Home Trending Latest threads New posts Developer diaries

Performance Megathread

Jump to latest Follow Reply
  • 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.
Stellaris 52835 Bug Reports 33973 Console edition 1287 Suggestions 20886 Tech Support 3133 Multiplayer 390 User Mods 4761 Stellaris AAR (After Action Reports)
Show only dev responses
VoidEmperor said:
whats your gpu? Integrated Graphics or dedicated? Did you try repairing/reinstalling the game files?
Click to expand...
My MSI GeForce RTX 2070 VENTUS GP graphics card has I believe 8G dedicated ram. I reduced the graphics performance and it works better, I also play on a 1000 star card. I have already planned to increase my configuration for the future because obviously paradox is rather difficult to optimize its games, given what is happening at home for imperator (lag of 5 seconds every month)
 
klingonadmiral said:
Someone on reddit already did some test and 3.0 reached the year 2400 about 12 minutes earlier than 2.8.1 - from 2:09:43 down to 1:57:32 in a 600-star galaxy.
Click to expand...
This is stupid. It was never about reaching 2400 and thet figure is pointless without the galaxy size and configuration, total pops and playstyle used.
 
  • 1
Reactions:
klingonadmiral said:
600 stars, elliptical, 9 AIs, 2 advanced starts, 2 FEs, 2 Marauders



Galaxy-wide pops in 2.8 were 10915, compared to 6205 in 3.0



AI only.
Click to expand...
Works as intendent, problems start much later and at a far greater pop count.

At such low levels you wouldn't see much difference. Read several pages above for reference test data, with x5 x10 pops.
 
  • 1
Reactions:
PedroLuiz said:
hey, did anyone on this thread make a study of how much Habitats contribute to Lag compared to other sources?
Click to expand...

For a given number of pops I'd expect habitats to make little/no measurable difference to planets or ringworlds in terms of performance. Depending upon the specific algorithms they could be more computationally efficient.

I like playing with the Void Dwellers origin a lot and I've not noticed a performance different to using that vs planets, though I wouldn't expect an individual player empire to have much effect on total game performance.

PS On a non-performance note, I've found the Void Dwellers origin is a lot more limited in 3.0 which makes them rather less fun. For some builds they're even more powerful and I've gotten some pretty broken results with experiments but it feels to me that the number of good builds for them has been significantly reduced. Going from Void Dwellers origin to more standard planet based origin feels quite liberating.
 
  • 1Like
Reactions:
Aleriez said:
117 min * 60s = 7020s
200
Click to expand...
idk, i was tired and sometimes my brain bugs out
CPR said:
For a given number of pops I'd expect habitats to make little/no measurable difference to planets or ringworlds in terms of performance. Depending upon the specific algorithms they could be more computationally efficient.
Click to expand...
pops hurt performance on an exponential base depending on the number of pops in a planet + free jobs being bugged
CPR said:
I like playing with the Void Dwellers origin a lot and I've not noticed a performance different to using that vs planets, though I wouldn't expect an individual player empire to have much effect on total game performance.
Click to expand...
a positive or negative difference?
 
I will say pop growth change has done wonders for performance. The machine I'm using is a potato, so I still see a somewhat noticeable slowdown towards end game, but the lag is no where near as bad.

One issue I did notice is that when I'm assimilating pops, each time a new year rolled in, I would get hit with a massive lag spike. So something might be wonky there with ow the game checks things.

As for future stuff, I've been thinking on how pops affect performance. Correct me if I'm wrong but the calculation for each pop is:
-Check to see if they're employed
-Check to see if the job is most suitable to them
-Check to make sure they nutrition
-Check to make sure they have proper amenities
-Check to make sure they have proper consumer goods
-check to see if they have proper housing
-check to see what their living standards are
-check planet's suitability for pops habitability
-check to see if they are unhappy or happy
-check to see if they have ethics
-check to see if there is something that will result in them being more happy or less happy because of their ethics
-check to see if they belong to a faction
-check to see if their faction will result in them being more or less happy
-check to how much political power the pop has

That is a ton of calculations; especially, when the average developed world for the AI in end game is about 80 pops with the new system. I know people have been clamoring for better internal politics, I'm one such person. I feel if the devs aren't already considering it, that the bold italic stuff could be something that gets removed from base pop calculations. IMO add a new tab to colonies called VIP. On this tab you have influential planetary individuals. Obvious each ruler pops will generate one such VIP, but you could have calculations to determine how many other VIPs you get from each stratum (ruler, specialist, worker and slave) and maybe other things I've missed. In short, VIP are what decide ethics, faction approval and political power that might impact planet stability. Granted all calculation might not be equal on processing power and I suspect the 5 bolded and italicized items might be some of the more involved things. So getting those off of base pops, which might average in like the 20k range for a large galaxy at end game could be huge performance boosts if those are only being done at a tenth of their current rate. Pretty much, a continuing issue with system performance is that base pops do too much and maybe a new system could have them do less without ruining things.
 
  • 1
Reactions:
Is there anything that can be done for lag on the Species screen? Late game I end up with tons of species from migration and that screen drops to about 10 FPS (vs. a smooth 60 FPS for rest of the game).

It is markedly worse with the UI Dynamic Scaling mod and even worse when I turn on Planetary Diversity and Ethics and Civics. I can reliably reproduce that one screen chugging by starting a new came using the console `communications` command to to meet all empires and then showing galaxy species.

- Vanilla it is about 20-30 FPS
- UI Dynamic it goes to 15-20 FPS
- Planetary Diversity and/or Ethics and Civics drops it to 8-10 FPS

I'm assuming the issue is maybe all the portrait animations? I am playing on an RTX 2070 and lots of RAM and CPU so it doesn't seem like a thing that should cause it to die?

Are there any ways to disable animation there?

Are there other calculations happening there that could be contributing?
 
  • 2
Reactions:
Millbot said:
As for future stuff, I've been thinking on how pops affect performance. Correct me if I'm wrong but the calculation for each pop is:
-Check to see if they're employed
-Check to see if the job is most suitable to them
-check to see what their living standards are
-check planet's suitability for pops habitability
-check to see if they are unhappy or happy
-check to see if they have ethics
-check to see if there is something that will result in them being more happy or less happy because of their ethics
-check to see if they belong to a faction
-check to see if their faction will result in them being more or less happy
-check to how much political power the pop has
Click to expand...
Most of them are checked during different times throughout the month and dont affect performance that much
Check to make sure they nutrition
-Check to make sure they have proper amenities
-Check to make sure they have proper consumer goods
-check to see if they have proper housing
Click to expand...
I'm pretty sure those are made on a per planet/empire basis
 
  • 1Like
Reactions:
Kolath said:
Is there anything that can be done for lag on the Species screen? Late game I end up with tons of species from migration and that screen drops to about 10 FPS (vs. a smooth 60 FPS for rest of the game).

It is markedly worse with the UI Dynamic Scaling mod and even worse when I turn on Planetary Diversity and Ethics and Civics. I can reliably reproduce that one screen chugging by starting a new came using the console `communications` command to to meet all empires and then showing galaxy species.

- Vanilla it is about 20-30 FPS
- UI Dynamic it goes to 15-20 FPS
- Planetary Diversity and/or Ethics and Civics drops it to 8-10 FPS

I'm assuming the issue is maybe all the portrait animations? I am playing on an RTX 2070 and lots of RAM and CPU so it doesn't seem like a thing that should cause it to die?

Are there any ways to disable animation there?

Are there other calculations happening there that could be contributing?
Click to expand...
You might need to report this as a bug. I've never had such a thing and I had over 200 species and variants.
 
Kolath said:
Is there anything that can be done for lag on the Species screen? Late game I end up with tons of species from migration and that screen drops to about 10 FPS (vs. a smooth 60 FPS for rest of the game).

It is markedly worse with the UI Dynamic Scaling mod and even worse when I turn on Planetary Diversity and Ethics and Civics. I can reliably reproduce that one screen chugging by starting a new came using the console `communications` command to to meet all empires and then showing galaxy species.

- Vanilla it is about 20-30 FPS
- UI Dynamic it goes to 15-20 FPS
- Planetary Diversity and/or Ethics and Civics drops it to 8-10 FPS

I'm assuming the issue is maybe all the portrait animations? I am playing on an RTX 2070 and lots of RAM and CPU so it doesn't seem like a thing that should cause it to die?

Are there any ways to disable animation there?

Are there other calculations happening there that could be contributing?
Click to expand...
I've seen this too with UI Overhaul Dynamic (not Dynamic Scaling). Not sure if it's a mod issue or not.
 
I don't understand, why Stellaris developers won't use Factorio's developers' experience as reference (especially, because Factorio developers are kind enough to delve into technical details of the issues they've encountered and solutions they've found). Here they describe what was the issue and how they resolved it and here somebody else playtested and described how effective the change was.

As much as I understand that issue with all pops (correct me if I'm wrong) is that all pops are processed globally by the same thread. Why not do the threading of pops on the same planet and use something similar to "Wake-up Lists" where main thread that just collects results of the pops calculations on the planets, when they are finished and processes them? And if pop calculates something for entire country, just move it to the thread with pops that do same country-wide calculations. Such thing can be applied to anything that interacts only with limited objects. And even if something interacts with multiple objects, something similar to "Wake-up Lists" can be used to flag when calculations are done so main thread will collect and process the results.

I'll leave most important quote from their blog here just in case:
To solve this, when working, the threads don't wake the entities immediately, but instead add the wakeup requests to a list to be processed later. After all threads are done, the main thread collects and merges all those requests, and wakes up the entities in a deterministic way based on the group update order number.
Click to expand...
I feel that this solution (if applied correctly) can resolve a lot of performance issues Stellaris is facing right now. Pop reduction bandaids aren't really a solution, IMO.
 
  • 2
  • 1
Reactions:
WarStalkeR said:
I don't understand, why Stellaris developers won't use Factorio's developers' experience as reference (especially, because Factorio developers are kind enough to delve into technical details of the issues they've encountered and solutions they've found). Here they describe what was the issue and how they resolved it and here somebody else playtested and described how effective the change was.

As much as I understand that issue with all pops (correct me if I'm wrong) is that all pops are processed globally by the same thread. Why not do the threading of pops on the same planet and use something similar to "Wake-up Lists" where main thread that just collects results of the pops calculations on the planets, when they are finished and processes them? And if pop calculates something for entire country, just move it to the thread with pops that do same country-wide calculations. Such thing can be applied to anything that interacts only with limited objects. And even if something interacts with multiple objects, something similar to "Wake-up Lists" can be used to flag when calculations are done so main thread will collect and process the results.

I'll leave most important quote from their blog here just in case:

I feel that this solution (if applied correctly) can resolve a lot of performance issues Stellaris is facing right now. Pop reduction bandaids aren't really a solution, IMO.
Click to expand...
The problem is that factorio is a single game, and you can optimise that to the death for its specific mechanisms. The PDX engine, clausewitz, is used on many titles as a base upon which a game is built and as such, it's really hard/near impossible to tinker with the code. You could do it, but it would be very expensive and you would end up with something that would be fragile, you can't change or expand upon and require a lot of work each time a DLC is released.

it's a different business model of a game, built on an existing multimillion lines code base.
 
GnoSIS said:
The problem is that factorio is a single game, and you can optimise that to the death for its specific mechanisms. The PDX engine, clausewitz, is used on many titles as a base upon which a game is built and as such, it's really hard/near impossible to tinker with the code. You could do it, but it would be very expensive and you would end up with something that would be fragile, you can't change or expand upon and require a lot of work each time a DLC is released.
Click to expand...
One or another way game engine is always adjusted for new games, new DLCs and etc. There is absolutely no reason not to optimize game engine for a game that they are releasing. I'm pretty sure Clausewitz Engine that they're using for Stellaris and Clausewitz Engine they are using for Crusader Kings III are already very differnt, even if initially they were derived/forked from the same game engine.
 
  • 1
  • 1Like
Reactions:
PedroLuiz said:
pops hurt performance on an exponential base depending on the number of pops in a planet + free jobs being bugged
Click to expand...
Almost certainly not exponential. (If it was exponential, the game would've been grinding to an absolute halt vastly faster and more spectacularly than it was.)

Probably quadratic; I could see a particularly degenerate case ending up quartic.
 
  • 2
Reactions:
PedroLuiz said:
Most of them are checked during different times throughout the month and dont affect performance that much

I'm pretty sure those are made on a per planet/empire basis
Click to expand...
I'm aware the calculation aren't being done every day. Anyways for the bolded, even if we assume that they aren't applicable to gestalt pops. If we assume that a medium galaxy only has 10K pops that do those calculations once a month. That's still several thousand calculations that really don't need to happen.

Just creating characters that deal with ethics, factions and whatever else is needed for internal politics and that does mean coming up with something for gestalts. Could easily cut down on thousands of calculations that pops are making. It's not that each calculation takes up tons of processing power, it's that it's doing it for each pop and at a certain point that just ends up being murder on many people's machines.

Honestly for stability the only things where general pops should be a factor is their happiness based on whether they are employed, have food, have adequate housing, adequate amenities, living standard and if applicable, adequate consumer goods.
 
You must log in or register to reply here.
Top Bottom
Back