turn time since patch

Hajee

Prince
Joined
Nov 4, 2014
Messages
361
Has any one else notice the turn times since the winter patch have gotten worse?

I went ahead and did the AI benchmark. Prepatch (dx12 and dx11) I got round 17 secs. After the winter patch this jumped 27 secs on both dx12, and dx11.
 
Has any one else notice the turn times since the winter patch have gotten worse?

I went ahead and did the AI benchmark. Prepatch (dx12 and dx11) I got round 17 secs. After the winter patch this jumped 27 secs on both dx12, and dx11.

What is an "AI benchmark" ?

I am not sure what DirectX version has todo with it. But a jump from 17secs to 27secs is considerable. Are using the same savegame? any AI mods? What CPU do you have? --Because generally CPU becomes the bottle neck when dealing with AI.
 
What is an "AI benchmark" ?

I am not sure what DirectX version has todo with it. But a jump from 17secs to 27secs is considerable. Are using the same savegame? any AI mods? What CPU do you have? --Because generally CPU becomes the bottle neck when dealing with AI.

there are two benchmarks one for fps, and one for AI which test turn times.

it is not cpu because as mentioned it went from 17 to 27.

I was just asking if others noticed it too. The devs could have put a new .exe for the game which means they worked on the engine and possible missed something which lowered performance
 
or they added something that increased performance reqs.. Your concern about FPS and Dx version suggest that you are skidding on the minimal requirements of GPU performance, however, graphical fidelity isn't everything.. More often CPU performance --especially in late game and on larger maps-- is what counts. Because better AI often means it spends more time calculating the best choices, which would effect end turn time, which is when all AI players need to move units and manage their empires.

My main computer has new I7 cpu and even i have lategame lags in certain games.
 
Last edited:
or they added something that increased performance reqs.. Your concern about FPS and Dx version suggest that you are skidding on the minimal requirements of GPU performance, however, graphical fidelity isn't everything.. More often CPU performance --especially in late game and on larger maps-- is what counts. Because better AI often means it spends more time calculating the best choices, which would effect end turn time, which is when all AI players need to move units and manage their empires.

My main computer is has new I7 cpu and even i have late game lags in certain games.

It it not a FPS issues, as if you reread the OP that was never said. Thus why I did AI benchmark A=and not the graphic benchmark.

With that I am know technical components and know its not my cpu, because before patch the performance in AI processing was better after the patch it got worse. Best example I can give is it the delay it starting the AI process has increase by alot.

Before patch the AI benchmark it would take maybe 10 sec for turn 1 to start, and and less there fore after. After the patch there is this hang time in-between each turn greater then 20 secs although after the 1st turn I notice the hang time be about 15sec, as before was still 10.

Now the AI benchmark in a stress test to the extreme however it does a good job of representing late game maps.

With all that said, I would agree with you in before patch was 30 secs and after patch was 30 secs which would mean yes cpu is not good enough. However since before patch it was fine, and after the winter patch the turn time increased by 50%ish I would say they made in error somewhere.
 
same playing on a huge map is a nightmare i don't know how i see some guys are playing with every civ active on huge maps their turn time must be horrendous i get lag with just 8 civs total! and it's not my rig it can more than handle it. it can play planet coaster at ultra settings in 2k smoothly!
 
However since before patch it was fine, and after the winter patch the turn time increased by 50%ish I would say they made in error somewhere.
Why?
Of course, if the AI shows just exactly the same "performance" (functionality, not speed), it is worse to waste now 1/3 of the computing power. But who knows, what the AI can, what is achieved (or not) exactly?

Theoretical question to all: Would you be satisfied / satisfiable with waiting 8 MINUTES between turns provided, the AI plays 'ok' in your standards?
 
Last edited:
Now the AI benchmark in a stress test to the extreme however it does a good job of representing late game maps.
I still don't understand what is this "AI benchmark" you are talking about.. Are you running some external benchmark program, some internal civ6 consul function, or just using your phone to time turn length?!

Because increase in end-turn time can be attributed to many factors that are not mutually exclusive, especially if you are skidding on minimal requirements in any of the categories. I only suggest CPU, because it is the usual suspect in all CPU intensive games in lategame on larger maps, whether you playing Civ6, or Hears of Iron. Because even with best optimization techniques the more entities and the bigger area they are spread out along the more data you need to crunch every time.

Based on your anecdotal evidence I can't what is the culprit and sure it is possible they <snip> something. But I also know that people have been clamoring for many improvements (some like better AI can increase performance reqs) and it is possible they added some of that. And since there aren't forum full of people raging about 10sec flat increase in turn times, occam's razor suggest that this increase is likely mainly due to your system specs..

I know, I have a very old work laptop that I run modded civ4 on, and I tend to kill several process to get a little better performance out of it.

Moderator Action: Please make your point without swearing. Swear word snipped.
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889
 
Last edited by a moderator:
I still don't understand what is this "AI benchmark" you are talking about.. Are you running some external benchmark program, some internal civ6 consul function, or just using your phone to time turn length?!

Because increase in end-turn time can be attributed to many factors that are not mutually exclusive, especially if you are skidding on minimal requirements in any of the categories. I only suggest CPU, because it is the usual suspect in all CPU intensive games in lategame on larger maps, whether you playing Civ6, or Hears of Iron. Because even with best optimization techniques the more entities and the bigger area they are spread out along the more data you need to crunch every time.

Based on your anecdotal evidence I can't what is the culprit and sure it is possible they <snip> something. But I also know that people have been clamoring for many improvements (some like better AI can increase performance reqs) and it is possible they added some of that. And since there aren't forum full of people raging about 10sec flat increase in turn times, occam's razor suggest that this increase is likely mainly due to your system specs..

I know, I have a very old work laptop that I run modded civ4 on, and I tend to kill several process to get a little better performance out of it

As stated ABOVE the Ai benchmark is built into the game, and on the main menu. If you cant see that or dont know it is there then please do alittle research before reply! So this not some made personal program I am using.

As I said I noticed worse performance after the pacth, so I rerun the benchmark GIVEN by the devs and noticed the numbers were worse.

And again yes I know about cpu and cpu usage in civ games. BUT BUT BUT BUT BUT! since the cpu is the same as before the patch and the benchmark is worse one would say the devs added something and or an error in the patch to reduce performance.

I built my own machine (10 years running), and researched all the parts. I know components, I know how they work and I know to measure performance on my end. That is why I always stated, and key thing here, the before and after numbers.

I am not mad or upset, I just made the OP to see if others noticed the same performance drop as I have.

Moderator Action: Quoted swear word snipped.
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889
 
Last edited by a moderator:
same playing on a huge map is a nightmare i don't know how i see some guys are playing with every civ active on huge maps their turn time must be horrendous i get lag with just 8 civs total! and it's not my rig it can more than handle it. it can play planet coaster at ultra settings in 2k smoothly!

I want to quote this because, this post as do others are showing a poorly optimized game. And I personally have seen since the patch the performance go down. This does not mean the game is bad, just it is poorly optimized, and they could have made it worse
 
or they added something that increased performance reqs.. Your concern about FPS and Dx version suggest that you are skidding on the minimal requirements of GPU performance, however, graphical fidelity isn't everything.. More often CPU performance --especially in late game and on larger maps-- is what counts. Because better AI often means it spends more time calculating the best choices, which would effect end turn time, which is when all AI players need to move units and manage their empires.

My main computer has new I7 cpu and even i have lategame lags in certain games.

Why?
Of course, if the AI shows just exactly the same "performance", it is worse to waste now 1/3 of the computing power. But how knows, what the AI can, what is achieved (or not) exactly?

Theoretical question to all: Would you be satisfied / satisfiable with waiting 8 MINUTES between turns provided, the AI plays 'ok' in your standards?

The AI doesn't work that way in this game. It doesn't check for x # of combinations and choose the best choice or anything remotely like that. It follows a linear progression through a deterministic algorithm with seeded rolls generating results (or "decisions") from a decision tree. It's not smart, it's designed for predictability (fewer desyncs) in multiplayer where every computer calculates the end step independently. Because of this every calculation in the end step must be made in the same order every time which means that the end step must be executed in a single thread which is why it can take a relatively noticeable period, especially with larger maps.
 
It's not smart, it's designed for predictability (fewer desyncs) in multiplayer where every computer calculates the end step independently.

Interesting about multiplayer, does that include thing like path-finding calculations\caching or just strategic\tactical AI? And can you explain offer a little more detail on why its unique? as far as I know many game started to use multi-thread\cpu but to my knowledge its often for auxiliary task and subsystems, all the AI stuff still run on single thread, no?

Also I am not sure what you mean by smart, but surely the quality of the AI decisions can be improved by considering more variables, and certainly the more agents it need to run for the more time it takes.

I want to quote this because, this post as do others are showing a poorly optimized game. And I personally have seen since the patch the performance go down.

No just another anecdotal evidence. We agree that the patch changes something and that you see worse benchmark. But just because you see a decrease of performance on your 10 year old rig which you don't share its specs, doesn't necessarily mean its poorly optimized as you assert.

And from the feedback I seen here about the AI, I personally would prefer if they improve it, even if it means that few extra seconds for people playing huge maps on 10 year old rigs.
 
Last edited:
No just another anecdotal evidence. We agree that the patch changes something and that you see worse benchmark. But just because you see a decrease of performance on your 10 year old rig which you don't share its specs, doesn't necessarily mean its poorly optimized as you assert.

And from the feedback I seen here about the AI, I personally would prefer if they improve it, even if it means that few extra seconds for people playing huge maps on 10 year old rigs.

let fix that, as i been building for 10 plus years, the machine I have is only 2 years old.

" poorly optimized" is true and not anecdotal evidence. It is okay a point flaws in the game, but still like the game. No need to defend everything about even with the facts point else where.

Fact is after the patch the performance went down
 
It follows a linear progression through a deterministic algorithm with seeded rolls generating results (or "decisions") from a decision tree.
Looks a bit abstract, but methinks, we need a more complex decision tree. Some more branches, many more twigs, a lot of more leaves ... no?
(implying 'perhaps' more computing power)

It's not smart, it's designed for predictability (fewer desyncs) in multiplayer where every computer calculates the end step independently.
Well, yes, 'desyncs in multiplayer' make my sleepless nights ... especially when I'm playing CIV, you know I, PC, myself and me: all playing CIV.

It's certainly alright, that the different multiplayer results are consolidated and give a consistent and reliable result to all human players ... but do you try to tell me, because of that the AI "is not smart", i.e. the moves of the AI players are fine insync, but unfortunately themselves decided by throwing dice?
 
I've noticed a decrease in performance since the patch as well. Not overly significant but enough to have me stay on DX11 with my outdated GPU.
 
Interesting about multiplayer, does that include thing like path-finding calculations\caching or just strategic\tactical AI? And can you explain offer a little more detail on why its unique? as far as I know many game started to use multi-thread\cpu but to my knowledge its often for auxiliary task and subsystems, all the AI stuff still run on single thread, no?

Also I am not sure what you mean by smart, but surely the quality of the AI decisions can be improved by considering more variables, and certainly the more agents it need to run for the more time it takes.
As stated above, the only variable considered in a given calculation is the random seed, the rest is scripted. IE, the seed generates a random number which is applied to a table (with static modifiers) to see where a unit moves, whether a civ offers peace etc. All you can do is script the actions better and change the % chance of a given decision. Nothing else is possible under this system, which can't/won't be changed now. So expect refinements but not miracles. Given the additional movement restrictions present now I doubt they could even reach the level of Civ 5 without a *lot* more effort though.
 
Last edited:
Looks a bit abstract, but methinks, we need a more complex decision tree. Some more branches, many more twigs, a lot of more leaves ... no?
(implying 'perhaps' more computing power)
Possible yes, but as the results still have to be calculated one at a time, sequentially, then you need to be careful how complex you make that. There are gains to be made here for sure, but they are limited in scope.

It's certainly alright, that the different multiplayer results are consolidated and give a consistent and reliable result to all human players ... but do you try to tell me, because of that the AI "is not smart", i.e. the moves of the AI players are fine insync, but unfortunately themselves decided by throwing dice?
This is essentially correct. There is no intelligence in this game, just a way for the "AI" to select from various scripted actions in the form of what is called a multi (or split) level decision tree. So you separate Macro (Strategic) and Micro (Tactical) decisions, but that's about it. It's old tech and not what is considered to be an AI (or attempt at AI) by modern standards. But then this is a game, not a true attempt at simulation.
 
As stated above, the only variable considered in a given calculation is the random seed, the rest is scripted. IE, the seed generates a random number which is applied to a table (with static modifiers) to see where a unit moves, whether a civ offers peace etc. All you can do is script the actions better and change the % chance of a given decision. Nothing else is possible under this system, which can't/won't be changed now. So expect refinements but not miracles. Given the additional movement restrictions present now I doubt they could even reach the level of Civ 5 without a *lot* more effort though.

So instead of the server making the decisions and sending the outcome to everyone, they are using a shared seed to generate "random" numbers guaranteeing the AI reach the same outcome on all the peers. But other than how is it different from most other games? I used to mod paradox games and its strategic AI logic is exactly that, a big script of states and chances. Obviously no one expects miracles but with a lot of tweaking you can give a better appearance of intelligence.

But back to my original comment, you seem to contradict the suggestion that increase in complexity of AI decisions, number of agents and size of map (which should effect various logarithm like path finding) wouldn't increase performance, and I am not sure why. --the second part of your comment, about the implementation forcing it to run many things on a single thread\core, sound plausible (to my ignorant ears) which would means CPUs with higher clock frequency's would fare better explaining what is observed by many others.
 
Last edited:
Has any one else notice the turn times since the winter patch have gotten worse?

I went ahead and did the AI benchmark. Prepatch (dx12 and dx11) I got round 17 secs. After the winter patch this jumped 27 secs on both dx12, and dx11.

I got 18.82 seconds before the patch and 24.06 afterwards (both numbers are average of three test runs, using i5-3570K@4.2GHz). Not as drastic increase as yours, but it's there.
 
Back
Top Bottom