your explanations seems logical. I hope that the diagnosis is correct.

[Wikkus]

I don't see a lot of I/O activity in the process statistics to support any speed of disk access theories. I suppose anyone who as mounted RW on a solid state drive could help provide some insight.

I trapped a bunch of log data yesterday whilst trying to trace a CTD issue; I was focusing on page faults and system I/O activity related to transitions between heavily populated tiles. There was a definite spike (10 or more times greater than "average") in Page Faults (and, of course, I/O activity) just prior to the CTD that I was seeing. I have a fast 10k RPM SATA RAID0 rig...

My testing is far from conclusive, but part of the issue seems to be (for me) when transition between tiles with over 2k assets showing in editor.

Rik.

[paulz6]

I'm not an expert on 3D graphics programming, but surely the shader processor switch theory would occur on every frame, and hence lead to slow frame rates rather than producing the odd occasional stutter?
What is causing the stutter must be something periodic, holding up the display of the next frame. I find it corresponds to working set memory delta's which are several megabytes in size. Finding clever ways to avoid dynamic memory allocation in the run-time environment can significantly improve the performance of software.

[paulz6]

I don't see a lot of I/O activity in the process statistics to support any speed of disk access theories. I suppose anyone who as mounted RW on a solid state drive could help provide some insight.

I trapped a bunch of log data yesterday whilst trying to trace a CTD issue; I was focusing on page faults and system I/O activity related to transitions between heavily populated tiles. There was a definite spike (10 or more times greater than "average") in Page Faults (and, of course, I/O activity) just prior to the CTD that I was seeing. I have a fast 10k RPM SATA RAID0 rig...

My testing is far from conclusive, but part of the issue seems to be (for me) when transition between tiles with over 2k assets showing in editor.

Rik.

A page fault dosen't necessary mean an increase in I/O activity, unless it is a major page fault. Minor page faults occur without any I/O activity. It does look like there is an increase in page faults associated with the changes in the working memory set delta, which doesn't surprise me. The I/O activity on the RailWorksProc.exe is far more sporadic than the incidences of stuttering.
I'd agree that the problem is probably due to the dynamic nature of the loading of tiles. I would be looking at why a run-time system's memory management needs to be so apparently dynamic.
Software is complex and very hard to diagnose by process statistics. I do not know enough about the impacts of 3D graphical systems, but their appears to be consistently dynamic changes in the working memory set. This would be my starting point of any investigation, and asking why it is so necessary.
If it was disk based, then a SSD should probably resolve the problem. However, this does not seem to be the case.

[paulz6]

Also, I believe that all assets are loaded into memory in RW during the route loading process. The more assets you have in a route, the longer it takes to load. These assets are not been loaded from disk on the fly.

[SaMa1]

Also, I believe that all assets are loaded into memory in RW during the route loading process. The more assets you have in a route, the longer it takes to load. These assets are not been loaded from disk on the fly.

Hmm, I've also wondered that, why does route with various assets load longer than route without any assets if they are loaded before every tile. What about loading textures into GPU memory? Usually textures take the most space in routes.Someone should really make test objects with various textures and shaders to experiment the stuttering. At this time the game is pretty much unplayable if you are driving over 160 mph and the route has more objects in high speed sections. Altough the stuttering time stays very much the same its far more noticeable in higher speeds.

[Basherz]

Also, I believe that all assets are loaded into memory in RW during the route loading process. The more assets you have in a route, the longer it takes to load. These assets are not been loaded from disk on the fly.

Hmm, I've also wondered that, why does route with various assets load longer than route without any assets if they are loaded before every tile. What about loading textures into GPU memory? Usually textures take the most space in routes.Someone should really make test objects with various textures and shaders to experiment the stuttering. At this time the game is pretty much unplayable if you are driving over 160 mph and the route has more objects in high speed sections. Altough the stuttering time stays very much the same its far more noticeable in higher speeds.

Certainly, all the Templates used for the route(ticked boxes) are loaded at this time.

[ightenhill]

On SADS blog, is he indicating by putting every asset in one folder it increases loading speed, or am I reading it wrong, the route certainly runs smoothly, I dont even get a micro stutter on it.. though that may be to do with its length (or is that largely irrelevent as well)