Railworks performance

[peterhayes]

Doc
Hope the 460 arrives soon - it has left Australia but probably travels to Brazil via Greenland.

Papa Express thanks for the links.
IMHO frame rates are a poor indicator of performance and smoothness, even the counter can have an impact on the rate and only recording differences of 1 fps is pretty qualitative. (to me a micro stutter would indicate something was happening at the microsecond level, and a macro stutter would be happening at the 1 second level). I'm always amused when fraps records 2560 fps or more when loading TS2012 it makes you wonder what is going on. However these two post have at least given us some benchmarks and the results there look consistent.
Personally I would rather see frame (dwell) times (difficult to measure), with 100% dwell time of say less that 20ms, but that's Nirvana. If you have 100 frames and 99 have a frame time of say 20ms but one has a frame time of 50 ms or more then you are likely to see the dreaded stutter.

I think the reason that the GTX 680 handles RW3 so well is that Kepler has improved the latency of these cards over Fermi and they possibly handle vertex buffer objects (less stutter) than previous cards especially when running Ts2012 with everything to the right. Right now I'm waiting for the next series (110 ) which I thought would be the GTX685 but that may not be the case it could even be called the 785, but it should show a major advance over previous series with updated architecture.
Thanks again
pH

[snowstorm]

IMHO frame rates are a poor indicator of performance and smoothness, even the counter can have an impact on the rate and only recording differences of 1 fps is pretty qualitative. (to me a micro stutter would indicate something was happening at the microsecond level, and a macro stutter would be happening at the 1 second level). I'm always amused when fraps records 2560 fps or more when loading TS2012 it makes you wonder what is going on. However these two post have at least given us some benchmarks and the results there look consistent.
Personally I would rather see frame (dwell) times (difficult to measure), with 100% dwell time of say less that 20ms, but that's Nirvana. If you have 100 frames and 99 have a frame time of say 20ms but one has a frame time of 50 ms or more then you are likely to see the dreaded stutter.

Chiming in a bit late here as I haven't been checking the forum (or playing RW) much recently. Just wanted to address a couple of things:

- I agree about the shortcomings of FPS as a measure of stuttering. But keep in mind the standard, accepted performance reporting methodology around here involves posting frame rate figures based on glancing at the game's FPS counter, and maybe some subjective impressions. Charts based on consistently recorded frame rates over time are quite a step up from that, and didn't do too bad a job showing slowdowns and instances of major stuttering in a relatively objective way.

- I didn't get around to charting frametimes as there's only so much one can fit in their spare time, but did add some figures to later tests. They're in the later charts in the thread Papa linked to, and the follow-up tests I did. Knowing what I do now, I'd probably have put a greater focus on frametimes from the beginning, because stuttering issues are what really got my goat in this game (and MSTS before it).

Something that had a definitive impact on stuttering, on my system, was an upgrade from slow mechanical drive to an SSD. Had really been looking forward to showing off the results of that upgrade - did a full suite of tests (including load times) and was putting the finishing touches on the charts, but after all the effort, I suddenly lost interest and motivation. Simply never got around to completing the job and putting it all together in a post.

[Machinist]

Subjective impressions are relevant because is the way that human mind works... In my case I was used to play in the bracket of say average 12-16fps. If I jumped for a while in the same scenario to 20fps and then back to 12fps I felt more of the sttutering thingy. This is why I've locked my fps to 15. Past weekend I installed a new graphic card GTX460 (I had an old 9500GT) and my average range raised to 18-25fps. After couple hours playing in the new fps zone, I causually loaded a heavier scenario and got fps dropped to the old zone (12-16) and then I felt much more the sttuters than when I was used to play all time long in the lower range.

I friend of mine reported with one of his rigs (I5 and HDD) an average 23fps with some minor sttuters from time to time. And with another and better rig (I7 and SSD) an average 18fps and absolutely no sttuter (running as silk, he said). Guess?! He rather play with the SSD rig, even with average 25% lower fps, than with the higher fps HDD machine.

[peterhayes]

Snowstorm
Great points - I entirely agree.
The water also gets a little muddier in the fact that with the 301.xx drivers you can set the maximum fps via vsync say at 1/2 the monitor refresh rate and you can also do this in NVidia Inspector. I seem to see more smoothness but have not carried out any tests to substantiate that.
As Doc says, I found that even though my high end machine has lower frame rates than my "business" machine it is smoother due probably to the fact that RW3 is installed on an SSD. I see the an occasional 1 fps drop but nothing like the 5+ fps drop I see using a conventional HDD.
Regards
PeterH

[micaelcorleone]

Thre is also a pre-load Steam command, so you don't need to check boxes on splash screen everytime you start TS2012:
On Steam | Library | Installed Games list (to the left of screen) | right click on Train Simulator 2012 | left click Properties | General tab | Set launch options (for advanced users only! stated by Steam on the screen!)

Code: Select all

-FPSLimit=30

Do you know some other commands to enter in there? Can the AsyncKeys option also be pre-set this way?

[Machinist]

Hi Timm,

Here they are...

-DisableEAX
-DisableSignals
-DisableSound
-DontUseBlueprintCache - awesome with older engines (especially RW2), but causing problems (artefacts and SBHH) with some rigs combined with modern engines (RW3), once it looks they need cache to load properly (It looks like RSC has changed things without warning, again). Use with caution and by your own risk. Deactivate if you get video issues
-EnableSoundDebugDialogs
-ForceSWMix ( that one should disable anything more than basic playback through your soundcard, if you're having problems )
-LogLocStrings
-LogMate
-NoPlayerTrain
-NoWagonBraking
-QuickStartSteam
-ResetAchievements
-ResetStats
-SetFOV
-SetLogFilters
-ShowControlStateDialog
-UseFastBlueprintCache
-ValidateNetwork
-ValidateSignals
-VerboseAudioDebug
-IgnoreTrackTypes
-lua-debug-messages
-nvperfhud
-FPSLimit=xx xx = a value like 30. The purists will tell you that the fps limit ideally should be an integer of the refresh rate, ie 15 (too low), 20, 30 and 60. That way there are less calculations going on between the cpu and gpu. (note by Peter Hayes)

So far, unfortunately I didn't find the code for AsyncKeys, which I use (ie I have to check) every time I launch the game.

Cheers,
Doc.

[arizonachris]

I just today installed my new EVGA GTX670 2Gb, and it didn't make any difference in RW's performance. My other games, however, are just really awesome now. Price had dropped $200 so I wanted it anyways, what the heck, go for it. Nice card!