[1.09e] Out of Memory

Large map, 7 AI factions, Vista 64 / E8400 / 6GB ram / Ati 5870. Details in debug.err: http://pastebin.ca/1964113

Thank you for your continued efforts to improve the game.

5,934 views 12 replies
Reply #1 Top

Since this is the 1.09e memory thread, I figured I'd post that I haven't seen memory-related crashes since this patch.  I am still getting the occasional 'in combat' freezeup though (quit & restart game time).  Thank goodness for autosave.

I regularly play on large maps with 10 opponents.  I've made it past Turn 300 a couple of times now before deciding I've pretty much won and start a new game.

AMD Turion 64x2, 2 gigs memory, ATI Mobility HD2600, decent free hard drive space.  Windows Vista Home Premium.

I use the old 'define min & max virtual memory at same values' trick, paging file size set at 16,384.  Dunno how much of an effect this is having on memory performance since the patch, but I figure I'd mention it in case it might help anyone.

It looks like they are at least getting a handle on this for most of us.  For those still having problems, posting your system specs and error files is always helpful to the developers, so they can try to figure out what configurations are still having trouble.  As the OP on this thread did.

 

Reply #2 Top

I spoke too soon.  Ran a whole bunch of turns back to back, and just hit the out of memory error.  Debug.err here:

http://pastebin.ca/1965549

I suspect I'll be able to load an autosave and continue.  If not, I'll post again.  In the meantime, no worries!

 

Also, if my debug.err is correct, apparently I haven't updated video drivers in a while (autoupdate fails again).  I'll  check the AMD/ATI site for info.

Edit: Checked AMD site.  I have manufacturer specific drivers apparently.  Checked Toshiba's site, no newer Vista graphics driver updates at all (figures)... I could roll the dice with the vanilla CCC, but am not feeling that frisky today.

Reply #3 Top
We have a new theory on this. Could you try a game where you do no tactical battles but rather just auto resolve. We are seeing some systems where the texture memory on the video card doesn't get cleared after a battle but gets swapped to main memory. It's only on some systems but we reproduced it finally. If someone can verify that it would be helpful.
Reply #4 Top

Game reloaded fine, and I played a few more hours before the next crash, which was instigated when the antivirus decided to start a full scan at the same time I was attempting a save (so not your guys's fault).  The first crash was 'between game turns', but I'm sure you picked up on that with the debug.err.

In my case, I've found if I shut down all other 'background' programs (antivirus, etc.), the game is stable for much longer periods, and haven't seen many crashes.  I usually quit after 2-3 hours of play, though, and take a break.  Pretty happy with the stability so far.

Reply #5 Top

Quoting Frogboy, reply 3
We have a new theory on this. Could you try a game where you do no tactical battles but rather just auto resolve. We are seeing some systems where the texture memory on the video card doesn't get cleared after a battle but gets swapped to main memory. It's only on some systems but we reproduced it finally. If someone can verify that it would be helpful.
End of Frogboy's quote
Do I have to start a completely fresh game for this test? I reloaded my old autosave for the game mentioned in the OP and got an out-of-memory again, even though I auto-resolved every battle. For completeness I should mention I upgraded my audio drivers. debug.err here: http://pastebin.ca/1965819.

Trying to interpret the debug.err myself, it seems especially the last turns show dramatic memory increase for example growing 90MB at turn 216 / 217. It may have resulted from researching Housing? I may try a completely fresh game later in which I will also autoresolve everything.

Reply #6 Top

Quoting Frogboy, reply 3
We have a new theory on this. Could you try a game where you do no tactical battles but rather just auto resolve. We are seeing some systems where the texture memory on the video card doesn't get cleared after a battle but gets swapped to main memory. It's only on some systems but we reproduced it finally. If someone can verify that it would be helpful.
End of Frogboy's quote

I did a test game using autoresolve on harpo's weirdie large with 10 ai custom sov & race, win7 64bit, ati5850 12gb ram, got to turn 230 and OOM

similar number of turns to fight each battle within 10%

added dxdiags,saves,mapseffects,items,maps,mods,quests,race,racetypes,stamps,tiles & units folders to zip

harpo

 

Reply #7 Top

I have this error too. However, The out of memory crash will occur ever 20 min to two hours. Its as simple as saving frequently, then going out back and beating the hell out of the wood pile when I have to take over an empire.. again.

 

Note: I have only seen the "out of memory" box 27 times since 1.09e was installed on my mashien, but the timed crash keeps occuring.

Reply #8 Top

I only really seem to have oom problems late in the game, but i always play on huge (or bigger) maps and usually with a full docket of opponents.

Reply #9 Top

I have found that the LAA trick DOES help on 64 bit windows machines, but does NOT help on 32 bit machines (approximately doubling the number of turns till OOM). the downside of the laa is IT IS modifying the EXE for elemental and technically in breach of the licence agreement with stardock.

harpo

 

Reply #10 Top

What is the LAA trick?

Reply #11 Top

xghost, do a forum search for LAA, you will find a thread that offers the LAA trick, I do not have a link to it, so can not post a link.

harpo

 

Reply #12 Top

I wouldn't worry about the breach of licence thingamajig, it's technically not. You're modifying the header of the executable file, not anything *in* the executable - this is why you can find an app that was written for another game, and use it to turn on LAA for Elemental - the header is universal and used by Windows, it is not specific to Elemental.