• We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
You got to love a high horse. But point taken
Why would I need any high Animals? I simple speak about maters of fact.

The big issue is usually visualizing the issue for laymen.
 
Because Multitthreading is not a magical bullet that makes thing go faster.
After all, 9 Women can not get a child in 1 month either.
I agree but bad analogue; multithreading CAN speed up things if done properly and thoroughly while women can never give birth faster.
Also, it's "matter", not mater. I gotta smart ass that one since you been doing that mistake since I started reading the forums for more than 1-2 years.
 
  • 1Like
Reactions:
I agree but bad analogue; multithreading CAN speed up things if done properly and thoroughly while women can never give birth faster.
Yes, 9 women can produce 9 Children in paralell faster then 1 woman can do sequentially.
That still does not allow 9 women to produce 1 child in month.

And it is the perfect example for Multitasking. 9 Months. 9-900 Milliseconds. Slightly different orders of magnitude, but the same thing in the end.
 
In any case, it's a bad analogy because producing children is not a situation that can be multithreaded while Stellaris is, and quite trivially. The correct response to Ringworm's question is "because Stellaris is badly coded".
 
Hi All,

Not after a debate but just wondering if there is a reason Stellaris does not utilise my CPU? Rarely gets above 40% and never above 50% . I have a load of resources available why isn't it using them?

Many thanks

Have you checked the load at each core? I have one core at 100% and the other 7 around 15%
 
In any case, it's a bad analogy because producing children is not a situation that can be multithreaded while Stellaris is, and quite trivially. The correct response to Ringworm's question is "because Stellaris is badly coded".
This is trivial if the tasks to be completed are not dependent on each other. So, you can run a thread to calculate the pop-job allocation in one thread, and trade in another - except the amount of trade is dependent on the amount of pops producing trade value... Some of the tasks may be interconnected without actually needing to be. In the example above, the trade value in a day might be dependent on the day before, at which point they could run separately.

How about overclocking a woman to have 2 or more babies at a time? I need more of that in my Stellaris games dammit!.
/thread derailed
This is called twin/triplets/quadruplets(/etc)... And I guess that would be kinda like running multiple instances of Stellaris on the same PC...
 
is there an easy way to fill all vacant jobs ? 'cause in my about 40k pops game the pace become very fast very much slower, and what it would be with filled jobs and if the lag for resettlement also goes away. But I also think I'm finished with my game ... the pain and torture became too much
 

Attachments

  • 2501.02.13_final.sav
    2,5 MB · Views: 6
  • littlemod2.7z
    30,7 KB · Views: 6
Hi,

I have read quite a bit of this thread (not all of it, I admit). As far as I understand the performance issue had been widely discussed. Still it seems to me that nobody tried to provide a second-best solution. Ater reading, I still have a couple of questions

1) Let us assume we want to improve late game performance at the cost of giving up some features; which of these fix would you suggest as the most effective?

- A) Do not use all the DLC. Still, which DLC do you feel affects the most late game performance?
- B) Play a small galaxy, ok... how much smaller than the average one?
- C) Someghing else: please articulate.

2) Do you think giving up stellaris soundtrack can improve performance? Usually I listen my music while playing and I would be happy to give up in game sounds/music completely; unfortunately no such option is available. What do you thind about it?

I thank you in advance for any suggestion.
 
This is trivial if the tasks to be completed are not dependent on each other. So, you can run a thread to calculate the pop-job allocation in one thread, and trade in another - except the amount of trade is dependent on the amount of pops producing trade value... Some of the tasks may be interconnected without actually needing to be. In the example above, the trade value in a day might be dependent on the day before, at which point they could run separately.

We have tasks that are grouped so that they are trivially completed without being dependent on each other. We call them "planets", and they create a situation that can be considered embarrassingly parallel. 200 colonized planets should trivially scale to 200 entirely separate threads (if desired) that only need to by synced to a few global gamestate conditions once during the intra-month period. There's zero reason for pop jobs to be slowing down the game with free CPU time available.
 
We have tasks that are grouped so that they are trivially completed without being dependent on each other. We call them "planets", and they create a situation that can be considered embarrassingly parallel. 200 colonized planets should trivially scale to 200 entirely separate threads (if desired) that only need to by synced to a few global gamestate conditions once during the intra-month period. There's zero reason for pop jobs to be slowing down the game with free CPU time available.
The "few" will break this back. See Amdahl's law (which you probably know, but should keep in mind that the percentage to spoil efforts is rather minimal). Also, overhead, synchronisation issues and so on and so forth. There is a reason why the most programmers among the forists don't advocate this approach, even if they are otherwise rather vocal about the current state of affairs. Based on the stuff @GnoSIS wrote a few pages back, this approach would mitigate the problem but not solve it. You'd simply open more windows which would become clogged in time.
I'd wager that a switch to an event-based system (or hybrid) could yield better results. Of course, you'd have to try. And gauge which approach will result in less overall mess due to unintened side effects.

Edit: Making a per-sector split may be a viable middle-of-the-road solution.
 
The "few" will break this back. See Amdahl's law (which you probably know, but should keep in mind that the percentage to spoil efforts is rather minimal). Also, overhead, synchronisation issues and so on and so forth.

Considering they only need to be synced once per in-game month, it is a trivial amount of overhead. Obviously I'm not saying a galaxy with 1000 planets should have 1000 separate threads, just that it would be possible. Reasonably speaking having somewhere around 10-40 would probably be enough to balance the load adequately with near non-existent overhead.

Based on the stuff @GnoSIS wrote a few pages back, this approach would mitigate the problem but not solve it. You'd simply open more windows which would become clogged in time.
I'd wager that a switch to an event-based system (or hybrid) could yield better results. Of course, you'd have to try. And gauge which approach will result in less overall mess due to unintened side effects.

Agreed, but the point is that the reason Stellaris is not currently using all available CPU time efficiently is because it wasn't coded to do so. Stellaris is clearly capable of running significantly better using the current system. Certainly a better system could scale better, up to 10k systems or what have you, but that's not currently a requirement for the game.
 
Agreed, but the point is that the reason Stellaris is not currently using all available CPU time efficiently is because it wasn't coded to do so. Stellaris is clearly capable of running significantly better using the current system. Certainly a better system could scale better, up to 10k systems or what have you, but that's not currently a requirement for the game.
While the first part is nothing to loose sleep over, thats the bit which makes me leery.
I have made the experience that in a drive to use the most of a resource that can backfire badly. I had a colleague who complained that his minimum-wage gofers had half-hours of downtime and since that was a strain on his budget, that some work be provided for them. It backfired badly (means: a few quite-a-bit-more-than-minimum-wage people had to put in extra hours), so it was a net loss.

I would duly recommend against making maximum CPU usage at all times a goal instead of an effect.
 
Dear All,

I understan that all the above debate on parallel computing is very stimulating. Nevertheless, while we wait for new improvement from the Devs, I still wonder if there are some "tweaks" or workaround that can help the average player to deal with the late game slowdown. I am really curious to know if anybody has an answer to my former questions. Just for your convenience, let me quote myself.

Best

Hi,

I have read quite a bit of this thread (not all of it, I admit). As far as I understand the performance issue had been widely discussed. Still it seems to me that nobody tried to provide a second-best solution. Ater reading, I still have a couple of questions

1) Let us assume we want to improve late game performance at the cost of giving up some features; which of these fix would you suggest as the most effective?

- A) Do not use all the DLC. Still, which DLC do you feel affects the most late game performance?
- B) Play a small galaxy, ok... how much smaller than the average one?
- C) Someghing else: please articulate.

2) Do you think giving up stellaris soundtrack can improve performance? Usually I listen my music while playing and I would be happy to give up in game sounds/music completely; unfortunately no such option is available. What do you thind about it?

I thank you in advance for any suggestion.
 
It's highly alarming to read your post suggesting nobody seems to be offering workaround solutions, whilst also admitting that you haven't read through the available material.

On the first page of this thread there are at least three posts that should provide food for thought:
https://forum.paradoxplaza.com/forum/index.php?threads/performance-megathread.1253705/#post-25868197
https://forum.paradoxplaza.com/forum/index.php?threads/performance-megathread.1253705/#post-25870082
https://forum.paradoxplaza.com/forum/index.php?threads/performance-megathread.1253705/#post-25870654