Performance Megathread

[Varileztra]

Yeaaaa, my game runs like a snail starting between 2500- 2600, no matter if MP or SP.

3 seconds per day... did some quick maths in bed when i was falling asleep last night... would be 90 seconds per month, 1080 seconds or 18 minutes per year... over about 600 more years for a long campaign would be 10800 minutes ... 180 hours...
and that does not account for stuttering and game freezes when larger events pop or the game saves.

Meanwhile i look at my computers ressource screen and see, that 4/5 of all my ressources are not being utilized. Mate told me today morning, that for him in 1 cpu core is fully used and a second one a bit, the rest sat at 0% usage.
I always thought hearts of iron was bad after 1945 without a mod to not have those small countries spam out divisions for no reason, but this is way worse and its an absolut interest and mood killer.

 

[MojaveDreams]

Yeaaaa, my game runs like a snail starting between 2500- 2600, no matter if MP or SP.

3 seconds per day... did some quick maths in bed when i was falling asleep last night... would be 90 seconds per month, 1080 seconds or 18 minutes per year... over about 600 more years for a long campaign would be 10800 minutes ... 180 hours...
and that does not account for stuttering and game freezes when larger events pop or the game saves.

Meanwhile i look at my computers ressource screen and see, that 4/5 of all my ressources are not being utilized. Mate told me today morning, that for him in 1 cpu core is fully used and a second one a bit, the rest sat at 0% usage.
I always thought hearts of iron was bad after 1945 without a mod to not have those small countries spam out divisions for no reason, but this is way worse and its an absolut interest and mood killer.

This is the reason why late game performance should be king priority. There is no fun on small maps for me and i am sick to death of quitting games because it is just too slow to do anything

 

[KingAlamar]

Seeing that feature-updates tend to break the state of the game when it comes to performance I think that the "POPS & JOBS" rework needs to be looked at again and redesigned so it's not nearly as impacting as before.

At the heart of the problem was the good-on-paper, terrible-for-performance decision to model each-and-every POP like they were it's own person so-to-speak while at the same time drastically increasing the number of POPs in the game. One or more things that could [should?] be done are:

• Cutting the number of POPs && JOBs by 50-75%. For example "rural" districts [mines, farms, energy] should provide 1 job & 1 housing instead of two. Fully upgraded buildings that provide jobs should provide base 1 job + 1 job per building upgrade. Similar cuts can be made to City districts, non-standard districts, etc.
• Reducing available "districts" naturally occurring per planet to max of 20. Reduce the number of potential "buildings" to HALF the number of current Districts.
• Limiting the number of "artificial" colonies that can be produced by the AI --- habitats & Ring Worlds for example. The only time AIs should be able to take "habitats" for example is when a CUSTOM race with the voidborn (??) origin is allowed to spawn. Then that AI race should be allowed to pursue habitat and potentially Ring World techs if & as appropriate.
• No routine job or faction shuffle allowed. For Jobs check once per month [per planet] -- Employed POPs will ONLY check for PROMOTIONS -- Unemployed will take ANY open job. Factions will be more restrictive -- once born into a faction you stay in the faction. Events can allow for faction reassignment including things like Edicts expiring, being nerve stapled, changes in "species / citizenship rights", being mind-controlled-by-lasers, etc.
• Rework "factions" top to bottom so that faction membership does not have to be traced all the way back to the individual POP.
• Rework "species" top to bottom so that species membership does not have to be traced all the way back to the individual POP.
• Limit the species-bloat by banning Genetic modification, Xeno Compatibility, and events that would create new species for the AI. Updates to PSI Ascension could follow the species "galaxy-wide" so that each species ascends as a whole no matter where they are.

While I find most of the above changes distasteful the combination of the above should give the game's main-thread the breathing-room it needs so that the DEVs can add new features without that breaking the performance again.

 

[Thinghunter]

What about changing to a pop system like in Victoria 2? I think this could solve the lag issues and fix the pop growth at the same time.

Victoria 2 has the same scope as stellaris:
- about 3000 provinces -> even in 1000 star system galaxies you wont have that many planets/habitats for a very, very long time
- about 200 different cultures and like 10 religions which makes 2000 possible cultures as pops get seperated by culture and religion-> a galaxy with 30 empires, all enabled extra species like FEs and pre FTL civilizations has ~ 70 species at max- even with a lot of gene modding and xeno-intercompatibility it will take very long untill you reach 2000 species
- about 20 jobs in 3 strata -> stellaris has much more jobs, especially with mods, but not all empires have access to every job, so in fact you have ~20 different jobs per empire

To sumn it up: Victoria is an almost 10 years old game providing the same amount of different pops as stellaris does on the biggest possible galaxy with no lag at all and a far better simulation of the population (pop growth, assimilation, migration). Switching to a Vicky2 like could enhance the gameplay of Stellaris by a lot and would especially remove the tedious micromanaging aspects of stellaris. Of course, changing to this system would require major changes- I think even more than the jump to 2.0- but in my opinion it would be worth it.

 

[KingAlamar]

To sumn it up: Victoria is an almost 10 years old game providing the same amount of different pops as stellaris does on the biggest possible galaxy with no lag at all and a far better simulation of the population (pop growth, assimilation, migration). Switching to a Vicky2 like could enhance the gameplay of Stellaris by a lot and would especially remove the tedious micromanaging aspects of stellaris. Of course, changing to this system would require major changes- I think even more than the jump to 2.0- but in my opinion it would be worth it.

I would assume that it would be a much larger job to scrap the existing models and implement Vicky2 onto the current Stellaris engine?? Then again I am calling for a rework of Species & Factions -- as they relate to jobs esp. -- so we don't need to count every single POP in the game.

It would be interesting to see what Vicky2 does and compare that to how Stellaris is implemented. I've got a BAD feeling that Stellaris has / had a lot of slopped-together-code and effects as opposed to being DESIGNED from the beginning to be this type of game it is currently which explains WHY it's behaving the way it is.

 

[Thinghunter]

I would assume that it would be a much larger job to scrap the existing models and implement Vicky2 onto the current Stellaris engine?? Then again I am calling for a rework of Species & Factions -- as they relate to jobs esp. -- so we don't need to count every single POP in the game.

This is basically the Vicky2 approach- but instead of lowering the amount of pops, you increase them up to the real world pops (1 pop ingame = 1 human in real life)

 

[GnoSIS]

This is basically the Vicky2 approach- but instead of lowering the amount of pops, you increase them up to the real world pops (1 pop ingame = 1 human in real life)

Yea, but with this way they lost all performance gains from grouping pops. In Victoria a pop is moddeled in the same manner but also has a number that represents population so you can have 10.000 workers being managed as one pop. So the choice to deal with DISCREET pops was totally wrong. It was ok while we had tiles and only 25 per planet and no jobs.

I and many others have outlined solutions here many times. If they don't re-engineer the system (which is expensive) they need to reduce pop counts/growth in game and boost pop productivity/efficiency instead. They also need to make the universe and war more hostile to life. Right now pops are really immortal 99% of the time.

To recap & summarise:

1. Hard caps per colony based on size or other factors, even tech.
2. Mid game/late game pandemic crisis that is repeatable and triggered once total galactic pop count exceeds a threshold.
3. fast growth on start, non existant growth near colony cap - as all populations behave in biology. To fix the slow start that is the first 100 years, but keep the game rolling after 2350.
4. bombardment should be x10 more lethal for pops.
5. army invasions should also kill pops!
6. More planetary events and ecological events/disasters to kill pops. System level events as well.
7. Abstract pops inside megastructures, without incuring job processing.
8. More static infrastructure that produces resources without jobs and an alternative tile unlock system/district exploitation system.
9. Simplify the job system.

Alternatively we can all wait 10 years when we'll all have computer chips at the angstrom scale to play with, but even then, this piece of software would play up to year 2800 at most.

 

[Thinghunter]

Alternatively you could reduce the number of species:
- On Galaxy generation, include a slider of how many different species you want. If you e.g. put the slider to "low", a species is used for 2 empires.
- Make a system which automatically changes the subspecies of a species to the dominant one, so you do not have like 10 subspecies per species, but at max 1 or 2
- Introduce a "mixed" species for minorities without distinct traits. This species includes all pops who are under a certain threshold empire wide or planetary wide (1% or 5%). These "mixed" pops are used for migration and jobs, but you store the real species numbers in the background. Once a species is above the threshold, the species gets fully represented, otherwise not.

This could reduce the number of species per planet/empire by a lot- if it is possible to cut the species relevant for job calculation & economy by 90%, you should get a huge performance boost.

 

[GnoSIS]

@Thinghunter they said that it doesn't have any impact on processing lag.
The gui lag, says otherwise though not to say anything about managing it all.

 

[caffran]

Well, I rolled back to version 2.6.3 I think it was, and it's like night and day.

The slowest speed (x1) from 2.6 was quicker than the 'fastest' (x3) speed on 2.7, and I am utterly baffled as to how that is possible. how can the galactic senate be using so much resources that it slows the game down that much?

I'm actually asking.

 

[Sam Illingworth]

Yeaaaa, my game runs like a snail starting between 2500- 2600, no matter if MP or SP.

3 seconds per day... did some quick maths in bed when i was falling asleep last night... would be 90 seconds per month, 1080 seconds or 18 minutes per year... over about 600 more years for a long campaign would be 10800 minutes ... 180 hours...
and that does not account for stuttering and game freezes when larger events pop or the game saves.

Meanwhile i look at my computers ressource screen and see, that 4/5 of all my ressources are not being utilized. Mate told me today morning, that for him in 1 cpu core is fully used and a second one a bit, the rest sat at 0% usage.
I always thought hearts of iron was bad after 1945 without a mod to not have those small countries spam out divisions for no reason, but this is way worse and its an absolut interest and mood killer.

Well that’s depressing. I came to thread to ask what sort of CPU people are finding gets the best performance for Stellaris (more cores or faster cored), but seeing your post I thought I’d check my own utilisation. It’s not pretty:

To be fair, you can see when it auto paused you can see an effect on every core, it’s just that there’s clearly one process that’s a bottleneck for the rest.

 

[Varileztra]

Ye, something has to be done about that bottleneck asap,
5 AI and me left on a medium map, year is 2440 and the game just ground to a freaking halt. Its not"unplayable", but its unenjoyable. There is like no notable progress happening on the map anymore, having a freaking science ship travel just one system takes a minute at best. earlier i waited for a good half an hour for my fleets to make a 1000 days travel accross my empire
it feels so incredibly unresponsive, just because everything is happening in this super slowmotion.
the fact that i synthed and even without robot assemblys i now have somewhere between 6-10 pop growth from robots alone on every single freaking planet doesn't help either. Im spamming all my avaiable district slots on all my planets and yet i dont seem to make any progress on getting away fro my unemployment rates as it feels like, because in real time i waited an hour and ingame maybe three years passed, so at best i finished up like 5 districts, which doesnt really help when i have a good 10 planets far beyond their capacities with over 100 pop, that i need to reshuffle
(also thanks at this point for the nice Greater than ourselves edict, that got rejected in the galactic community as every AI voted against it and i then had to wait over 5000 days to repropose it... it got rejected about 4 playtime hours ago and the last time i checked i still had to wait for over 1000 days.

And just to add how they look for me...

On an unrelated note
i probably should restart my pc tonight

 

[Sam Illingworth]

Ye, something has to be done about that bottleneck asap,
5 AI and me left on a medium map, year is 2440 and the game just ground to a freaking halt. Its not"unplayable", but its unenjoyable. There is like no notable progress happening on the map anymore, having a freaking science ship travel just one system takes a minute at best. earlier i waited for a good half an hour for my fleets to make a 1000 days travel accross my empire
it feels so incredibly unresponsive, just because everything is happening in this super slowmotion.
the fact that i synthed and even without robot assemblys i now have somewhere between 6-10 pop growth from robots alone on every single freaking planet doesn't help either. Im spamming all my avaiable district slots on all my planets and yet i dont seem to make any progress on getting away fro my unemployment rates as it feels like, because in real time i waited an hour and ingame maybe three years passed, so at best i finished up like 5 districts, which doesnt really help when i have a good 10 planets far beyond their capacities with over 100 pop, that i need to reshuffle
(also thanks at this point for the nice Greater than ourselves edict, that got rejected in the galactic community as every AI voted against it and i then had to wait over 5000 days to repropose it... it got rejected about 4 playtime hours ago and the last time i checked i still had to wait for over 1000 days.

And just to add how they look for me...
View attachment 581619



On an unrelated note
i probably should restart my pc tonight

Perhaps the solution is to play four games simultaneously? Hmm...

 

[CPR]

I've not even played 2.7 - I basically gave up due to the poor performance. I've been looking around to see if there's any other interesting games out there but none have sparked my interest yet.

 

[mercav]

I just got an SSD to improve a different games performance (it has a lot of loading screens) and was wondering if anyone knows if moving Stellaris to the SSD would help late game performance. Can anyone help me with this?

 

[Sam Illingworth]

I just got an SSD to improve a different games performance (it has a lot of loading screens) and was wondering if anyone knows if moving Stellaris to the SSD would help late game performance. Can anyone help me with this?

I highly doubt it. I think the CPU is the limiting factor, not storage. I have it running from SSD and on a large galaxy late game I’m still waiting around 5 seconds per day.

 

[grommile]

I just got an SSD to improve a different games performance (it has a lot of loading screens) and was wondering if anyone knows if moving Stellaris to the SSD would help late game performance. Can anyone help me with this?

As far as I know, the only disk access Stellaris makes during play is saving the game.

 

[mercav]

Thanks gommile and Sam

 

[Varileztra]

can at the very least confirm, that having it on an SSD if anything, does not make a huge difference as i have a designated games only SSD in my pc and...ye.

 

[DiscoRay]

I think the CPU is the limiting factor, not storage.

CPU per say isn't the problem, it's more the Stellaris.exe file not making full use of all cores/threads available. Generally just 1 core takes all the load while the rest sit and watch.

I run the game from a Samsung NVME drive, and can also confirm that apart from the initial loading, there isn't any noticeable performance gain. Open world games the likes of Far Cry, Shadow or Mordor, Witcher III etc. get more benefits running from an SSD/NVME drive. (Along with a meaty GFX card!)