[roeter]

Hello,

lately I noticed something very peculiar with regards to FPS rates.
When OR is just started, I get a steady FPS of just above 60.
But after some minutes, at the same location without actually moving, the FPS starts to drop markedly.

The pictures tell the story.
The first is taken just after starting - camera position 8, little scenery (route is still under construction), and a handfull of AI in the area. FPS is quite steady around 65.
The second picture is about 15 mins. later. Nothing has moved - player train has remained steady, no camera switches or movement etc. The AI trains in the area are part of a short local suburban shuttle service, so the total number of AI trains is about constant.
Yet the FPS is now varying a lot, but it hardly reaches over 45 and has dips as low as 25.
Total memory has hardly changed, what is most notable is the high number of "GC".
What is this and what could be causing this steady drop and increased fluctuation in FPS?




Regards,
Rob Roeterdink

[James Ross]

 roeter, on 03 April 2014 - 05:13 AM, said:

Total memory has hardly changed, what is most notable is the high number of "GC".
What is this and what could be causing this steady drop and increased fluctuation in FPS?

The GC numbers look perfectly normal to me and they're not occurring more frequently in the graph.

What does stand out though is the step-changes in frame time graph. It seems to be, periodically but not completely consistently, jumping up to 1.2-1.5x the frame time or worse. It would be interesting to know if the render primitives values change much during this time from their apparently-normal 4.6k.

My guess-in-the-dark is that the GPU is powersaving itself because it thinks (mistakenly) not much it going on. I don't know if this is a desktop or laptop, but I am pretty sure such behaviour occurs in the mobile Intel GPUs.

Have you tried the same test on older versions and gotten any idea if/when this started?

[That Genset Foamer]

 James Ross, on 03 April 2014 - 05:30 AM, said:

The GC numbers look perfectly normal to me and they're not occurring more frequently in the graph.

What does stand out though is the step-changes in frame time graph. It seems to be, periodically but not completely consistently, jumping up to 1.2-1.5x the frame time or worse. It would be interesting to know if the render primitives values change much during this time from their apparently-normal 4.6k.

My guess-in-the-dark is that the GPU is powersaving itself because it thinks (mistakenly) not much it going on. I don't know if this is a desktop or laptop, but I am pretty sure such behaviour occurs in the mobile Intel GPUs.

Have you tried the same test on older versions and gotten any idea if/when this started?

I've experienced this notable FPS decay happen ever since using a version which came out around 10/2 or thereabouts. I'm using a Dell Inspiron N4110 with W7-SP1 installed and using the default Intel graphics stuff.

[James Ross]

 That Genset Foamer, on 03 April 2014 - 09:19 AM, said:

I've experienced this notable FPS decay happen ever since using a version which came out around 10/2 or thereabouts. I'm using a Dell Inspiron N4110 with W7-SP1 installed and using the default Intel graphics stuff.

Curious. Would you mind checking a few versions from http://james-ross.co...jects/or/builds to see when it might have started? I would recommend X2012 as a possible "before" and X2021 as a possible "after". If you have the time to figure out exactly which build it started in (use a binary search to save time) that would be fantastic. (The same goes for Rob, since a binary search of every version back to X1889 will only need 8 or 9 builds checking.)

To be clear, I have not observed any of this kind of behaviour, on my Windows 8.1 + NVIDIA GPU system.

[roeter]

I think the problem has been around for quite some time now, but I'll try and see if I can find a change somewhere going through the old versions. Might take a few days, though, due to other work.
By the way, my system is a laptop, ACER 771G, Intel i7-3610QM (8 core), Windows 7 tries to control it all and an NVIDIA GT 650M makes it all visible.
I also checked on the render primitives values but these hardly change between the 'peaks' and 'valleys'.

Regards,
Rob Roeterdink

[engmod]

For me the FPS problem is here:-

http://www.elvastowe...-track-monitor/

cheers
Derek

[rdamurphy]

More than likely, James is right. Either the CPU or GPU is throttling down (feature of all i-series CPU's) since in reality there's no reason for them to maintain FPS when nothing is happening. So, do the FPS go back up when you hit a key or move the train?

My i7 throttles from .799G to 4.4G (Turbo Mode) depending on load... I believe, I'd have to check my Radeon 7790 goes from about .3 G to 1.075G - depending on load.

I would suggest running CPU-Z and GPU-Z with OR in a windowed mode to see the reactions of the processing units.

You could also try changing your Power Options in Windows Control Panel, but use caution, this could cause overheating since your CPU will no longer throttle back with inactivity.

Robert

[rdamurphy]

Also, just curious, you mentioned AI trains - are the FPS drops coinciding with "spawning" of AI trains? You can use Shift-F5 to scroll through the debug screens to the one that shows Player and AI trains...

Robert

[James Ross]

 rdamurphy, on 03 April 2014 - 05:22 PM, said:

I would suggest running CPU-Z and GPU-Z with OR in a windowed mode to see the reactions of the processing units.

I don't think the CPU is doing it, since the process times are the same, but I'd certainly suggest running GPU-Z and checking if the GPU clock speeds are changing at the same times as the frame rates drop/restore. (Thanks for reminding me about these tools!)

[rdamurphy]

I checked with CPU-Z and saw no drop, OTOH, after 5 minutes of complete inactivity, I had no drops in FPS either. I have my screensaver set for 5 minutes.

I guess my question is do framerates go back up when something changes, such as keypress or mouse movement?

Robert