Not buying this unless..

dissentience

Chieftain
Joined
Nov 6, 2010
Messages
3
Hello Civfanatics,

I would very dearly like to know, if anyone that already has the new game expansion can take a moment to clear this up for me, whether the issue with the civ 5 .exe only using one core has been fixed. I played many many hours of Civ 5 and only after getting many hours into my last game did I find out that it was practically impossible to play late game, due to the 15 min waiting time between turns. Seeking the answer, I looked through a couple forums and finally confirmed that the culprit was due to the application only using a single core out of my 4 cores by going into task manager and seeing the .exe CPU usage maxed out at 25%.

So, now that Gods & Kings has come out, I'm interested in giving it a go. But I don't think I'll buy it unless that issue was fixed.

I think the simpliest way to test this would be to open up task manager after clicking "NEXT TURN" in game, and looking if the .exe goes above 25% CPU usage.

Correct me if I'm wrong! Thanks.
 

Attachments

  • 25.jpg
    25.jpg
    171.7 KB · Views: 146
Yeah, my G19 has a Performance Monitor for the CPU/Ram usage and I'm seeing both cores of my CPU being used in G&K.
 
Please let us know what the problem was and how you fixed it.. if someone comes here through Google they're going to curse you.

"Hi, I have [problem X]"
"Fixed it, nevermind"
 
Basically, my patch version was 1.511 at the time when i was having those issues. I just hadn't bothered to check if the current version still had the problem. It uses multiple cores properly now, like MadDjinn said.
 
Huh. I've noticed pretty terrible turn times since G+K, and so I was looking at its CPU usage yesterday for the first time, and noticed that all 4 cores were at around 25-30%, so I assumed that the majority of its CPU time was spent in a single thread bouncing across all 4 cores.

I wonder if I have this option disabled for some reason - does anyone know how to change it?

Edit: I seem to have "MaxSimultaneousThreads = 8", so I guess that's not it. I wonder why it doesn't max out my cores then...
 
I just have a dual and both ping to 100% when I click Next Turn :P
 
I decided to try and quantify this, so I loaded up my most recent game for a couple of benchmarks.

Using Windows task manager I set the CPU affinity for the game so as to force it to use 1, 2, 3, and all 4 cores to compare the turn time.

In absolute terms, these numbers aren't bad, so you might say 'what are you complaining about?', but this example isn't all that taxing so the important point is the relative times (playing the renaissance scenario is a rather different matter; benchmarking that would take...a while).

So the numbers:
1 core - 46s
2 cores - 37s
3 cores - 37s
4 cores - take a guess...37s

Only in the 1 core case did it max out the core in question, which leads me to conclude that turn times aren't CPU bound beyond that. I did also try in strategic view, because some people apparently find that faster: 37s. So I would hope that means it's not GPU bound either.

Maybe all the AI calculations happen in one thread, and everything else happens in a second thread but isn't too intensive; that would explain why adding a second core gives a modest improvement.

Or maybe the game isn't correctly detecting all 4 cores. I'll try investigating further...

Edit: Okay, based on a very quick look with process explorer, I get the impression that the rendering[i/] is multi-threaded, so if graphics are your bottleneck then extra cores will help, but the AI calculations are done in a single thread, so if that's your bottleneck (as in my case) then more cores won't help; they need to be faster.

I'm still surprised though that trading up from a Q6600 (a five year old quad core) to an i5-3750k (a five week old quad core, near enough, which is far faster per core) brought me worse turn times; of course, I didn't try it on the new system before the patch on the 14th. At some point I may try G+K on my old system to see how it fares.
 
Back
Top Bottom