Open GL - out of memory error.

[StuartC]

Im getting this problem more and more frequently, and at increasingly shorter intervals.

"Warning: Detected OpenGL error 'out of memory ' at after RenderBin::draw(..)"

I have reinstalled FG 2.8.1, reinstalled my video drivers for my Nvidea 550TI card and its actually getting worse.

Help !

[Hooray]

You need to tell us a little more about the problem and the circumstances (when it appears), including the settings that seem to have an effect (or not), such as aircraft/airport, rendering related stuff.
Please see the red box at the top of the page.

Obviously, anything directly related to rendering (especially shaders) would seem pretty relevant here.
Once you do a forums earch for "out of memory error", you'll see that this has occurred a number of times in the past (for some people/configurations).
During the previous release cycle, this was almost exclusively related to Stuart's new random buildings code, however this has been recently improved/optimized, so that this should not be the cause here anymore.
OTOH, if the problem disappears without random buildings enabled, that would seem like a good hint.

Finally, it would be great if you could use the issue tracker to file an issue report, so that core developers can have a look - especially because of the upcoming release, so fixing such issues prior the release would be awesome obviously.

Your card/GPU seem pretty powerful actually. On the other hand, you didn't post any useful information regarding your startup/runtime settings. So while your GPU should definiely be able to 2.8.1 - it really all boils down to your settings (resolution, bpp, shaders, rembrandt etc) - using the "right" combination of settings, you can bring even the latest GPUs down to their knees easily - especially when using Rembrandt + maximum shaders + advanced weather ...

[StuartC]

I have been through all the options Trees, building, normal/advanced weather, rembrandt and shader slider values, on and off, makes no difference at all.
The crash produces the text I printed above, no extra hints as to who what or where this happens.
Last night it started off happening after a very short time in flight, but by the end of the night, it went a full 20 min without a crash. Same environment, location, aircraft, everything.

Ths is the config file from last night

Code: Select all

; FLTK preferences file format 1.0
; vendor: flightgear.org
; application: fgrun

[.]

fg_exe_init:g:\\games\\FlightGear-28\\bin\\Win32\\fgfs.exe
fg_exe:g:\\games\\FlightGear-28\\bin\\Win32\\fgfs.exe
fg_root_init:g:\\games\\FlightGear-28\\data
fg_root:g:\\games\\FlightGear-28\\data
fg_scenery_init:g:\\games\\FlightGear-28\\data\\Scenery;g:\\games\\FlightGea
+r-28\\scenery;g:\\games\\FlightGear\\terrasync
fg_scenery:g:\\games\\FlightGear\\terrasync;g:\\games\\FlightGear-28\\d
+ata\\Scenery;g:\\games\\FlightGear-28\\scenery
ts_dir:1
ts_exe_init:g:\\games\\FlightGear-28\\bin\\Win32\\terrasync.exe
ts_exe:g:\\games\\FlightGear-28\\bin\\Win32\\terrasync.exe
ts_dir_init:3
x:549
y:312
width:1088
height:613
fg_aircraft:g:/games/FlightGear/FGUKaircraft;g:/games/FlightGear/Dave;g:
+/games/FlightGear-28/data/Aircraft
aircraft:T38-yasim
airport:EGNL
airport-name:WALNEY ISLAND
carrier:
parkpos:
runway:30
clouds3d:1
horizon_effect:1
enhanced_lighting:1
specular_highlight:1
ai_models:1
time_of_day_value:noon
time-match-real:0
time-match-local:0
start-date-sys:0
start-date-gmt:0
start-date-lat:0
time_of_day:1
terrasync:1
callsign:WF01
multiplay1:out,10,mpserver08.flightgear.org,5000
multiplay2:in,10,,5000
fullscreen:1
geometry:1680x1050
fetch_real_weather:1
show_cmd_line:0
browser:
control:joystick
lang:
config:
game_mode:0
splash_screen:1
intro_music:1
mouse_pointer:0
random_objects:0
random_trees:1
panel:1
sound:1
hud:0
antialias_hud:0
hud_3d:1
auto_coordination:0
distance_attenuation:1
failure:0
failure_pitot:0
failure_static:0
failure_system:0
failure_vacuum:0
fdm:jsb
no_trim:0
model_hz:120
speed:1
on_ground:1
in_air:0
wind_heading:0
wind_speed:0
turbulence:0
ceiling:
random_wind:0
freeze:0
fuel_freeze:0
clock_freeze:0
lon:
lat:
altitude:
heading:0
roll:0
pitch:0
vc:
uBody:
vBody:
wBody:
vor:
ndb:
fix:
offset_distance:
offset_azimuth:
glideslope:
clouds:1
skyblend:1
textures:1
wireframe:0
fog:fastest
shading:smooth
visibility:120
visibility-units:miles
view-offset:
bpp:32
fov:60
texture-filtering:1
time-offset:
start-date-sys-value:
start-date-gmt-value:
start-date-lat-value:
httpd:0
props:0
jpg-httpd:0
proxy:
io-count:0
property-count:2
env-count:0
log-level:alert
trace-read-count:0
trace-write-count:0
nav1:
nav2:
adf:
dme:disabled
layer-0-elevation-ft:0
layer-0-thickness-ft:0
layer-0-coverage:clear
layer-0-transition-ft:0
layer-0-span-m:0
layer-1-elevation-ft:0
layer-1-thickness-ft:0
layer-1-coverage:clear
layer-1-transition-ft:0
layer-1-span-m:0
layer-2-elevation-ft:0
layer-2-thickness-ft:0
layer-2-coverage:clear
layer-2-transition-ft:0
layer-2-span-m:0
layer-3-elevation-ft:0
layer-3-thickness-ft:0
layer-3-coverage:clear
layer-3-transition-ft:0
layer-3-span-m:0
layer-4-elevation-ft:0
layer-4-thickness-ft:0
layer-4-coverage:clear
layer-4-transition-ft:0
layer-4-span-m:0
season:summer
property-item-1:sim/rendering/multi-sample-buffers=true
property-item-2:sim/rendering/multi-samples=4
scenario-count:0
materials-file:
scenario-item-1:droptank_hunter_demo
frame_rate_limiter:1
frame_rate_limiter_value:30
version:20800
ai_traffic:0
scenario-item-2:eisenhower_demo
scenario-item-3:foch_demo
scenario-item-4:ICE3_demo
scenario-item-5:KRHV_towing_demo
scenario-item-6:KSFO_AirTrain
scenario-item-7:lead_aircraft
scenario-item-8:load_demo
scenario-item-9:nimitz_demo
scenario-item-10:PAVictoria_demo
scenario-item-11:railway_demo
scenario-item-12:refueling_demo
scenario-item-13:refueling_demo_1
scenario-item-14:refueling_demo_2
scenario-item-15:sanantonio_demo
scenario-item-16:seahawk_wingman_demo
scenario-item-17:ship_demo
scenario-item-18:steam_train_demo
scenario-item-19:thermal_demo
scenario-item-20:vinson_demo
scenario-item-21:wingman2_demo
scenario-item-22:wingman_b29_2_demo
scenario-item-23:wingman_b29_3_demo
scenario-item-24:wingman_b29_2_demo
scenario-item-25:wingman_b29_3_demo
scenario-item-26:wingman_b29_demo
scenario-item-27:wingman_demo


Reest of the system spec :-
Operating System: Windows Vista™ Business (6.0, Build 6002) Service Pack 2 (6002.vistasp2_gdr.120402-0336)
Language: English (Regional Setting: English)
System Manufacturer: GBT___
System Model: GBTUACPI
BIOS: Award Modular BIOS v6.00PG
Processor: AMD Phenom(tm) 8650 Triple-Core Processor (3 CPUs), ~2.3GHz
Memory: 3198MB RAM
Page File: 2349MB used, 5565MB available
Windows Dir: C:\Windows
DirectX Version: DirectX 11
DX Setup Parameters: Not found
DxDiag Version: 7.00.6002.18107 32bit Unicode

Display Devices
---------------
Card name: NVIDIA GeForce GTX 550 Ti
Manufacturer: NVIDIA
Chip type: GeForce GTX 550 Ti
DAC type: Integrated RAMDAC
Device Key: Enum\PCI\VEN_10DE&DEV_1244&SUBSYS_15563842&REV_A1
Display Memory: 2335 MB
Dedicated Memory: 992 MB
Shared Memory: 1343 MB
Current Mode: 1680 x 1050 (32 bit) (59Hz)
Monitor: BenQ FP222Wa
Driver Name: nvd3dum.dll,nvwgf2um.dll,nvwgf2um.dll

-------------
Sound Devices - Incidentally, MP soupd will cause a random crash as well with this sound config.
-------------
Description: Speakers (Realtek High Definition Audio)
Default Sound Playback: Yes
Default Voice Playback: Yes
Hardware ID: HDAUDIO\FUNC_01&VEN_10EC&DEV_0885&SUBSYS_1458A002&REV_1001
Manufacturer ID: 1
Product ID: 100
Type: WDM
Driver Name: RTKVHDA.sys
Driver Version: 6.00.0001.5473 (English)
Driver Attributes: Final Retail
WHQL Logo'd: Yes
Date and Size: 8/28/2007 09:46:38, 1951000 bytes
Other Files:
Driver Provider: Realtek Semiconductor Corp.
HW Accel Level: Basic
Cap Flags: 0xF1F
Min/Max Sample Rate: 100, 200000
Static/Strm HW Mix Bufs: 1, 0
Static/Strm HW 3D Bufs: 0, 0
HW Memory: 0
Voice Management: No
EAX(tm) 2.0 Listen/Src: No, No
I3DL2(tm) Listen/Src: No, No
Sensaura(tm) ZoomFX(tm): No

Description: Realtek Digital Output (Realtek High Definition Audio)
Default Sound Playback: No
Default Voice Playback: No
Hardware ID: HDAUDIO\FUNC_01&VEN_10EC&DEV_0885&SUBSYS_1458A002&REV_1001
Manufacturer ID: 1
Product ID: 100
Type: WDM
Driver Name: RTKVHDA.sys
Driver Version: 6.00.0001.5473 (English)
Driver Attributes: Final Retail
WHQL Logo'd: Yes
Date and Size: 8/28/2007 09:46:38, 1951000 bytes
Other Files:
Driver Provider: Realtek Semiconductor Corp.
HW Accel Level: Basic
Cap Flags: 0xF1F
Min/Max Sample Rate: 100, 200000
Static/Strm HW Mix Bufs: 1, 0
Static/Strm HW 3D Bufs: 0, 0
HW Memory: 0
Voice Management: No
EAX(tm) 2.0 Listen/Src: No, No
I3DL2(tm) Listen/Src: No, No
Sensaura(tm) ZoomFX(tm): No

[ot-666]

The error is related to your fgfs running out of memory (but not your gpu memory)!

This will happen with detailed aircraft or detailed scenery.
Custom scenery like the one from papillon81 repo will cause this happen realy quick.
Using a detailed plane like the 787 from omega makes it worse.
(will crash fg on startup at LOWI for example / or cause scnery and some textures to be not loaded and leads to a crash a couple of minutes later)
This happens without any eyecandy activated!

If you have a 64bit windows you can prevent the error / crash by using a fgfs 64bit version.

Oliver

[StuartC]

I only have 32 bit windows.
So, its a hard coded problem with FG thats causing this??????????

[ot-666]

I only have 32 bit windows.
So, its a hard coded problem with FG thats causing this??????????

Depending on the scenery view distance and and diabling mp aircraft that use a lot of memory, trees and buildings you can make it work.

This error was discussed at the forum a while ago:http://www.flightgear.org/forums/viewtopic.php?f=68&t=17114#p163132

Oliver

[StuartC]

This is not a good situation then. How can we push the quality of objects, aircraft it terrain in FG when this problem exists. Is it something that the core dev team are looking at or not????????

[Thorsten]

My understanding is that in a 32bit OS, a running process cannot have more than ~4 GB of memory associated with it. This has nothing to do with FG or Windows and is the same in a 32bit Linux installation. There is hence no fix other than using a 64bit OS or to run FG with memory-conserving restrictions. You can't beat the numbers - if you want to develop a scene beyond a certain complexity, you run into this problem.

[Gijs]

I remember something about loaded scenery tiles not being freed from memory once they are loaded. So if you fly some distance, memory usage keeps on increasing until you reach the limit. On 32 bits this limit is much lower (and thus earlier reached) than on 64 bits.

Not sure if this is still the case though...

[Johan G]

My understanding is that in a 32bit OS, a running process cannot have more than ~4 GB of memory associated with it...

So then, in a far future, when FlightGear have been made more modular with subsystems being able to run as separate processes/threads, this could potentially be less of an issue?

[Thorsten]

@Gijs: I think the major leaks of that kid are plugged, but I guess there's some memory issues. Anyway, I've been flying with a 32bit system until this weekend, and I would argue that it's possible to get a nice experience out of it.

So then, in a far future, when FlightGear have been made more modular with subsystems being able to run as separate processes/threads, this could potentially be less of an issue?

I think in a far future, 32bit computers will have died out. My guess is that this'll happen before a modular FG solves any memory problems. Besides, I don't see it - if it's in the scene, it needs to go into the rendering pipeline. So vertex coordinates and textures need to be available in memory. You can't feed the rendering pipeline from many different sources (?) - so if you have million trees in the scene, you have a million coordinates for the trees, which take memory. And so on.

[Johan G]

...if you have million trees in the scene, you have a million coordinates for the trees, which take memory. And so on.

Ah, I see.

[Hooray]

It's true:

Memory: 3198MB RAM
Page File: 2349MB used, 5565MB available

In 32 bit mode (regardless of the OS being used), applications like FlightGear cannot (normally) address memory beyond 32 bit: http://en.wikipedia.org/wiki/32-bit

This is not a good situation then. How can we push the quality of objects, aircraft it terrain in FG when this problem exists. Is it something that the core dev team are looking at or not????????

A while ago, I talked to Stuart about it, so core developers are aware of the issue. The real problem is the way memory is simply just allocated in FG, without using a real memory manager subsystem that's aware of object lifetime and that could de-allocate memory. For more details, please see this thread: viewtopic.php?f=5&t=16083&p=164936&hilit=swapping#p165125

However, this should probably be better moved to the issue tracker - the real solution is for FG to not just keep on allocating more memory than the process/OS can handle, but instead keep track of the amount of memory, write it to the property tree and let subsystems free resources more aggressively as needed.

At the moment, fgfs is still pretty dumb about allocating memory - even for the rare cases where we are using smart pointers. See for example Gijs posting:

I remember something about loaded scenery tiles not being freed from memory once they are loaded. So if you fly some distance, memory usage keeps on increasing until you reach the limit. On 32 bits this limit is much lower (and thus earlier reached) than on 64 bits.

Not sure if this is still the case though...

And there are more of such examples, where memory is never freed and basically leaking - some of this was recently discussed by Tim and Mathias.

There are far better memory allocation schemes available. Even using a garbage collector like the BoehmGC would be better than what we're currently doing.

But this should be really moved to the issue tracker in the form of a feature request: http://flightgear-bugs.googlecode.com/

My understanding is that in a 32bit OS, a running process cannot have more than ~4 GB of memory associated with it.

This is generally true, but there's one exception: "PAE" see: http://en.wikipedia.org/wiki/Physical_Address_Extension
On Linux, PAE makes it possible to access 64bit of memory on a 32bit system.

So then, in a far future, when FlightGear have been made more modular with subsystems being able to run as separate processes/threads, this could potentially be less of an issue?

Sort of, because each subsystem running in a dedicated process could then have its own area of "private" memory that could be up to 4gb in size.
On the other hand, Thorsten is right: This is still going to take a long time (probably another 2 years at least), and only a handful of subsystems will be ported then probably, such as the AI traffic subsystem and other "memory hogs". Mathias has recently committed a prototype of a HLA-based AI traffic system which will probably become the foundation for related efforts.

Also, Thorsten is right when he said that in another 2-3 years, it will be rather difficult to find people still running 32 OSes or even 32 bit computers. Especially people who are hard core gamers and flight sim enthusiasts.
And obviously, all rendering related objects need to be in a allocated/available in a single process like Thorsten said.

I only have 32 bit windows.

On 32 bit windows, there's another issue: http://en.wikipedia.org/wiki/3_GB_barrier
Which is often the reason why the remaining gb of RAM is used as shared memory for graphics cards.

So, its a hard coded problem with FG thats causing this??????????

It's not a FG-related problem at all - FG just makes it obvious, because it tries to use more memory than can be handled by your current OS.

However, it seems you already have a 64bit processor that can address more RAM:

Processor: AMD Phenom(tm) 8650 Triple-Core Processor (3 CPUs), ~2.3GHz

See:
http://en.wikipedia.org/wiki/AMD_Phenom
http://products.amd.com/pages/DesktopCP ... eSupport=1
http://www.sharkyforums.com/showthread. ... nom-64-Bit

In other words, you can easily install a real 64 bit OS in parallel to your old OS and use it for FG. Note, that you'll probably want to upgrade your RAM, too.
Otherwise, FG would still use "swap" memory - i.e. virtual memory taken from your HD.

[StuartC]

I have been looking at OS upgrades but according to Microsoft, I cant upgrade to a 64bit OS. It needs a clean install. As im running a business and I am likely to loose a number of software packages due to misplaced licence numbers I cant take the risk of a fresh install.
So, im still stuffed.

[Hooray]

As im running a business and I am likely to loose a number of software packages due to misplaced licence numbers I cant take the risk of a fresh install. So, im still stuffed.

You should still be able to install another OS in parallel to your existing/old OS. That's how you can have even 10+ different operating systems on a single computer. You would then be asked during boot-up to chose one of the installed systems. Obviously, that would require sufficient free disk space to install an additional OS next to Windows.

For testing purposes, you could simply download a free 64 bit Linux image such as Lubuntu: http://lubuntu.net/
During installation, it will then detect your existing Windows installation and ask you if you want to erase everything on the old disk (NO!!!!) or simply install Linux next to the old OS. If you have previously partioned/formatted a hard disk (even just a flash disk or USB drive), the whole process will seem fairly familiar. Nevertheless, it's always a good idea to do backups regularly.

On the other hand, if you know nothing about partioning/formatting, I'd advise you to be very careful here obviously.
In case of doubt, you could also set up a free partition under Windows first - and make it 10-50 gb in size, give it some unique name like "FlightGear", that's not used by any other partitions, so that you can easily identify it.

PS: It's never a good idea to mix private/work stuff on the same computer/user account. Accidents do happen, and it's always better to separate things, i.e. by having a separate "FG" account, or even a completely different OS - one that you use for private stuff (web surfing, games etc) and one that you use for work. Especially under windows this is increasingly important because of viruses, trojans, worms etc