Log In
Register Now!
Help
Hi
On some routes at very populated areas (lot of models and textures) i noticing continuous heavy stuttering in OR, while there is more than 1 gbytes of free ram available. It seems that if i turn the camera behind, the game unloads everything that can be seen at the front so when i turn back, it loads it again, and vice versa every time while i see that loading thread starts working too. While these happens the memory usage (according to the game's debug informations) is above 1100 mbytes, but never exceeds 1200mbytes. That's why it seems to mee, there is a limit at 1200mbytes, and the game unloads thinks when the camera turned away, to fit in this limit.
Memory usage limit in OR?
Rate Topic:
#1
- Foreman Of Engines
-
- Group: Private - Open Rails Developer
- Posts: 818
- Joined: 07-October 12
- Gender:Male
- Simulator:OpenRails
-
Country:
Posted 08 April 2013 - 01:27 PM
#2
- Open Rails Developer
-
- Group: Posts: Elite Member
- Posts: 5,514
- Joined: 30-June 10
- Gender:Not Telling
- Simulator:Open Rails
-
Country:
Posted 08 April 2013 - 01:35 PM
There is no unloading when you look away in OR, that only happens when you move away. However, if the GPU doesn't have enough memory (VRAM) to hold everything then the GPU driver + OS work to move things in and out of VRAM so that what's currently visible is in VRAM. This could well affect performance when rotating the camera.
There is also not memory limit set in OR, but it will be limited by the OS and the fact that OR is a 32bit program (that is not marked 3GB-aware). What this means is that OR will never consume more than 2GB itself, usually a bit less due to fragmentation and stuff.
There is also not memory limit set in OR, but it will be limited by the OS and the fact that OR is a 32bit program (that is not marked 3GB-aware). What this means is that OR will never consume more than 2GB itself, usually a bit less due to fragmentation and stuff.
#3
- Foreman Of Engines
-
- Group: Private - Open Rails Developer
- Posts: 818
- Joined: 07-October 12
- Gender:Male
- Simulator:OpenRails
-
Country:
Posted 08 April 2013 - 02:11 PM
that's a strange thing, as in TS2013 where there are bigger textures and higher polycount objects, that 512mbytes vram is enough.
Also the loader process is always working (for minutes, while running at 30 kph). Isn't loaded textures and models cached in ram?
this is how it happens
(note the wheelslip while the reverser in neutral, and the brakes are released)
Also the loader process is always working (for minutes, while running at 30 kph). Isn't loaded textures and models cached in ram?
this is how it happens
(note the wheelslip while the reverser in neutral, and the brakes are released)
#4 Inactive_nyc01_*
Posted 08 April 2013 - 02:17 PM
disc, on 08 April 2013 - 01:27 PM, said:
Hi
On some routes at very populated areas (lot of models and textures) i noticing continuous heavy stuttering in OR, while there is more than 1 gbytes of free ram available. It seems that if i turn the camera behind, the game unloads everything that can be seen at the front so when i turn back, it loads it again, and vice versa every time while i see that loading thread starts working too. While these happens the memory usage (according to the game's debug informations) is above 1100 mbytes, but never exceeds 1200mbytes. That's why it seems to mee, there is a limit at 1200mbytes, and the game unloads thinks when the camera turned away, to fit in this limit.
On some routes at very populated areas (lot of models and textures) i noticing continuous heavy stuttering in OR, while there is more than 1 gbytes of free ram available. It seems that if i turn the camera behind, the game unloads everything that can be seen at the front so when i turn back, it loads it again, and vice versa every time while i see that loading thread starts working too. While these happens the memory usage (according to the game's debug informations) is above 1100 mbytes, but never exceeds 1200mbytes. That's why it seems to mee, there is a limit at 1200mbytes, and the game unloads thinks when the camera turned away, to fit in this limit.
I don't know how much it's going to help but if you're running on a 64-bit OS you could give the 4GB Patch a try -
http://www.ntcore.com/4gb_patch.php
#5 Inactive_nyc01_*
Posted 08 April 2013 - 02:17 PM
James Ross, on 08 April 2013 - 01:35 PM, said:
There is also not memory limit set in OR, but it will be limited by the OS and the fact that OR is a 32bit program (that is not marked 3GB-aware). What this means is that OR will never consume more than 2GB itself, usually a bit less due to fragmentation and stuff.
Is there any particular reason why OR hasn't been marked 3GB aware, are there any side effects to doing this?
Also does OpenRails always detect and use all available dedicated RAM on the GPU?
#6
- Member since Nov. 2003
-
- Group: Status: R.I.P. or just Retired
- Posts: 2,376
- Joined: 22-November 03
- Gender:Male
- Location:SP Pomona Div. MP 520.2
- Simulator:MSTS, OR
-
Country:
Posted 08 April 2013 - 06:14 PM
Any idea is the Large Address Aware (LAA) (see other forum topic) executable would be of any value?
#7
- Open Rails Developer
-
- Group: Private - Open Rails Developer
- Posts: 1,199
- Joined: 04-May 06
- Gender:Male
- Location:Thornton, CO
- Simulator:MSTS - OR
-
Country:
Posted 08 April 2013 - 11:40 PM
disc, on 08 April 2013 - 02:11 PM, said:
that's a strange thing, as in TS2013 where there are bigger textures and higher polycount objects, that 512mbytes vram is enough.
Why do you think the signals and AI don't work - and they wont' fix them?
Robert
#8 Inactive_nyc01_*
Posted 09 April 2013 - 07:21 AM
rdamurphy, on 08 April 2013 - 11:40 PM, said:
Why do you think the signals and AI don't work - and they wont' fix them?
Robert
Robert
How come the signals don't work?
I've been testing a route in TS2013 that Marc Nelson created by using railroad data from the actual sub division and so far he's been able to get accurate signal indications.
As far as the AI is concerned I've never seen anything created by anyone who actually really knew what the AI was supposed to do based on the real world anyway.
#9
- Open Rails Developer
-
- Group: Posts: Elite Member
- Posts: 5,514
- Joined: 30-June 10
- Gender:Not Telling
- Simulator:Open Rails
-
Country:
Posted 09 April 2013 - 08:48 AM
nyc01, on 08 April 2013 - 02:17 PM, said:
Is there any particular reason why OR hasn't been marked 3GB aware, are there any side effects to doing this?
I'm not 100% sure, but I don't think we can do this directly from Visual Studio - at least not for the Express editions without the C++ compiler toolchain.
In general, though, there should be no side effects, but remember the configuration may not be supported by the CLR so there could always be issues. It can't hurt to try, anyway.
However, in this case, the VM is only getting up to ~1100MB so the likelihood of it making any difference here is slim.
nyc01, on 08 April 2013 - 02:17 PM, said:
Also does OpenRails always detect and use all available dedicated RAM on the GPU?
Although we detect the VRAM, we make no use of this in our loading patterns currently. We load everything in the "tile space" (current tile and the 8 around it) and that's it.
There is nothing in OR (or should certainly not be!) that unloads/loads things based on viewing direction. If there is, it's a bug. That does make the video a little puzzling. I believe I have seen reports that ATI cards are more aggressive at unloading "unused" VRAM and are known to often have lots of stuttering when rotating the camera in open-world style games (I've personally seen it with World of Warcraft and it's pretty bad sometimes). While that may not be everything going on here, I'd suspect it of not helping.
If you haven't already, make sure you're on the latest graphics card drivers too, just in case. :rotfl:
#10
- Foreman Of Engines
-
- Group: Private - Open Rails Developer
- Posts: 818
- Joined: 07-October 12
- Gender:Male
- Simulator:OpenRails
-
Country:
Posted 09 April 2013 - 01:16 PM
rdamurphy, on 08 April 2013 - 11:40 PM, said:
Why do you think the signals and AI don't work - and they wont' fix them?
Robert
Robert
What signal, and AI? how comes this here?
#11
- Superintendant
-
- Group: Posts: Elite Member
- Posts: 1,849
- Joined: 25-November 11
- Gender:Male
-
Country:
Posted 09 April 2013 - 01:18 PM
James Ross, on 09 April 2013 - 08:48 AM, said:
I'm not 100% sure, but I don't think we can do this directly from Visual Studio - at least not for the Express editions without the C++ compiler toolchain.
In general, though, there should be no side effects, but remember the configuration may not be supported by the CLR so there could always be issues. It can't hurt to try, anyway.
However, in this case, the VM is only getting up to ~1100MB so the likelihood of it making any difference here is slim.
Although we detect the VRAM, we make no use of this in our loading patterns currently. We load everything in the "tile space" (current tile and the 8 around it) and that's it.
There is nothing in OR (or should certainly not be!) that unloads/loads things based on viewing direction. If there is, it's a bug. That does make the video a little puzzling. I believe I have seen reports that ATI cards are more aggressive at unloading "unused" VRAM and are known to often have lots of stuttering when rotating the camera in open-world style games (I've personally seen it with World of Warcraft and it's pretty bad sometimes). While that may not be everything going on here, I'd suspect it of not helping.
If you haven't already, make sure you're on the latest graphics card drivers too, just in case. :p
In general, though, there should be no side effects, but remember the configuration may not be supported by the CLR so there could always be issues. It can't hurt to try, anyway.
However, in this case, the VM is only getting up to ~1100MB so the likelihood of it making any difference here is slim.
Although we detect the VRAM, we make no use of this in our loading patterns currently. We load everything in the "tile space" (current tile and the 8 around it) and that's it.
There is nothing in OR (or should certainly not be!) that unloads/loads things based on viewing direction. If there is, it's a bug. That does make the video a little puzzling. I believe I have seen reports that ATI cards are more aggressive at unloading "unused" VRAM and are known to often have lots of stuttering when rotating the camera in open-world style games (I've personally seen it with World of Warcraft and it's pretty bad sometimes). While that may not be everything going on here, I'd suspect it of not helping.
If you haven't already, make sure you're on the latest graphics card drivers too, just in case. :p
Hmmmmmmmmmmmmmmmm........... , I will have to give this a test on both my systems. I have long had the feeling that the NVidia card (One system has a NVidia GTX560SOC the other a Radeon 5870) does a good deal better job on OR than the the 5870. As you mentioned "stuttering" is a major problem with the radeon. I will check this though, i have not run the NVidia system much recently as the fans in the case keep giving problems and have yet to fix this issue.
#12 Inactive_nyc01_*
Posted 09 April 2013 - 01:40 PM
James Ross, on 09 April 2013 - 08:48 AM, said:
I'm not 100% sure, but I don't think we can do this directly from Visual Studio - at least not for the Express editions without the C++ compiler toolchain.
In general, though, there should be no side effects, but remember the configuration may not be supported by the CLR so there could always be issues. It can't hurt to try, anyway.
However, in this case, the VM is only getting up to ~1100MB so the likelihood of it making any difference here is slim.
In general, though, there should be no side effects, but remember the configuration may not be supported by the CLR so there could always be issues. It can't hurt to try, anyway.
However, in this case, the VM is only getting up to ~1100MB so the likelihood of it making any difference here is slim.
Thanks for the info.
I gave the 4GB patch a try but it's really hard to see if it's really doing anything on the machine I'm testing on.
It seems that with the draw distance increased (I'm using 4000m) I get slightly more stuttering when flying around fast in the #4 key view, the 4GB patch might be be helping to keep the stuttering down but again it's hard to tell.
I've only tested it on a GTX 680 based system so far, I'll have to give it a try on the AMD 7970 based machine and see if it helps with the AMD stuttering at all.
Quote
Although we detect the VRAM, we make no use of this in our loading patterns currently. We load everything in the "tile space" (current tile and the 8 around it) and that's it.
There is nothing in OR (or should certainly not be!) that unloads/loads things based on viewing direction. If there is, it's a bug. That does make the video a little puzzling.
There is nothing in OR (or should certainly not be!) that unloads/loads things based on viewing direction. If there is, it's a bug. That does make the video a little puzzling.
The reason why I asked is that with a couple of racing sims I run there is an option to load the entire track into the video cards RAM. Obviously this isn't going to work with some 100-200 mile route in a rail operation sim but I didn't know if it could help in other ways with OpenRails.
Quote
I believe I have seen reports that ATI cards are more aggressive at unloading "unused" VRAM and are known to often have lots of stuttering when rotating the camera in open-world style games (I've personally seen it with World of Warcraft and it's pretty bad sometimes). While that may not be everything going on here, I'd suspect it of not helping.
The ATI/AMD stuttering is in a long list of issues I've had with the 5870 and now the 7970 with older DirectX 9 sims/games. For a while now I haven't been able to use super-sampling AA with OpenRails no matter how the options are set up in the Catalyst Control Center. It's like the OpenRails exe is completely ignored.
My experience with the 7970 with DirectX 11 sims/games has been the total opposite, great performance and image quality.
Quote
If you haven't already, make sure you're on the latest graphics card drivers too, just in case. :p
I always keep my test machines up to date with the latest drivers, I'm using the latest 314.22 drivers on the Nvidia based system and the 13.3 beta 3 Catalyst drivers with the AMD 7970.
#13
- Fireman
-
- Group: Posts: Active Member
- Posts: 185
- Joined: 13-December 09
- Gender:Male
-
Country:
Posted 11 April 2013 - 09:31 PM
disc, on 08 April 2013 - 01:27 PM, said:
Hi
On some routes at very populated areas (lot of models and textures) i noticing continuous heavy stuttering in OR, while there is more than 1 gbytes of free ram available. It seems that if i turn the camera behind, the game unloads everything that can be seen at the front so when i turn back, it loads it again, and vice versa every time while i see that loading thread starts working too. While these happens the memory usage (according to the game's debug informations) is above 1100 mbytes, but never exceeds 1200mbytes. That's why it seems to mee, there is a limit at 1200mbytes, and the game unloads thinks when the camera turned away, to fit in this limit.
On some routes at very populated areas (lot of models and textures) i noticing continuous heavy stuttering in OR, while there is more than 1 gbytes of free ram available. It seems that if i turn the camera behind, the game unloads everything that can be seen at the front so when i turn back, it loads it again, and vice versa every time while i see that loading thread starts working too. While these happens the memory usage (according to the game's debug informations) is above 1100 mbytes, but never exceeds 1200mbytes. That's why it seems to mee, there is a limit at 1200mbytes, and the game unloads thinks when the camera turned away, to fit in this limit.
I have the same problem. Basically, the train moves in a jerky motion similar to a slideshow even though the HUD shows framerates in the 30s or 40s. I don't know how much this helps but I've been able to isolate what causes the problem. Whenever I notice this behavior one of the driving aids is on (usually the track monitor, but the other overlay type aids also cause it). If I turn all of the driving aids off, the behavior goes away. Interestingly, if I hit alt-enter to go into window mode, the problem goes away also, even with the driving aids still on. The HUD display, which is just text, doesn't cause this problem. Neither does any other info display which uses only text.
My system specs are Asus A85-F2A85-V PRO motherboard, A10-5800K APU (with built-in HD7660D graphics), 16 GB of DDR3-1600 RAM, Windows 7 Ultimate 64-bit O/S. I have the latest set of Catalyst drivers.
On a side note, I'm back into trainsimming after updating my system last year. My prior system couldn't run the latest releases of OR well so I took an extended hiatus from it. If there's anything I can do to resume helping the development team I would be glad to.
#14
- Conductor
-
- Group: Posts: Contributing Member
- Posts: 426
- Joined: 24-July 08
- Gender:Male
-
Country:
Posted 12 April 2013 - 04:46 AM
#15
- Open Rails Developer
-
- Group: Posts: Elite Member
- Posts: 5,514
- Joined: 30-June 10
- Gender:Not Telling
- Simulator:Open Rails
-
Country:
Posted 18 April 2013 - 12:58 PM
jtr1962, on 11 April 2013 - 09:31 PM, said:
Whenever I notice this behavior one of the driving aids is on (usually the track monitor, but the other overlay type aids also cause it). If I turn all of the driving aids off, the behavior goes away. Interestingly, if I hit alt-enter to go into window mode, the problem goes away also, even with the driving aids still on. The HUD display, which is just text, doesn't cause this problem. Neither does any other info display which uses only text.
Do you have the "Use glass on in-game windows" option enabled? There's a potential for that to cause issues like this, so try turning it off.