Civ5 performance analysis

The current design seems to be something along the lines of:

1 parent/overlord ai/thread
\- 1 diplo ai/thread
\- 1 builder ai/thread (cities, workers, whatever)
\- 1 military ai/thread
\- 1 civics/tech ai/thread

then each civilization gets processed by the parent.

that's fine. but the divisions are not equal. most likely the crunching happens in deciding where to move units -- more units over time -- which I theorize explains the slowdown over time.

so take the design the next step. create parallel overlord threads

\ master thread
-\ civilization 1 overlord
--\ diplo
--\ builder
--\ military
--\ civis
-\ civilization 2 overlord
--\ etc
-\ civilization 3 overlord
--\ etc

the sequential turns might make sense if you want AI #2 to respond to AI #1's previous move, but that should happen on the /next/ turn, not the current turn.
 
Hmmm, I'm not sure I ever want to see the AI relying on a simultaneous-moves mechanism, or even just pretending it's that way. Come to think of it, I wonder how civ4 AIs deal with it in simul turns MP games. :confused:

I'm sorry but I don't believe splitting up civs into the separate overlord threads is where the solution lies. I think there is some truth in the suggestion that the problem here boils down to the new unit rules (1upt, longer movement ranges, inability to easily group units into stacks, fewer units meaning individual units are required to multi-task better unlike civ4 units which each were created with a particular role etc.) and how it is difficult to multithread much of those processes.

Apart from the more demanding graphics, one could ask why it is that civ5 takes so much longer with its inter-turns than civ4, considering that most aspects of the gameplay apart from that dealing with units are essentially the same. In fact, if one accepts the common rhetoric around here that the game is less complex, it should be expected for the AI turns to take less time, especially with faster CPUs, than the times in civ4.
At this point in time, I'd most love to be able to look at the SDK so I can see just what the AI does about this.
 
Not sure.

you might be right about the more complex pathing with 1upt/hexes, but:

in an 8 AI, 16 city state game, or argument's sake, it takes:

: 5 seconds per AI
: 2 seconds per CS

and there is 1 "unit AI", then

+ 8*5 (assuming none are conquered yet) = 40s
+2*16 = 32s
+ 52s for sequential unit AI.

Imagine that was collapsed so that each unit AI could process simultaneously. Your CPU would very likely be topping out at 100%, but the turn duration would drop dramatically as a factor of the number of cores.
 
Running the game at the highest settings, dx11, 8xAA, 1920x1080, etc. and not a hitch at largest map with 11 AI opponents. I was wondering why so many people were complaining, and it seems that I somehow managed to hit the sweet spot! Pure luck.

Here's the sub $1k computer build I did a month ago just for CiV :old: Might be useful if anyone is thinking of upgrading.

  • i5 760 - OCed to ~3.8Ghz
  • GTX460 1gb - not OCed yet
  • 4gb gskill ripjaw 1333 - OCed to ~1600
  • asus p7p55d pro mobo
  • 950W corsair PSU (an overkill, was cheaper than 650W at the time I got it-650W should be more than enough)


This system runs CiV even at the most demanding settings. The AI also seems to be pretty smart too!! (the first two games were King) I wonder if that scales with the horsepower you have.. Anyone can confirm this?? :confused:
 
This system runs CiV even at the most demanding settings. The AI also seems to be pretty smart too!! (the first two games were King) I wonder if that scales with the horsepower you have.. Anyone can confirm this?? :confused:

I don't like to spoil your wishful thinking, but I am 99.99% (or more :p) sure that your computer's power has no effect on how smart the AI is. It is simply not how games like this have been built or designed.

However, it may be that you enjoy significantly faster turn-times (when you press "next turn" and have to wait for the AI).
 
However, it may be that you enjoy significantly faster turn-times

Yep, definitely could be a 'positive' side effect of waiting only 5 seconds between turns :) It's just that they seem to attack where I'm the weakest, and they do utilize ranged units really well! Almost got butchered yesterday. But that's probably a game specific event, considering so many people are complaining about it..

I remember waiting for at least half an hour in Civ 3 between the turns at later stages on my P3 rig. Ah good ol' days... Don't miss them one bit :lol: Overall no sequel matched the speed of Civ 1 on my 486 :)
 
Running the game at the highest settings, dx11, 8xAA, 1920x1080, etc. and not a hitch at largest map with 11 AI opponents. I was wondering why so many people were complaining, and it seems that I somehow managed to hit the sweet spot! Pure luck.

Here's the sub $1k computer build I did a month ago just for CiV :old: Might be useful if anyone is thinking of upgrading.

  • i5 760 - OCed to ~3.8Ghz
  • GTX460 1gb - not OCed yet
  • 4gb gskill ripjaw 1333 - OCed to ~1600
  • asus p7p55d pro mobo
  • 950W corsair PSU (an overkill, was cheaper than 650W at the time I got it-650W should be more than enough)


This system runs CiV even at the most demanding settings. The AI also seems to be pretty smart too!! (the first two games were King) I wonder if that scales with the horsepower you have.. Anyone can confirm this?? :confused:

I don't suppose you could upload a save of this game. I get so much lag beyond the game AI stuff itself I can't even play past turn 250 without long lag spikes.

Right now I'm trying a game huge/pangea with no saves at all and not turning my computer off and just leave the game running 24/7. I'm at turn 198 and about 2/3 of the map revealed. Barbs keep killing my scouts and warriors are too slow and die often as well. I hear so much talk about these saves messing things up. so we'll see if this is my issue.
 
I just did an old fashioned stopwatch benchmark on a late huge stl earth map.
I checked 3 turns and made sure nothing was different, I tested the run of 3 turns 2 times for dx9 and 2 times for dx10/11 modes, my gpu is a gtx 260 only dx10 but it did show something really interesting, apparently the game uses cuda through the dx10/11 api to aid turn/ai calculation.

My general fps were sad with dx10/11 mode but it cut a good part from turntimes which I find way more interesting in civ than fps as those times can run into minutes on very late huge games.
After the turn timetest on each run I alt tabbed, started fraps and ran a 30 second benchmark of zooming in and out and scrolling the huge map zoomed out.

Here are the results:

DX9
turn 896 89 seconds
turn 897 89 seconds
turn 898 88 seconds
FPS min/max: 7/61
FPS average: 27.7fps

Identical turns and ai decisions insame game on dx10/11 on a dx10 gtx260:
turn 896 45 seconds
turn 897 46 seconds
turn 898 44 seconds
FPS min/max: 6/28
FPS average: 17.4fps


I'd suggest if you have at least a dx10 cuda enabled card and you don't mind your tiny animations to run a bit sluggish zoomed out but are getting fed up with long turn times to use dx10/11 mode, it feels more sluggish but AI turns go quicker.

My system details:

AMD Phenom II X4 955 BE (at stock, need to get better cooling)
AMD 790GX based motherboard.
3 GB DDR2 RAM
XFX GTX 260 216 cores 896 mb videoram black edition (core 685 mhz, shaders 1470mhz, memory 1005mhz)

I thought I'd give the oc details for comparison.

There is a load of disk loading going on a well sometimes, perhaps it's just having 896 videoram, the grey tile loading in dx9 seems to suggest the vram is full and being swapped.
I haven't done a decent defrag of the harddisk yet after I got ciV, if that changes a lot for this game I'll post the benchmarks here.

Cheers!
 
Highly interesting stuff.
I seriously doubt though that it's due to GPGPU, 2k would have marketed that all over the place. Is that even possible with DX?
I would guess that DX9 mode is sacrificing CPU power for better graphics performance, but it is extremely astonishing that it takes twice the time. Could you post the save?
And/or maybe try out how DX9 mode fares with this?
 
Well these are my measurements in dx10/11. I would not know another explanation, the measurements stay consistent and there is really not much that directx10+ could otherwise be doing that would give such a positive performance impact on turn times I think. DirectX does not control in any way how well the game is using my quad core cpu so that can't be it, (cpu usage in taskmanager actually went down a bit compared to dx9.)

But I did find the a (part of the) solution to bad dx10/11 fps/sluggishness in general on dx10 cards, it seems to have put my dx10 card back to fps I had in dx9 but still giving me the win on the turn times using dx10/11 api:

go into your my documents\my games\sid meier's civilization 5 -folder, open the "GraphicsSettingsDX11.ini" file and find the line that says "BicubicTerrainTessSubdiv = 2" and change it to 0.

My theory on why this is: It's doing tesselation on dx10 hw which is akin to, in the past, trying to run Hardware T&L through software, it's too hard for hardware not made for it.
The terrain polygon detail goes back down to dx 9 levels and my fps on my gtx 260 are back up to what they were in dx9 except now my turn times are still almost halved or indeed halved in some measurements.

To recap my benchmarks including the tweaked dx10/11 one:

Code:
            DX9  DX10/11 on gtx260 DX10/11 on gtx260 with tweak
Turn 896    89s     45s            45s
Turn 897    89s     46s            45s
Turn 898    88s     44s            44s
FPS min/max 7/61    6/28           12/62
FPS average 27.7    17.4           27.4

Cheers!

Edit, to answer your question, technically yes, also it wouldn't be a problem to add the cuda api to the game if it weren't included in the drivers of the card. I'll have to research some to be really sure though.
It's just my theory to explain the facts I'm seeing on my pc, my turn times have almost halved, that's a huge difference, my reaction time on the stopwatch which should be somewhere between less than half a second and 1,5 sec for most people isn't even significant when the result is going from 1,5 minutes to under a minute to do the same turns over and over (to see if its consistent and allow for ai making perhaps a different decision in some turns).
I wasn't expecting this myself at all when I decided to benchmark dx9 vs dx10/11 on my system.
I was getting tired of the long turns, when I read half a book playing yesterday between turns on dx9 I thought it became worth checking if you can tweak those turn times a bit.
I'm googling a bit to see if I can find anything. I don't know if they would tout the acceleration as obviously but I also can't think of anything else that would slice my turn times in half especially since dx10 performance is usually worse as a trade-off to nicer graphics.

Anyway I thought it was a discovery worth sharing on civfanatics, perhaps it will help some people as it did me.

I know I am a happy emperor! :)

last edit lol:
The tweak can't harm your game to try, just don't turn it higher than the preset value of 2.
If it works in any way like 3d studio max's subdivision/tesselation I use occasionally to create more poly's/faces, it's times 4 for every level of tesselation:
e.g. if 0 = 200 faces then 1 is 800, 2 is 3200, 3 is 12800, 4 is 51200 faces etc.

[edit]or not, as there hasnt been a reply yet, I've just played my game for 20 turns more, the sluggishness isn't completely gone with my tweak but the fps still are a lot better then they were at first with dx10, I'm still looking if I can find something else. One thing that is really bothering me with dx10/11 is the way it apparently deals with textures not loaded yet, it prevents you from scrolling over them where you would see grays disappearing, so okay no greys like dx9 but a stutterish scroll instead, really annoying I don't mind the occasional gray tiles filling up, at least scrolling was smooth and easy on my eyes and the gray tiles tell me where the problem is for when I have cash to upgrade.

If I find anything else that helps performance with dx10 cards I'll post again, cheers.
 
I don't suppose you could upload a save of this game. I get so much lag beyond the game AI stuff itself I can't even play past turn 250 without long lag spikes.

Right now I'm trying a game huge/pangea with no saves at all and not turning my computer off and just leave the game running 24/7. I'm at turn 198 and about 2/3 of the map revealed. Barbs keep killing my scouts and warriors are too slow and die often as well. I hear so much talk about these saves messing things up. so we'll see if this is my issue.

I don't think it has anything to do with the savegame. What's the hardware you're running it on?

Right now I'm at turn 315, and it takes about ~15 seconds from turn-to-turn. This is on the largest map, in a standard game. I can upload the savegame when I get back from work, but your problem is probably a hardware issue.

go into your my documents\my games\sid meier's civilization 5 -folder, open the "GraphicsSettingsDX11.ini" file and find the line that says "BicubicTerrainTessSubdiv = 2" and change it to 0.

Kaell, the results you got are really interesting! I don't know why people haven't checked this out, but I believe this is a significant find. I'm gonna do some testing when I have a chance.. Wonder what could be causing such a huge difference?!
 
I don't think it has anything to do with the savegame. What's the hardware you're running it on?

I agree on the save game, was just hopeful I guess.

Rig #1: i7 920, 6GBB RAM, 5870 OC, and 160GB SSD are main hardware.
Rig #2 i5 655k, 4GB RAM, 5670 Go Green, and HDD

both machines crap out the same, except the i5 is less lag because it's faster @3.2GHz vs i7 @ 2.66GHz

Right now I'm at turn 315, and it takes about ~15 seconds from turn-to-turn. This is on the largest map, in a standard game. I can upload the savegame when I get back from work, but your problem is probably a hardware issue.

Yes, please upload a save game, I would greatly appreciate it.

you sound worse than me. I was getting around 8 seconds per turn up until turn 330ish and bam, instant 30sec lag spikes with task manager reporting civ v was not responding and everything locks up while lag spike happens, then continues normally.

Kaell, the results you got are really interesting! I don't know why people haven't checked this out, but I believe this is a significant find. I'm gonna do some testing when I have a chance.. Wonder what could be causing such a huge difference?!

as for the "BicubicTerrainTessSubdiv = 2" is that disabled by just turning off AA in your options? or does the high terrain effects reenable it by default? It just sounds like x2AA is enabled to me.

Here is a link to the save game and the issues I'm having.
http://forums.civfanatics.com/showthread.php?p=9757801#post9757801
 
Thanks for the linkage :)

I really hope it will be of some help to people, I really want to get it to run with playable crispness/laggyness and fps on dx11, because for whatever reason the turn-time being almost-halved remains true in dx11 for me.

I started a new game now, not specifically for testing but couldn't resist checking dx9 vs dx11 on it too, I'm still getting my turn times 40-50% faster in dx11, even when the game goes horribly sluggish as far as interface, graphics and feel go the turn times are 40-50% faster than dx9. I'm still betting on some cuda code being the reason for this, as it can't use cuda through dx9 directly but it has been a part of at least the nvidia driver set post-dx10/11 which can use gpgpu code directly this way afaik. (to be honest I really don't care, i'm getting half my turn times which is a godsend for someone playing marathon games on huge maps, my first game lasted almost 2000 turns until I blasted off into space and left the suckers behind ;).)

There is still more going on though, I've been playing with other *.ini settings.
If I can pinpoint any concrete effects on how playable to game is using dx11 on older dx10 cards to get the shorter turn times I will post or if I find the reason for the slowdown.

What the setting did for me after longer play, getting on with my empires instead of testing, is make dx11 playable for a few hours most of the time instead of being ridiculously sluggish from the get go.
Now I can play for an hour up to a few hours and suddenly for some reason it goes from really crisp to sluggish like hell again, I just have to save, exit civ, restart civ and I can play crisp for another few hours again. (memory leak?)

Turn times do go up as the turns count up but that's to be expected, the turn times really seem to come from AI calculations, how much I see of the map does not seem to matter (Which is right if coded well, whether I see it or not the AI still has to decide on a move and make that move anyway.)

I'm still under a minute for a turn on dx11 being at turn 1100-ish now with 19 civs inc. mine on the huge stl earth mod/map. It's getting close to 2 minutes if I play the same save in dx9.

One thing I found weird comparing the dx9 and dx11 ini's is they seem to have doubled most "Terrain Page in Rate" values in dx11's ini, as if all dx11 cards are going to be more horsepower than a dx10 card by default. (which is far from true, dx11 is just the api the card can work with/up to, it says nothing about the HP on the card.) I have no idea what the devs were thinking, I put all those rate things back to the levels they have in the dx9 ini, unsure about the performance impact, still tweaking but doesn't seem to hurt visually or performance wise but I need to play with them more to figure out how it impacts the game exactly.

I also turned down "TerrainPageinSpeedStill" and "TerrainPageinSpeedMoving" to dx9 values, also still tinkering with that but it seems to help in keeping the game free from a sluggish interface/feel longer if you don't mind seeing some tile and texture loading after just loading a game or starting a new one. It only seems to be on the initial load.

Another setting worth trying and turn off, as it's name does not really truly describe what it does exactly but does include it in the potential suspects for sluggish performance on dx10 cards. Turning it off seems to have helped in postponing sluggishness in dx11 and I don't seem to see any visual difference, perhaps aa can't be turned on with it off.
"TerrainUseAdvancedGPU=1" to 0


------------------
@mark

AA has been off in dx11 mode in all my tests because it would be an unfair comparison otherwise.
I'm looking into getting the dx11 api version to perform better on my dx10 card so I can benefit from the faster turn-times, not get more/better visual effects (unless I can get them in non-sluggish ways, dx11 reflections do not seem to have anything to do with the sluggishness in dx11 mode btw ;) )

They had no reason to disable AA in dx9 (and you can force it fine through control panel) anyway.

If I left AA on in dx11 in my comparisons, I wouldn't be comparing api performance but comparing performance with or without FSAA convoluted by different api's.

The setting has nothing to do with AA, Tess stands for tesselation, subdiv stands for subdivision both are terms of geometry in this context.
Doing this tweak gives me hours of crisp game time before the mystery slowdown happens instead of having the slowdown right after loading my saves.

About the save game, is your save game growing in size disproportionately to what you'd expect? There seems to be a save bug using quick saves and loads, dumping a whole lot of duplicate junk data in the save game.

I don't have any issues loading games takes a little but not too long in both dx9/11
saving takes mere seconds.

I have finished to 2000+ turn marathon games on huge maps now and have less ram to boot.
Is your OS 64 bit? else the game will effectively have just 3 to play with (still should be enough because it's enough on my system)
----------------------


I forgot to mention what settings I play at in my previous posts listing the hardware used:
1680x1050, everything maxed except AA which is off because of comparing dx9 to dx11 performance.

For now I'm mostly going with my measurements, nothing beats measuring/testing, ask any scientist :)
My theories on why turn-times are much faster in dx11 mode, and why dx11 on dx10 cards is unbearable might be completely wrong, but the numbers are what they are. I hope someone more technical might offer a better concise theory on why this happens.

I'm just going to enjoy the fact I can now play for hours in dx11 mode getting half the wait time between turns and only get a sluggishness-attack after a few hours (seems to come out of the blue.) which is fixed by just restarting ciV and the savegame, not even a reboot needed and hope this information might have the same results for some other people having the same unreasonable sluggish performance on their still decent dx10 cards like the gtx 260.


Something is wrong I think, the sluggishness in dx11 mode in ciV on dx10 cards is beyond proportiante to the horsepower o some of these "old" cards.
My oc'd gtx 260 slaps a low end budget 4xx card silly and then uses it a a lawn ornament on pure horsepower (fillrate etc.)
Combined with being able to delay the sluggishness for hours with my tweaks on my system makes me thin it's a bug e.g. a memory leak or something either in civ, how it deals with dx11 or dx11 itself.
Maybe compounded or even caused by not detecting if the hardware is dx10 or dx11 (dx11 is backwards compatible but you have to make sure it doesn't try to send dx11 instructions if on a dx10 card and/or doesn't give the dx10 gpu stuff to do it cant really do in hardware.
(kinda like "software-hardware t&l", anyone who's been around a bit longer knows what I'm talking about, the only thing the ability to run the instructions over cpu pre-hwt&l gpu's served was to demonstrate its too much for anything not having hardware made to do it.)

Personally I think I found the compounding factor, the "BicubicTerrainTessSubdiv" setting which made a big impact on my experience with the dx11 build.

I hope my long writings don't annoy anyone.

One last thing though, anyone who's going to edit his dx11 ini a good practice to do is copy the line you are going to edit and comment out one, change the other so you have the original value still in the file.

It should look like this in the file, for example:

; Terrain Page in Rate
;TerrainPageinSpeedStill = 4
TerrainPageinSpeedStill = 2



I really need to sit down a weekend and test each value/setting seperately, but it's running quite well with the tweaks now so I'm not to hot on toying around more.


Here's all the changes in my dx11 ini file and my specs once more

Code:
[..]

; Terrain Page in Rate
;TerrainPageinSpeedStill = 4
TerrainPageinSpeedStill = 2

; Terrain Page in Rate
;TerrainPageinSpeedMoving = 2
TerrainPageinSpeedMoving = 0

; Use more advanced GPU features which can cause some drivers to crash
;TerrainUseAdvancedGPU = 1
TerrainUseAdvancedGPU = 0

[..]

; Terrain Page in Rate
; Terrain64Chunks = 4096
Terrain64Chunks = 2048

; Terrain Page in Rate
; Terrain128Chunks = 512
Terrain128Chunks = 256

; Terrain Page in Rate
; Terrain256Chunks = 256
Terrain256Chunks = 128

; Terrain Page in Rate
; Terrain512Chunks = 64
Terrain512Chunks = 32

[..]

; Default level of patch subdivision
;BicubicTerrainTessSubdiv = 2
BicubicTerrainTessSubdiv = 0

It's become a bit many changes, my plan was to return everything to original values one by one and test between each, but that will take very long because at the moment the sluggishness attack is postponed usually for up to 4 hours before it suddenly hits. (it's also completely separate from the amount of activity or how revealed the map is, huge saves are crisp just as long as new games before it hits.)

How the ini file is made up makes me think the BicubicTerrainTessSubdiv setting should not have an impact at all because tesselation is turned off in the file by default, but it still has the big impact for me so what should be and what is once again don't quite match it appears.

Once again sorry for my extensive writing style.
Hope it helps!
 
I think one of the most telling things of the whole Techspot performance review article is the last sentence in "Final Thoughts", where it says:
Rarely do we see such a powerful desktop chip pushed to its limits, but when clocked at 4.0GHz, that's exactly what we saw when testing the late game performance of Civilization V.
 
@mark

AA has been off in dx11 mode in all my tests because it would be an unfair comparison otherwise.
I'm looking into getting the dx11 api version to perform better on my dx10 card so I can benefit from the faster turn-times, not get more/better visual effects (unless I can get them in non-sluggish ways, dx11 reflections do not seem to have anything to do with the sluggishness in dx11 mode btw ;) )

They had no reason to disable AA in dx9 (and you can force it fine through control panel) anyway.

If I left AA on in dx11 in my comparisons, I wouldn't be comparing api performance but comparing performance with or without FSAA convoluted by different api's.

The setting has nothing to do with AA, Tess stands for tessellation, subdiv stands for subdivision both are terms of geometry in this context.
Doing this tweak gives me hours of crisp game time before the mystery slowdown happens instead of having the slowdown right after loading my saves.

I know a few games actually enable tessellation in the game through AAA. I just assumed it was an extension of AA and thought it may me a rudimentary tessellation feature. But looking at my ini file I see mine is set to 2 as well, so I'll give it a try next restart and see if it makes a difference for me.



About the save game, is your save game growing in size disproportionately to what you'd expect? There seems to be a save bug using quick saves and loads, dumping a whole lot of duplicate junk data in the save game.

I don't have any issues loading games takes a little but not too long in both dx9/11
saving takes mere seconds.

My saves seem to be fine. I tried playing a game without saves and crashed to around turn 300ish I think as my ram hit 2GB. It has done this a few times and I tried CFF Explorer to make sure greater than 2GB was enabled, and it was already. The weird thing is now my games go past 2gb and my current game task manager reports 3.5GB for the civ V dx11 executable. I don't know why it suddenly goes past 2GB, when before 2GB caused crashes.

My load times don't see abnormal on either machine and are roughly the same.

I have finished to 2000+ turn marathon games on huge maps now and have less ram to boot.
Is your OS 64 bit? else the game will effectively have just 3 to play with (still should be enough because it's enough on my system)
----------------------

Wow! 2000 turns. I never made turn 400 yet. lol

I'm using win 7 ult x64 and I'm installing win 7 home 32-bit on my dual core rig for testing as well. I see more 32-bit people having success on laptops most it seems.

and don't worry about long post. It is helpful sometimes to figure these things out.
 
Yes, please upload a save game, I would greatly appreciate it.

I've attached the savegame..I wanted to play a bit to test kaell's finding but you know the 'one more turn' syndrome..:) I'll get there!!

Anyway the savegame is at ~355th turn. I did the stopwatch and now the turns range between 34-40 seconds. 4 superpowers, large armies, etc. slowed it down a bit, but still very bearable considering that the end is close. It's on a 64 bit win7. Hardware-wise my hw sits right between your first and second rig. I wonder what your turns will be like.

Btw 60+ hours with CiV, and the game hasn't crashed even once.. No freezes, no odd slow downs, etc.

Have you checked out the CPU/GPU temperatures? That used to cause a lot of issues with me in Civ 4..Needles to say I was better prepared this time. I've conducted some temperature analyses (using cpuid's HWmonitor) and the game seems to bee as taxing, if not even more than Mass Effect 2 with ultra high setting everything at 1920x1080 res :eek:
 

Attachments

I've attached the savegame..I wanted to play a bit to test kaell's finding but you know the 'one more turn' syndrome..:) I'll get there!!

Anyway the savegame is at ~355th turn. I did the stopwatch and now the turns range between 34-40 seconds. 4 superpowers, large armies, etc. slowed it down a bit, but still very bearable considering that the end is close. It's on a 64 bit win7. Hardware-wise my hw sits right between your first and second rig. I wonder what your turns will be like.

Btw 60+ hours with CiV, and the game hasn't crashed even once.. No freezes, no odd slow downs, etc.

I downloaded twice and still no love. Maybe it got corrupt? Can you re-upload it in zip format? You are around the turns I start experiancing lockups depending on how much map is revealed and the number of cities. My first game started in at turn 250, but I had built six scouts right off the bat and had the whole world revealed early on.

This last game was around turn 330 before lag, but I didn't have the map revealed until then.

Have you checked out the CPU/GPU temperatures? That used to cause a lot of issues with me in Civ 4..Needles to say I was better prepared this time. I've conducted some temperature analyses (using cpuid's HWmonitor) and the game seems to bee as taxing, if not even more than Mass Effect 2 with ultra high setting everything at 1920x1080 res :eek:

Yes, my system is fine and very stable. My temps are way low as I bought water cooling for my gaming rig to keep heat down in my small room. You'd be surprised at how much heat a 5870 can put out under load.
 
kaell said:
; Default level of patch subdivision
;BicubicTerrainTessSubdiv = 2
BicubicTerrainTessSubdiv = 0

I did this and now have to terrain features at all! I have trees/rivers. no hills/plains/grassland/mountains.

are the other changes mandatory as well? I was just trying one at a time to see the effects. It's hard to pinpoint issues with many changes at once.
 
kaell said:
; Default level of patch subdivision
;BicubicTerrainTessSubdiv = 2
BicubicTerrainTessSubdiv = 0

I tried ths and lost all terrain completely. I get forests/rivers still, but that's about it.

 
Back
Top Bottom