A Lie Revealed: Slowdown Bug

[Capt Dizle]

There are two versions of Firaxis explanations going around about the cause of slowdown in the game:

1. Caused by advanced AI pathfinding
2. Caused by culture calculation and map redraw

Some might want us to believe its # 1 but it is almost exclusively #2

Proof:

Once slowdown starts it occurs on exit of city screen during player turn...this is because the culture routines are triggered every time you access it...

No units are moving at this point are they?

Culture, the most despised game concept ever implemented, is the cause of this game breaking flaw.

Firaxis has no choice but to rework the game and remove all aspects of culture so as to restore fun to the genre.

[Worthingtons]

I agree that the game needs to be sped up, but culture shoudnt be abolished.

[Capt Dizle]

You can have borders without culture, what do you need culture for?

[asleepathewheel]

Originally posted by jimmytrick
You can have borders without culture, what do you need culture for?

what a set up. but...pass...

seriously, I don't know what you're talking about, slow down after exiting the city screen? It only slows down for me between turns, and only then when there are hundreds of units on a huge map.

Originally posted by jimmytrick

Firaxis has no choice but to rework the game and remove all aspects of culture so as to restore fun to the genre

again I'm sure they'll fix that now that YOU have brought it to their attention!

[Aeson]

I've noticed unit (usually workers) movement as the main cause of slowdowns in my games. When manually moving my workers around on Huge railroaded maps, often 1 worker will take 2 or 3 seconds to move 30+ tiles. I've had waits of up to 10 seconds occasionally for a single worker to move. Usually by this time the AI are almost non-existant, and nothing else causes a slowdown at all. Even with my workers automated, they are by far the most time consuming portion of each turn. This is of course with all animations and show movement options turned off.

What culture routines are you talking about? I don't understand why the game would be doing any processing that deals with culture while entering or exiting the city screen, other than to display the culture value. That function shouldn't cause any noticeable slowdown. I would think the slowdown you are seeing upon entering/exiting the city screen is mostly graphical in nature.

[Capt Dizle]

It is graphical in a sense but it is tied into culture. The borders are drawn based on culture and its a series of mathematical calculations based not only on the culture of an individual city but all surrounding cities. Essentially, every time you or the AI does anything to a city it triggers a check of the culture ratings of every city on the entire map and this is done many times each turn thus the increased time based on the number of cities not necessarily based on map size (except more cities on larger maps) and it has nothing to do with movement which is separate and less important (because its less crippling) issue.

[Über KruX]

let's not go into the culture sucks thing

let's go into how military occupation should push back borders

[Capt Dizle]

Aeson, you are referring to the railroad bug, what are the specs of your computer.

[Aeson]

1.33GHz T-Bird/512MB RAM/ATI Radeon 8500

In my last Huge map game, worker (~500 of them) movement accounted for about 15 minutes per turn near the end of the game. Everything else combined maybe took a minute to process. This isn't counting the time I spent manually setting builds, which took a few minutes depending on how many cities had completed their previous build.

I don't understand why you think that all the cultural borders are checked every single time you exit a city screen. Nothing you can do in the city screen is going to change the culture for that turn. There would be no purpose to doing the calculations at that time. I'm not saying it's impossible, useless calculations have been known to crop up in software before, but highly unlikely that two such unrelated processes would be linked.

[Capt Dizle]

Your config is powerful enough to blow through what I am talking about. What you are experiencing is not as important because few folks are ever going to endure this mess long enough to get to the huge map 500 worker level. But it should be fixed.

What I am lamenting is exactly what you say it is...there is no purpose to the culture calculations being done in this manner, but they are, and this slowdown affects folks on the lower end of the computing power spectrum. Folks with the amount of juice you have don't see this as a problem at all.

It might be unlikely but it is happening.

[DrFell]

When you exit the city screen it redraws the map, the culture routines might be accessed then to display the borders.

[Capt Dizle]

Right.

[notyoueither]

Which version(s) of Windows for those experiencing slow downs?

I have far less than Aeson. No significant speed problems for me. I do get the lag for a sec or so when exiting city screens from time to time. No big deal for me.

I'm using ME on the comp that plays civ3.

Salve

[Aeson]

Well, the cultural borders do have to be redrawn quite often. Do you experience the same delays while scrolling? Each time you move the focus of the map, the cultural borders on screen need to be redrawn as well. It would stand to reason that this process is the same as when coming from the city screen, or any other screen for that matter.

[Capt Dizle]

No it does not occur when scrolling. Scrolling is perfectly smooth.
This points to less than optimal code. The routines are triggered when dealing with cities. There lies the problem.

[ACooper]

Just to wanted to mention that just because something needs to be redrawn does not mean it has to be recalculated. The values would be stored for later retrieval.

To do it the way it was suggested would be absolutely stupid and I think we should all give the programmers more credit than that.

[Capt Dizle]

Aeson, do you see a long pause when the map is redrawn after you take or raze a city. You may not notice this with your computing power.

This kind of event triggers a recalc. It is longer in this case than when I exit a city screen because the map HAS to be redrawn, when I exit a city screen the routines are obvisouly being run but the delay is shorter because the map DOESN'T HAVE to be redrawn. This is the problem, they need to rework the code to take this constant routine out of the loop.

[Thrawn05]

I have to agree with DrFell that it is a redraw issue. If you look at the folder with all of the bitmaps, you can see that most screens have layer upon layer of bitmaps. The reason that scrolling on the map is not slow is because it is all redrawn when you leave what ever screen you were on before.

I personal would like culture to stay, it beats having a rival civ plant a few cities right next to your captial.

[Capt Dizle]

Cooper you always defend them. I am not calling them stupid, they just made a mistake. I used the "Lie Revealed" stuff just to draw attention to this. If I am mistaken let them go on record and I will eat dirt.

[Aeson]

You mean no delays at all on a Huge map with 100's of automated workers notyoueither? What are your system specs? I can't imagine ME being the "best" for anything..

I'm running Windows XP. Civ3 seems to run pretty well on my system. Other than a bug in 1.07f where it would crash to desktop when I tried to auto move a settler over a river, I haven't had any crashes at all.

[asleepathewheel]

Originally posted by jimmytrick
do you see a long pause when the map is redrawn after you take or raze a city. You may not notice this with your computing power.

Ah, so that's what you were talking about. That i have witnessed, and yes it is annoying. Sometimes the pause for me is 30 sec to a minute. So in this case, I concur with you Jimmytrick that it is a problem, but to scrap the whole system? a little harsh for me. And I have a 1.7ghz w/ 384 ram.

[Aeson]

No delays at all when razing a city or when cultural borders expand. It's just the unit movement that slows things down.

[ACooper]

Originally posted by jimmytrick
Cooper you always defend them. I am not calling them stupid, they just made a mistake. I used the "Lie Revealed" stuff just to draw attention to this. If I am mistaken let them go on record and I will eat dirt.

I defend what is right.

You are right that when a city is taken by another civ the culture values (and borders) have to be recalculated.

All I'm pointing out is that otherwise it would not need to be recalculated.


Of couse a MPG of you (or anyone else) eating dirt is always fun.

[Thrawn05]

Originally posted by Aeson
You mean no delays at all on a Huge map with 100's of automated workers notyoueither? What are your system specs? I can't imagine ME being the "best" for anything..

I'm running Windows XP. Civ3 seems to run pretty well on my system. Other than a bug in 1.07f where it would crash to desktop when I tried to auto move a settler over a river, I haven't had any crashes at all.

I'm playing Marla's Word Map with 14 other civs. I'm in the industrial age already, and I clocked turn times to be about 25-ish seconds for me.

I'm Win98, P3 550MHz, 256MB 100MHz SDRAM, for your FYI. I guess i'm just lucky.

[Aeson]

I'm talking about the time needed for your own automated workers, when you have 100's of them. The time needed to process turns is usually only 10-30 seconds even in the late game for me.

[notyoueither]

Aeson.

OK. I never have 100s of automated workers. I'm too much of a control freak for that. Also, I am playing large maps with 12 civs.

However, I never notice any lag in moving workers or any other units.

Perhaps your bug is in the auto worker routines and it affects all workers when you have so many?

Salve

[notyoueither]

Thrawn05.

Do you ever get 5 or 10 second delays coming out of city screens?

Does moving a worker ever take 10 seconds or so?

Do you use automated workers? How many?

Salve

[Aeson]

The delay happens whether or not the workers are automated for me. It almost always occurs when I am moving a worker more than ~30 spaces in a single turn. Anything under ~30 and the movement is almost instantaneous.

I usually keep control of my workers as well. I was using timing from automated workers as I can accurately state how long it takes them to complete their tasks. When manually setting the workers tasks, much of the lapsed time is from my decision making, and it's hard to keep track of the overall worker response time.

[notyoueither]

Hmmm. I'm commonly zipping workers 20 to 40 squares along rail. All instantaneous. I love that stack command... zip, 12 workers from point a to point b in 0.597 seconds.

I'm thinking that XP is the issue. The game would have been designed during the reign of ME (and 98). Makes sense that it works well under those two. In fact 98 and ME are very similar.

XP is a dog of a different colour. Many software vendors have had some problems adapting their software to it. All the Anit virus packages would be a case in point.

Hence my interest in which versions of M$ people are running.

Salve

[ICMB]

Hmm.

I've *never* seen the slowdowns that I've heard people complain about. It must be something specific to the system config that triggers it.

For what it's worth, I'm on Win2K. I play mostly on a 774mHz CPU with a GF2, and I've had huge maps where I have hundreds of units and never noticed any particular slowdown.

- ICMB