Performance improvement lategame. How to fix?

[A.J]

Yoyo,

Currently at year 2400 in the game and it has slowed down so much regardless of the game speed setting. I remember it was way better in the previous versions of the game. I'm wondering if there are any tweaks you can do to fix or get around this? I'm playing on a 800-star map for reference and the early game speed is completely fine.

It takes ages to do anything at this point and it's demoralizing to say the least.

[Darkrex]

so far i remember it used to be way worse with lags and stutters, yes its on the fastest speed it still feel like 1x speed later on but its by far not worse then it used to be. but then again it takes a lot of processing power to calculate all the moves but i too hope to see it speed up more

[A.J]

Originally posted by Darkrex NL:

so far i remember it used to be way worse with lags and stutters, yes its on the fastest speed it still feel like 1x speed later on but its by far not worse then it used to be. but then again it takes a lot of processing power to calculate all the moves but i too hope to see it speed up more

Performance was much better in 1.9

[tempest.of.emptiness]

No matter which version you play, performance is going to decline as the game progresses, because there are more pops/ships/whatever that the game has to process.

There's nothing you can do to change that fundamental fact, but you can ease the problem by doing things that either slow down progress or lower the ceiling on what is possible, such as:
1) Play on a smaller galaxy.
2) Reduce the quantity of habitable planets in the galaxy.
3) Decrease the end-game date.

Depending on your computer, you may also be able to improve performance by upgrading, although be aware that even the best computer is going to struggle with Stellaris in the end-game if you crank up the number of stars and habitable planets.

[Vyndicu]

Don't upgrade your PC yet. At least wait until the next patch to see if they fix it some more.

There are reports of high-end CPU struggling with Stellaris.

I have a CPU that can run a huge CPU-heavy Factorio game fast but micro-stuttering show up in Stellaris after only a few years in-game.

[Vale902]

There is nothing to do, unless you want some perfomance mods but they won't solve the problem totally, given it comes from the population system. You may look for a mod which deactivates habitats. AI players are used to build many of them and mixting species, so going forward with the game, your cpu has to deal with all the calculation that comes in a big part from Habitats.

[Darkrex]

Originally posted by A.J:

Originally posted by Darkrex NL:

so far i remember it used to be way worse with lags and stutters, yes its on the fastest speed it still feel like 1x speed later on but its by far not worse then it used to be. but then again it takes a lot of processing power to calculate all the moves but i too hope to see it speed up more

Performance was much better in 1.9

well yeah 1.9 had less features compared to now. see it like this, i got 1 plate with 1 potato on it and 1 plate with 10, which one would be faster to eat, plate 1 or 2?

[Vyndicu]

Originally posted by Darkrex NL:

Originally posted by A.J:

Performance was much better in 1.9

well yeah 1.9 had less features compared to now. see it like this, i got 1 plate with 1 potato on it and 1 plate with 10, which one would be faster to eat, plate 1 or 2?

One of the goal behind the 2.0 patch overall was to free up performance bottleneck (Stellaris dev blog source).

Yet at the same time ending up worse off post-2.2 despite changing spec hardware to match the newer, back then, generation of CPU. In fact I believe there were multiple reports that higher end, at least back then, CPU were struggling to even process a modest galaxy.


Even after Federation patch, I am still having micro-stuttering, early game, on a two year old PC.

On the same PC, I can run a huge CPU-heavy Factorio mega-base (image thousand flying logistics drones and launching two rockets at once).

Stellaris team still have some ways to go.

[Kotori Asobi]

The issue is the high count of pops lategame and the change into 1.1. of the next year trigger assimilation, which may take a while, if there are 100k pops in the galaxy.

Each single pop has a own biography it will be updated everytime something is happening like assimilation, ownership change etc.

In 1.9, there was a maximum of 25 pops per planet, now, there can be more than 200 pops per planet

[EleventhStar]

Originally posted by Darkrex NL:

well yeah 1.9 had less features compared to now. see it like this, i got 1 plate with 1 potato on it and 1 plate with 10, which one would be faster to eat, plate 1 or 2?

But isn't it the cooks fault if he stacks more potatoes on the plate than the plate can handle?

honestly its the biggest reason keeping me form re-trying this game: i'm pretty sure the last patch peformance increases only brought it back to a point _after_ i quit because of performance issues.

[Thjaffi]

this game is just using 2 cores effectively (in 2020).
For a really large scale RTS it's a shame. Paradox should upgrade their engine instead of pushing more DLCs. (i guess it wouldn't happen before Stellaris 2)

Large and fully habitated galaxies or long games don't make any fun with stuttering game.
And small and fast games dont make fun (for me) either.

[Targonis]

Originally posted by Thjaffi:

this game is just using 2 cores effectively (in 2020).
For a really large scale RTS it's a shame. Paradox should upgrade their engine instead of pushing more DLCs. (i guess it wouldn't happen before Stellaris 2)

Large and fully habitated galaxies or long games don't make any fun with stuttering game.
And small and fast games dont make fun (for me) either.

Exactly. A Ryzen 9 3900x SHOULD handle this better than a quad-core i7-7700k if each faction in the galaxy were spun off to its own thread(which would clearly benefit over the next five+ years as we see core counts increase.

[Zane87]

Originally posted by Targonis:

Originally posted by Thjaffi:

this game is just using 2 cores effectively (in 2020).
For a really large scale RTS it's a shame. Paradox should upgrade their engine instead of pushing more DLCs. (i guess it wouldn't happen before Stellaris 2)

Large and fully habitated galaxies or long games don't make any fun with stuttering game.
And small and fast games dont make fun (for me) either.

Exactly. A Ryzen 9 3900x SHOULD handle this better than a quad-core i7-7700k if each faction in the galaxy were spun off to its own thread(which would clearly benefit over the next five+ years as we see core counts increase.

If it would be even remotely easy it would have been done already. And no this is not a fanboy thing, this is a pure logical reasoning.

The performance especially in lategame is one of the big 3 huge issues of this game, and for me the biggest. There are a ton of threads and complaints about it, justifiedly, and one main reason for frustration with the game.

A lot people would buy more dlc if they enjoyed the game more, lategame lag directly decreases that joy, therefore without it the positive effect on sales would be at least substantial.

Companies are not stupid when it comes to making more money. Therefore it's a safe assumption that this already is a pretty big thing on their agenda, with small successes but only that. Therefore there has to be other reasons why it's still mostly single core actually (not even really dual core, the other guy probably just mistook his hyperthreading and 2 threads at 50% as 2 cores, it's mostly one).

The harsh truth is: a lot of games like this mostly use one core. Period. The reason for that is because meaningful hyperthreading is hard, read some dev blogs or tech talks for that, pdx even admitted so themselves recently.

I am as pissed as you but still, reality is that meaningful multithreading in certain architectures is very hard. Especially if the different mechanics all rely on each other, thus can't be calculated simultaneously. (And even then they would need to be synchronized across cores, which can be even worse for performance than just a single thread. It's hard, it's really hard)...

[Nightmyre]

The latest dev blog does explain this a bit more for non-coders. Basically - it's really not as simple as just saying "okay processors, split this task up into fifty threads and just run them all at the same time, i have the memory to handle it, don't worry!".

The long story short is that multithreading requires you to fundamentally rethink how you're coding your methods. You can't simply assume that everything will operate in the order you expect it to, and therefore you have to start thinking about concurrent processes and whether your method is theadsafe or not.

As a professional programmer with over a decade of experience, let me tell you that the number of real-world scenarios where people use multithreading is vanishingly small. You generally just don't need the extra performance, and when you do, you'd better have thought everything through from the beginning, 'cause trying to be like "okay this one thing is having performance issues, let's make it multithreaded" is extremely difficult to pull off.

[Kayden_II]

Originally posted by Nightmyre:

you'd better have thought everything through from the beginning

Yeah, it's Paradox fault that they haven't thought this through right from the start ( Stellaris V1.0 ). But isn't it more important that Paradox attitude of "release yesterday and fix tomorrow" had given them a **** load of profits and us **** load of bloated content via 11 DLCs and a zillion of updates ? And on top of that: The recent dev-diaries had a lot of BLA-BLA-BLA, but no actual hints in regards to actual results in regards to actual performance-optimisations, but guess what ? New bloated content ( eventually in form of DLC number 12 ) was already spotted instead. This makes any further tries in regards to actual performance-optimisations "certainly" "easier", right ?

"Hurray", that's Paradox how it's known since it went public !