[markus_GE]

Just some thing I noticed when running the latest experimental:

Even on default routes, where I earlier used to get around 40 fps on my middle class laptop, fps are no going scuba diving to about 15 fps - and I´m not talking of things like the MONON or the surfliner 2, just default Marias Pass...
I some time ago, shortly before the first intriduction of SE also noticed one such "drop" in performance, and I think I remember having read somewhere about "sleep" times (i.e. the time OR waits before rendering the next frame) being changed.

Can this have something to do with my problem, or can anybody tell me what I´m doing "wrong"?

Yours, Markus

PS: None of the old settings had been changed after switching (except for one keyboard input)

[James Ross]

Could you provide the following details:

• Which versions have which frame rates.
  
• What the Render and Updater Process values are like in each in the HUD (F5) DEBUG view (Shift-F5 until it appears).
  
• What the GPU is in your machine - just the line from HUD DEBUG will do.

[markus_GE]

On Marias Pass default activity "Building and Sorting Cuts of Cars": (the "official" dev-releases not mentioned perform virtually the same as the one listed before it)


X.1455: around 60fps, dives to 40; render/updater: ~98/~19%;
X.1477: around 50fps, dives to 35; render/updater: ~99/~16%;
X.1502: around 45fps, dives to 35; render/updater: ~98/~23%;
X.1542: around 35fps stable; render/updater: ~99/~24%; Superelevation amount 3, cab sway "notch" 1, distant mountains on
X.1565: around 35fps, dives to around 30; render/updater: ~99/~25%; DM at 50km, ViewingDistance at 4000m, others as above
X.1576: around 30fps stable; render/updater: ~100/~24%; rest as in 1565
X.1587: around 16fps stable; render/updater: ~99/~18%; rest as in 1565

Quite a significant dicline throughout, and two major drops in fps... I would be ok with around 40 or so, so laso more intensive routes like the already mentioned monon will run at some 20 fps or so, stable...

Graphics Card is an ATI Radeon Mobility HD 5650 (1024 MB)


Hope this is the needed info...

[eric from trainsim]

I had a similar experience in the past couple days --- turned out that on the video tab, the value for view distance got dialed out to 10km...

[markus_GE]

Checked it, but was properly set to 4000...

[James Ross]

markus1996, on 30 April 2013 - 08:57 AM, said:

X.1542: around 35fps stable; render/updater: ~99/~24%; Superelevation amount 3, cab sway "notch" 1, distant mountains on
X.1565: around 35fps, dives to around 30; render/updater: ~99/~25%; DM at 50km, ViewingDistance at 4000m, others as above

Okay, so you're playing with all the experimental stuff too. What happens if you leave all these alone (so 2000m viewing distance, all the others disabled)?

While performance for experimental features is not ignored, it must be considered a lower priority than for non-experimental features, so it might be that fixing this is not done soon if they turn out to be the culprit.

[copperpen]

I have no problems with FPS falling through the different X versions. They have in fact remained fairly stable. The one item that does give me a severe drop in FPS is dynamic shadows. Turned on I get an immediate drop of 40fps in the activity named above.

[markus_GE]

Hm, but turning of dynamic shadows means, I won´t have any shadows anymore, doesn´t it? Please correct me if I´m wrong.

The shadow feature, however, is one of the stronger points why I changed to OR...

Yeah, I know that of the experimental features, however, as I tried out, superelevation did not really affect framerates (+/- 2), same for distant mountains (+/- 1 fo every 10km) and as for viewing distance, I´ll have to try it out.


Actually (I´m wondering if I forgot to state it in the original post) what I originally wanted to know is, if this will be "corrected" in future (near or not-so), or also, if I could correct it myself by preparing my "own", self-compiled experimentals changing the sleep times between the single frames.

[James Ross]

markus1996, on 30 April 2013 - 01:11 PM, said:

Hm, but turning of dynamic shadows means, I won´t have any shadows anymore, doesn´t it? Please correct me if I´m wrong.

Yeah, dynamic shadows are the only option for shadows in OR at present. They do have a performance hit.

markus1996, on 30 April 2013 - 01:11 PM, said:

Actually (I´m wondering if I forgot to state it in the original post) what I originally wanted to know is, if this will be "corrected" in future (near or not-so), or also, if I could correct it myself by preparing my "own", self-compiled experimentals changing the sleep times between the single frames.

I'm not sure where the idea that this is just a sleep time somewhere in the game, but that should never be the case. That would be artificially slowing things down.

As for correcting the drop you have seen - we still need to figure out why it has dropped off so much for you. Viewing distance can certainly affect it, depending quite a lot on the route though, but it's still a large drop-off.

[markus_GE]

Tried it with standard 2000m viewing distance, everything else set as was above:

~20 to 25 fps, render/updater: ~98/~20%
doesn`t really seem to be the actual cause, also, I remember having read in another thread here at ET that viewing distance up to 4000m should not increase the amount of tiles loaded, just affecting the count of objects drawn.


Viewing distance back at 4000, no (dynamic) shadows, rest as above:
~40fps, render/updater: ~99/~31%

This seems to be the deal, but for me I think is a bit of a compromise...

Thinking of compromises, the avoid trees on track thingy comes to my mind. It was often said, I remember, to have bad impact on fps, and for some time could be turned on or of in the options window. As I´m running none of the routes that suffer from that problem, I don´t really need this feature to be running all the time, so maybe just reimplementing a switch for it might help something? Again, correct me if I´m wrong, as I´m not running any routes that need it, it didn´t read much about this.


EDIT: I in the regisrty added a DWORD "ShadowMapCount" with value set to 2 (half of the default shadow maps) to check out he impact: about 5 fps higher, so another thingy to forget about...

Maybe, howeber, combined with lower viewing distance... I´ll take a try.