Performance Megathread

[Aurelius789]

Well, for a time Zen seems to be easily best option on market.

Closest benchmark to stellaris I found is civilization turn time benchmark (it is also CPU heavy simulation) :

As you can see, now is probably one of best times to buy new cpu if you want the game to be as fast as possible. Looks like there is 20% increase in single core preformance - this is most important statistic for stellaris speed.

$300 Ryzen 5 5600X easily beats ANYTHING on the market at the momment, awesome bang for your buck.
$550 Ryzen 9 5900X is even a bit faster, pretty much the best single core speed you can get now.
$800 Ryzen 9 5950X for gaming it is not worth it imo, single core is nearly same and you just get few more cores, that are useless for most people.

 

[Xmostera]

I would like to share one of the screenshot that they tested Paradox interactive : CK 3 performance in one of the chinese video.
Source:

Since CK3 has already changed their game engine, it was unexpectedly fast.

 

[Xmostera]

It literally only just came out yesterday!

Anyway, I'd be very surprised if the Zen 3 based CPUs are not the fastest option available. I'd also suggest that for pure gaming, the Ryzen 5600X is your best option and you'd be better off putting any money saved in a faster GPU instead.

Yes, pretty excited, because the performance would be so great with its low memory latency as well has bigger cache, and we can use higher ram speed with budget

 

[Ghost_Tomb]

Hey all,

Long time lurker here, in short thanks to my job I've gotten my hands on a AMD 5800x as a system upgrade.

I've not got the ram fully optimized yet but I have done some basic tests with it and my old system.

I did this because no one tests with Stellaris and what CPU tests we do see normally only compared with the last 2 gens.


My 6 year old system

4970 *not K as that one died and I had to switch to a backup

32GB ram DDR3 2400MHz

This is running the latest test version of the game (2.8.1 2190) , with the -dx11 parameter set

running off a regular SATA SSD at 500MB/s *I've tried it on that and a 3000MB/s M.2 and saw only a few seconds improvement for start times and zero change actual gaming performance.


New system is a 5800x, 32GB DDR4 3800MHz and same SATA SSD as before.


For perspective I ran a few simple CPU benchmarks.

WPRIME scores 32M / 1024M	i7-4790 (in seconds)	AMD 5800x (in seconds)
1 Thread	37.071 / 1150.203	28.274 / 906
4 Threads	10.333 / 325.505	7.51 / 233.015
8 Threads	7.017 / 210.17	4.573 / 131.097
16 Threads	n/a	3.62 / 88.212
		*Lower is better

I didn't have any really late games to try it with sadly, but I was playing on a 1000 star galaxy, with max Civs, fallen, marauder etc only thing that would improve performance is habitable planets being set to 0.5

	i7-4790 (in seconds)	AMD 5800x (in seconds)
cold start using resume	45.19 seconds	30.26 seconds
1 year from 2303-2304 *Fastest	1 minute 47 seconds	1 minute 19 seconds
1 year from 2304-2305 *Fastest	1 minute 44 seconds	1 minute 17 seconds

Do note that for the time tests on the 5800x only 18% of the total CPU power was in use *but I believe all threads were doing something. and 18% is nearly 3 cores at full load.


In a related note, as no one tests or benchmarks Warhammer 2,

	i7-4790 (in seconds)	AMD 5800x (in seconds)
Ironhide turn 1 mortal empires	19 seconds	11.5 seconds
Bonerattlaz turn 88 to 89 (also ME)	35 seconds	24 seconds

I was going to make a forum post but this is the performance thread

Hope folk find this useful.

 

[Moah]

when i do "nothing" but watch the game zoomed out my month takes 25 seconds.
but as soon as i just open a planet it takes 1 minute.

2.81. beta branch. year 2381 - 5304 pops. i am glad to see a month tick in 25 seconds, really am.

but something does happen when in planet view.

edit* also in starbase buildings/ modules & shipyard menu

The issue with planet view is known and a fix is coming.

 

[CPR]

The issue with planet view is known and a fix is coming.

That's nice.

Any idea if my experience of Stellaris 2.8.x on my 6 core MacBook Pro is normal or not? Basically, performance starts out better than before but rapidly deteriorates within a few decades. Overall CPU usage is much higher but performance is lower. I've not seen anyone else comment on Mac performance for 2.8.

 

[GnoSIS]

I'm the one who reported the bug with the planet view. Keep playing if you experience this. In my games the planet view lag went away after a few years.

Also, it doesn't matter if the game is paused or not, it has nothing to do with game proccessing.

 

[George_L]

Iam running a i-5 6500k with a rtx2060 and 16gb of ram.My fps on late game is around 40fps (besides doomstack battles) which is quite ok for me but when i try to move around the map the game behaves weird(stutter?).Since iam not planning to upgrade my pc the only thing i could do at this point is going from win7 to win 10.Anyone noticed any differnce in such transition?

 

[RONRONlepigeon]

I am playing on 2.1.3 and I don't have any performance issue, we juste passes 2500
There is a little lag but exept that everything is pretty good
ai is doing allright and I am still fighting the contingency

 

[WJLIII3]

Iam running a i-5 6500k with a rtx2060 and 16gb of ram.My fps on late game is around 40fps (besides doomstack battles) which is quite ok for me but when i try to move around the map the game behaves weird(stutter?).Since iam not planning to upgrade my pc the only thing i could do at this point is going from win7 to win 10.Anyone noticed any differnce in such transition?

I see the same, right around 2400 every time

 

[GnoSIS]

Well, I just reached the endgame, year 2430, medium galaxy, 0.75 habitability. There are about 18.000-20.000 pops, I have 10.000 and 85 colonies.

Guess what, the game still has performance issues, and the year proccessing rate, along with much of the UI has become extremely slow. I've stopped playing because I need to play for 4 hours to get the same passage of time as 1 hour some decades before - a 400%-500% slowdown in day proccessing.

CPU graphs show that the utilization of a single core is always 100%, but my other cores are not even 50% utilized. The patch was a much needed improvement and I want to indeed congratulate Paradox for their engineering effort, but this game still needs **subtantial work** or re-design.

1. Some of you, that bought recently a new $500-$600 CPU, might not be facing this. In this case, they need to raise the minimum specs of the game with whatever reprecusions this might have. As for me, my next upgrade will be around Jan/Feb. My current CPU is indeed old: i7-6700HQ, but I firmly believe that it's not the CPU's fault. I have no issues with other games, even newer ones, and other PDX games. Even if I had done the upgrade today, I'd still feel the need for this issue to be addressed, because it would mean more people being able to play and a larger community, and my powerful system would be able to play on bigger and grander galactic settings. So no, I'd still be here and not silent, nor turn against gamers with old systems and tell them to upgrade.

2. There is a hint about revisiting pop growth in the next expansion/version. Perhaps they will make it so that even by year 2500, with my settings above, you never reach 20.000 galactic pop counts. That would also be a satisfactory solution. I'd expect Y2500-Y2550 a reasonable timeframe to design and support the game for.

3. They might be still working on further enhancing multithreading. While I would love this to be the case, there's a point of diminishing returns, and the fact that with double the pops in a few decades, the game will still experience severe slowdown. But if they offload more tasks from that main thread, it will be a huge relief for all user systems.

4. I, and perhaps you, can **forget* about playing large or bigger than large with > 1.5 habitability. Even with a modern system, you will not have a performant game. If you have a save with those settings please share it. I expect it to have around 45.000 to 60.000 pops in it. There's no effing way that such a save would be playable, even on an Ryzen 9 5900X and especially for normal empires with trade and gateways - I challenge YOU!

5. Same old solutions apply: Mods, Genocide, stopping growth, colossus, playing gestalts (no trade), releasing the Galactic Krakens (or/and letting them do their job) etc. Of course a full game is possible with small galaxy and low habitability - it takes substantial time to reach the same pop counts and for the slow down to settle in. I'd suspect it would happen around Y2550, but I would need to try it out.

 

[Thinghunter]

I think a good idea for a better performance would be to (1) reduce the number of species (and subspecies) and (2) reduce the number of jobs on a single planet/in an Empire.

1) Reducing the number of species could be done in several ways:

1. Add a slider for the number of spawned species in addition to the number of Empires. Lets say you play with 20 Empires, 2 FEs, 5 pre FTL species & with the slider for the number of species set to 10, you get several Empires, pre FTL empires & co. sharing one single species (same name, same traits etc.) like the UNE & Empire of Man. This could explained e.g. by different origins giving a backstory why several empires share the same origin- like a civil war in a advanced empire where no side could win.
2. Reduce the number of trait picks for a species and instead add super traits which merge lower tier traits so you still can make very specialised species.
3. Give a huge penalty for having the same species with different traits in an empire or make a super fast conversion to the better trait selection.
4. Remove the traits for climate preferences and instead manage that via the adaptive trait. This means every planet gets a different base habitability depending on the planet type and the habitability is determined by the adaptibility of the species- perhaps in 7 tiers from very low to very high. I know, this reduces the complexity by a lot, but once you reach the lategame, habitability hardly matters and it might be of benefit to reduce the games complexity a little for a performance boost.
5. Make a trait rework and reduce the number of traits (e.g. merge the research traits Intelligent, Natural Engineers, Natural Physicists, Natural Sociologists, Erudite to a scale of traits ranging from 0 to 3 [0 would be no trait aka the basic research speed] and each species can only have one of these traits).

2) Reduce the number of Jobs:
This is very important as the automatic assignment of jobs scales exponentially with the number of species. Reducing this could improve performance significantly!
I proposed an idea for that in dev diary 190:

...

1. Remove the Crystal Miner, Translucer & Gas Extractor jobs & their corresponding buildings. Instead:
   If a planet has deposits for Exotic Gases, Rare Crystals & or Valatile Motes, the planet gets additional Mining District Slots & the production of Miners is changed so that the Miners produce an additional small amount of those special ressources (that amount could be depending on the size of the corresponding deposit).
   
   
2. Merge the Chemist, Gas Refiner & Translucer to an Industrial Worker job responsible for converting minerals into more specialized ressources. Determining what is produced by that job could be done via an empire wide output ratio (e.g. by making a slider or ascertaining the direct values). This would be in my opinion work very well because every job takes has the same input/output ratio & matrerials. You would save 2 buildings + Jobs & you would be able to tailor the production directly to your needs.
   
   
3. If you want to be very bold, merge the Artisan & the Metallurgist and decide the output ratio by a slider (empire wide/sector wide or planet wide).
   
   
4. It may very well possible to merge the jobs from changes 2 & 3 with a unified slider for all output so the number of jobs & buildings get reduced again. Reducing the number of jobs would make it easier to govern an empire for both players & the AI without reducing the depth of the gameplay as you create as many Artisan jobs as you have to, but as many Metallurgists as possible in any given situation (the same is true for the Chemist, Gas Refiner & Translucer which get created as few as possible). Last but not least, less jobs may very well increase the game performance substancially.
   
   
5. ...

Most of my ideas have the goal to reduce the complexity a little, while improving the performance, which is very important in the lategame.

 

[GnoSIS]

@Thinghunter the idea to reduce jobs, has been discussed to death. What we need is to reduc pop count, and increase job effects.

For a game that wants to have pops shine and have them as the back bone of progress, stellaris does a bad job on enhancing job output. there's not even an end game repeatable for it. idea being you stop growing pops, but you increase your pop efficiency.

 

[Thinghunter]

@Thinghunter the idea to reduce jobs, has been discussed to death. What we need is to reduc pop count, and increase job effects.

For a game that wants to have pops shine and have them as the back bone of progress, stellaris does a bad job on enhancing job output. there's not even an end game repeatable for it. idea being you stop growing pops, but you increase your pop efficiency.

I think it is more important to reduce the complexity (Species & Jobs) than the number of pops. The main reason is, that the number of species has a bigger influence on game performance than the number of jobs & pops:
The complexity of Job assignment highly depends on the number of Jobs & Species & Pops on a planet. As far as I understand, the performance scales very bad (exponentially) with number of species, but the number of pops does in worst case scale linearly (and less if you do an effective binning of the different pops which gets much more effective if the bin size is big & the number of bins is small- hence less different species = performance improvement). Same is true for the number of different jobs.
Why is the binning important? It enables parallelization! And i think you can see every planet & empire as a bin -> better possible multithreading

For me a reduction of the number of pops would be a game breaker- i want more pops, not less.

About the enhancement of job output: There are already several late game technologies enhancing the output per job and i think adding several repeatable more would be convenient (e.g less input per job needed ...).
Additionally, some jobs, could indeed deserve some love. E.g. Soldiers, for my part, are highly non efficient (i never build fortresses). When i want to increase my fleet, i build star bases but not fortresses.

 

[GnoSIS]

I think it is more important to reduce the complexity (Species & Jobs) than the number of pops. The main reason is, that the number of species has a bigger influence on game performance than the number of jobs & pops:
The complexity of Job assignment highly depends on the number of Jobs & Species & Pops on a planet. As far as I understand, the performance scales very bad (exponentially) with number of species, but the number of pops does in worst case scale linearly (and less if you do an effective binning of the different pops which gets much more effective if the bin size is big & the number of bins is small- hence less different species = performance improvement). Same is true for the number of different jobs.
Why is the binning important? It enables parallelization! And i think you can see every planet & empire as a bin -> better possible multithreading

For me a reduction of the number of pops would be a game breaker- i want more pops, not less.

About the enhancement of job output: There are already several late game technologies enhancing the output per job and i think adding several repeatable more would be convenient (e.g less input per job needed ...).
Additionally, some jobs, could indeed deserve some love. E.g. Soldiers, for my part, are highly non efficient (i never build fortresses). When i want to increase my fleet, i build star bases but not fortresses.

I don't think we'll see a major benefit with subspecies, because the issue with pops and lag is modifiers. A simpler pop system would make it possible to agregate processing, but would you like a simpler, more flat pop representation? This is the only way to have more pops.

Soldiers are cool. I build at least 1 fort on each colony, for naval cap, because the starbases are not enough! This is how you reach FE levels in 2340.

 

[GnoSIS]

@Thinghunter the idea to reduce jobs, has been discussed to death. What we need is to reduc pop count, and increase job effects.

For a game that wants to have pops shine and have them as the back bone of progress, stellaris does a bad job on enhancing job output. there's not even an end game repeatable for it. idea being you stop growing pops, but you increase your pop efficiency.

Quoting myself here, prophetic words indeed! Yes this can be put to rest, but after the next version and probably DLC comes out and we play a few games to see the reality of playing into 2500-2600 and concluding a game like it's 2016 Again

 

[MrFreake_PDX]

Quoting myself here, prophetic words indeed! Yes this can be put to rest, but after the next version and probably DLC comes out and we play a few games to see the reality of playing into 2500-2600 and concluding a game like it's 2016 Again

I hope this comment ages extremely well.

 

[Thinghunter]

Yeah, i have very high hopes for the upcoming patch regarding the performance- especially if it makes it more balanced regarding playing tall or wide (currently you are kinda forced to go wide because more planets are always better).

 

[Riince2]

Hello. If any gentlemen here are in possession of a Ryzen 5000 cpu and is not gpu bottenecked in stellaris in any way I would very much appreciate if they did a late game performance test by stopwatching how long it takes for x months or even a year to pass on average, preferrably over a few runs on Fastest, and have a 2-stick ram config and are willing to tell me what speeds this RAM runs at I would like to do some apples to apples comparison with Intel to determine whether it's a worthy upgrade to go from Intel 7th/8th/9th/10th gen up to Ryzen 5000, in other words whether it's one of those games AMD is leading by a few % or if it's one of the games they are leading by a large margin. For the good of the people who play stellaris that are considering whether or not to upgrade (including myself).

Please post your time results, ram speed, and the save file here if interested and I will run the same test on my 10700k to reach a conclusion about this. Make sure the save is late/heavy enough to actually be slowed down of course.

If you have a 4-stick ram config, I can't copy that for testing, but if you would be willing to test whether or not there is a difference between having single vs dual rank (running with 2 sticks instead of 4) to see if Stellaris is something that also benefits hugely from being able to simultaneously read and write to memory registers, that would also be very appreciated, but I don't expect anyone to actually go through that much effort.

You could also use my old performance-testing save from 2.8.0 but I am not 100% sure if it still works. Full vanilla. Thanks in advance.

 

[CrazyJ]

the next patch is looking very promising. admittedly i said the same thing before 2.2 dropped and boy was i wrong... but this patch is offering solutions to a few of the game's biggest issues, and unlike with megacorpse the dev's seem to be focusing more on fixes than adding new shinies.

I'm very happy with what the stellaris's future looks like. I'm glad paradox got their act together