[QJ-6811]

The last period I have occasional problems that OR stops (falls away). OK, I use a (detailed) route and Timetable mode, so an ever higher loading and memory usage. I also thought that I had now reached the limits? (the XNA 4Gb maximum).
However, when I look through task manager, I see:
- total memory uses 3.4 to 5.1 Gb of the 7.9 Gb installed.
- GPU memory 4Gb available, about 0.5Gb in use.
- processor CPU load about 30% (GPU 70-100%)

When I view the process "open rails activity runner" I see a memory consumption of about 2.1 to 2.8 Gb.

My conclusion is that it is not a "memory problem" (or processor problem) because there is still enough space available. (IOW, which part is actually the maximum XNA 4Gb, I assume what you see in the process "open rails activity runner")

Furthermore, I see that on my W10 PC is installed "Microsoft XNA framework redistributable 3.1"
On the Microsoft site I see 4.0 as the last release
Is it a recommendation for OR to update this?

Sorry, in this part of Windows and Graphics I am a bit lost and especially careful, to keep it working. :sweatingbullets:
I hope someone can help or advise me with this.

[Mike B]

 QJ-6811, on 29 August 2018 - 03:51 AM, said:

The last period I have occasional problems that OR stops (falls away). OK, I use a (detailed) route and Timetable mode, so an ever higher loading and memory usage. I also thought that I had now reached the limits? (the XNA 4Gb maximum).
However, when I look through task manager, I see:
- total memory uses 3.4 to 5.1 Gb of the 7.9 Gb installed.
- GPU memory 4Gb available, about 0.5Gb in use.
- processor CPU load about 30% (GPU 70-100%)

When I view the process "open rails activity runner" I see a memory consumption of about 2.1 to 2.8 Gb.

My conclusion is that it is not a "memory problem" (or processor problem) because there is still enough space available. (IOW, which part is actually the maximum XNA 4Gb, I assume what you see in the process "open rails activity runner")

Furthermore, I see that on my W10 PC is installed "Microsoft XNA framework redistributable 3.1"
On the Microsoft site I see 4.0 as the last release
Is it a recommendation for OR to update this?

Sorry, in this part of Windows and Graphics I am a bit lost and especially careful, to keep it working. :sweatingbullets:
I hope someone can help or advise me with this.

Some additional info as an attachment
MoreInfo.zip

In terms of hardware, you certainly have enough CPU and memory. As I'm not an OR developer (only a user), I can't speak to other needs/issues, but the only thing I might add to your general system to help might be to convert the disk used for Open Rails and routes/trainsets to a solid-state disk rather than a conventional hard disk. SSDs are still a little expensive compared to similar-sized hard disks, but have become more affordable recently. Open Rails, and especially MS Train Simulator, need to read data from the disk frequently during operation, and the SSD will speed up that process a lot. A SSD for the system disk will also speed your computer's startup and (if necessary - your computer looks like it has enough memory to minimize swapping in normal use) virtual memory swap speeds a lot.

I've used a SSD for the system and software disk in an old computer (with adequate memory) for some time. It and a reasonable nVidia graphics card (750ti, probably not as good as yours, certainly with regard to graphics memory) were what made Open Rails enjoyable to use, rather than just running at some barely-playable speed. OR in my computer still sometimes pauses, but not often after it's been running a little while; I attribute the general smoothness of running to the SSD.

[Csantucci]

A screenshot with the Debug HUD could be useful.

The logfile is full of instances of the same log warning, that is:

Warning: System.IO.FileLoadException: C:\Roger\3ORTS\Trains\trainset\SNCF_X2400\SNCF_X2419_ORTS_VL.eng ---> System.ArgumentException: Er is al een item met dezelfde sleutel toegevoegd.

If I understand well this dutch wording it means that an item with the same key has already been added, or something similar.

The warnings alternatively refer to SNCF_X2419_ORTS_VL.eng and to SNCF_X2404_ORTS_VL.eng, and are generated by the sound process. You could first try to remove the sound references from these two .eng files and check if something changes.

[QJ-6811]

 Mike B, on 29 August 2018 - 08:38 AM, said:

In terms of hardware, you certainly have enough CPU and memory.
convert the disk used for Open Rails and routes/trainsets to a solid-state disk rather than a conventional hard disk.

Hi Mike B,

Thank you for the reply. I am glad that I get a "confirmation" that it is not a hardware / memory problem. :sweatingbullets:
Concerning SSD is a good idea, it will in any case work faster and because of this less chance of getting "delay", especially when loading all files, such as objects and sounds.

I already had a similar situation, that "anti-malware" (virus scanner) was very busy at times during the ride. (Mb/s on hard disk) This probably when you reach a new area in the route (world file) and thus many new objects / files are loaded. In the last two days the directory of OR is set to "excluded". It does not prevent the above-mentioned problem, but gives at least less "delay". (the sometimes briefly haltingly image )

Actually also by removing all "unnecessary programs" (mainly preventing background actions) and turning off internet (Ethernet and Wi-Fi) ...! also gives positive results. :)

OK, find out what is possible. I think the simplest thing is to leave W10 on hard disk and everything from OR on an SSD. (?)

Regards,
Roger

[QJ-6811]

 Csantucci, on 29 August 2018 - 12:54 PM, said:

A screenshot with the Debug HUD could be useful.

The warnings alternatively refer to SNCF_X2419_ORTS_VL.eng and to SNCF_X2404_ORTS_VL.eng, and are generated by the sound process. You could first try to remove the sound references from these two .eng files and check if something changes.

Hi Csantucci,

Thanks for your information. Do you mean the F5 HUD with "Debug HUD"? This does not make sense after my opinion, because if things go wrong, OR falls away. (so no more image, no error as pop-up, etc.) IOW, back on W10 desktop with only the "Open Rails menu" window open.



I also noticed the error messages with "X2419" and "X2404", only the reason for "an item with the same key has already been added" was unknown to me. I checked the .ENG files several times, but did not find anything yet. If you think this has to do with the sound process, then it confirms that I had also this feeling. :)

Only ........... the curious thing is ....

In the focus on one crash, I came to the conclusion that it had to do with a "moment of time". Afterwards discovered that it went wrong when a new AI-train was placed / started on the route at that moment. Various tests (viewing range, objects, loading, etc.) had no effect until I replaced the AI train with a different train / consist. (This AI train has been in use for some time without any problems, and is still used at other time moments ....?)

When tested with two other 'crashes', both with different Player trains and time moments, these "X2419 + X2404" error messages always occur. Now no start of this AI train, but the "X2419 + X2404" train is nearby (+/- 5-8 km) of the Player train. The most remarkable thing is that the times of the crash are when this "X2419 + X2404" train leaves/start as an AI-train from a station ??? :ko2:
(after stop at red signal or waiting point)

Already tested by the same Player train and Timetable, only with replacement of "X2419 + X2404" train with another train. OK, no crash and problems at the moment. But I saw in the log file some other "not loaded .wav files" I certainly know exist?

I suspect that there is a kind of "memory problem", perhaps not with "Objects and trains" (.ACE files), but more focused on the sound files. (Loading of .WAV files).
Could it be (as you can see with ALT-S "Sound debug") that there is a maximum memory? (WAV Cache, sources, etc.)
Furthermore, I am also busy to check a lot of .SMS files, sometimes due to error messages from the log-file, or simply to count the number of 'Streams' and fill it in at 'Streams (x'.
Maybe that "missing or hanging sound files" still give problems at certain times and busy conditions?

But then, how can you "capture" this ....? :(


Regards, Roger

[roeter]

First of all, it's not crash, but a time-out. The main program checks all underlying processes which have to 'ping' within a certain interval, otherwise it is assumed to 'hang'. In this case, the sound process 'hangs' when it tries to process all triggers in all loaded files.
Whether it is indeed a real endless loop or whether it just take too long is not clear, that distinction can not be made.

The cause of this problem is probably indeed the many sound files it has (perhaps partially) loaded.

More curious, however, is the occurence of all the 'load failure' errors. Looking through the code it could well be that once such an error occurs, it will keep in occuring as long as that engine is within range. The trigger to load the sound-files occurs when an engine is loaded. An engine is only loaded if there is not yet such an engine available in memory already. So, when the engine is still near, it is available in memory and should therefor not be loaded again. But through some error the first time the engine is loaded, the loading failed and the engine is not properly included in the list of available engines. But the sound files apparantly have been loaded. So, next update, the engine is loaded again, and now the loading crashes on the sound file load error, and again the engine is not properly set in the list of availabel engines. And this happens again the next update, and the next ect. In the included logfile, it happens around 180 times for both engines before the program fails.

So the main question is : why does the engine load process fails the first time for both these engines?
I cannot find any clue to this in the logfile, so this remains a mystery. Perhaps someone with more extensive knowledge of the viewer process and in particular on loading of engine within this process can shed some more light on this.

Regards,
Rob Roeterdink

[QJ-6811]

Rob, thank you for the explanation. Today again some tests done, and can indeed conclude that you also indicate that it is not the 'time moment' of departure of the AI-X2400 train, but if it is within reach. This time a little faster drove and went wrong now 15 sec. before departure of the X2400. :)
A second test confirms this too, when I had modified my player train (freight train) with a passenger train, so I could now drive faster to 100km/h. Equal error at the same distance, of course now a few minutes earlier.

A test done with the X2400 train with other soundfiles, this also went well.

With the focus on the X2400 .ENG and .SMS files the following:
The X2400 has 4x sound.SMS files, the usual ENG.sms and CAB.sms, but also PAS.sms (passenger view) and TCS-France.sms

- X2400 with ENG + CAB, no problem, memory within range about 2Gb
- X2400 with ENG + CAB + PAS, no problem, memory within range about 2Gb
- X2400 with ENG + CAB + TCS, Crash (sorry, time-out), memory within range about 2.4Gb !!

The most incomprehensible at the moment, the TCS-France.sms consists of 2 wav-files (cabin signaling) of each about 25Kb. (and thus already causes an increase in memory of almost 400Mb ..... ?)
Also, why does TCS (or a TCS script) affect an AI train, and even better is that my options include "disable TCS scripts".

OK, eventually found the causer. As far as I am concerned I delete all TCS rules from the .ENG and .SMS files (I do not use it anyway ...), but it remains strange. Must still test whether other situations now remain good, but that takes much more time ....

If I have to test something, I can easily mimic it, then I'll hear it.

Thank you all for the help.

Regards/Roger


(My test with TCS.sms include: [attachment=92241:Test06 Train7 X2400 -TCS.zip] )

[Csantucci]

Where can the trainset be downloaded?

[QJ-6811]

The original trainsets are here: http://www.bb25187.eu/
( http://www.bb25187.e...MatRoulant.html )


I have put a download link in your email, which I use. (slightly adjusted)


regards/Roger

[QJ-6811]

An small update concerns my experiences.
The "X2400" problem does not occur anymore after I have removed the "TCS settings / sound", also no longer in my other 'crash situation'. Only the why? It remains a strange situation, because these trainsets in the timetable driving 6x a day up and down in the route, so there are plenty of other times with other Player trains that meet each other .... (?)

My '4th crash situation' has now been extensively tested. Again, it appears that the Player train is in range with another AI train. In this situation, two AI trains + Player train, which all happen to be "extended", thus require extra memory (.ACE and .wav consumption). From the error messages and log-file you see here that it goes wrong on "no memory more" / "loading, ACE not possible", etc.

A lot of adjustments made. (errors mapped in .ENG and .SMS files / unnecessary services stopped / swap-memory extended (1.5 x Gb) / Priority increased for Open Rails - Laa / Remove all trains-TCS settings, etc.) At this moment I can achieve a maximum Memory for OpenRails up to 3.3 Gb, even once to 3.55Gb succeeded! :)
OK, apparently this is the limit? Resolved my '4th crash situation' situation by creating new consists / .ENG, so shorter train, no use 3D cab, etc., so that these trains stay just together under the total memory 3.2Gb ..... (hummm). :ko2:

My only question remains, where are my remaining "500-800Mb" memory, if XNA should be able to process 4Gb?
I myself suspect that Windows10 is not as smart as they claim, and consume memory from other applications in this section, so that 4Gb can never be achieved. (???)
Or does anyone know how, is there still a smart way to reserve "4Gb" 100% only for XNA / Openrails?


(For me at this moment solved, no more problems, but am still curious / interested if maximum range is still possible .... :D )

Regards/Roger