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 52828 Bug Reports 33970 Console edition 1287 Suggestions 20882 Tech Support 3132 Multiplayer 390 User Mods 4761 Stellaris AAR (After Action Reports)
Show only dev responses
FacelessOne said:
Exactly that is what rally baffles me about you people...
This attitude is why never anything gets really fixed.
Click to expand...

If bad behavior is rewarded you'll get more bad behavior. It's funny when you see people saying that they HATE where the game is going but don't bother "voting" in the clearest way -- with their wallets :)

<SNIP>

FacelessOne said:
Also: I find the lack of Developer in this thread disturbing!
Click to expand...

Industrious81 said:
It's a containment thread, developer involvement is pretty much just moving trouble topics into this one.
Click to expand...

I hope that this isn't like another famous thread where those hoping for a change are shuffled off to be ignored and [hopefully] the issue -- in this case performance -- dies without being fundamentally addressed.
 
Riftwalker said:
edit: from the time I press play from the launcher to the time i get to the main menu, it took me exactly 2 minutes and 8 seconds. if yours starts in 6 seconds that's amazing, it's usually longer when i use mods, because there's more to compile. this is on a HDD, but otherwise has a very strong Hardware, so i hardly think this is an unusual occurance.

for comparison, it takes me 2 seconds to load up to the main menu for another game i have installed on my HDD.
Click to expand...


Interesting. I used to have "long" load times when I ran a traditional HDD. On my 7-8 year old system my current load time [with a not-fast SSD] is < 35 seconds. I know my load times improved a TON with an SSD ... I just don't recall by how much though :(
 
KingAlamar said:
Interesting. I used to have "long" load times when I ran a traditional HDD. On my 7-8 year old system my current load time [with a not-fast SSD] is < 35 seconds. I know my load times improved a TON with an SSD ... I just don't recall by how much though :(
Click to expand...

EDIT: I just did some searching from last year when I got my SSD. Based on my posts from early 2019 it seems the SSD likely helped a LOT. I seem to recall a very slow game start and getting an SSD did help. Can't recall now if it was 2 minutes plus or not but I'm confident I was well in excess of 1 minute start up ....
 
alexti said:
It's hard to make a guess. I would try to look at where the time is spent (like user vs kernel, possibly in what module, which might point to AV, for example). Also if it's spent in disk IO or not. The choice of the tool would depend on OS, but all of them have tools to monitor that. What OS are you using?
Click to expand...

Windows 10 Pro, thanks for the help on this.

KingAlamar said:
\
I hope that this isn't like another famous thread where those hoping for a change are shuffled off to be ignored and [hopefully] the issue -- in this case performance -- dies without being fundamentally addressed.
Click to expand...

i get the impression that just 99% of performance threads aren't actually useful to the devs, they know it's an issue, but most people don't actually know what the problems are, while they can probably get statistics about what function calls are taking up the most time and thus where they need to focus their work.

it's just performance updates imo, are the hardest thing to fix, as you usually have to restructure code to get significant gains as otherwise you need to do assembly tweaking or find a way to cull the number of cpu cycles spent on certain things.
 
FacelessOne said:
Exactly that is what rally baffles me about you people...
This attitude is why never anything gets really fixed. Example is Star Trek Online (which I currently revisit after 7 years): Half of the missions are either broken or massively stacked against the player. But instead of fixing it, they gave the player a "skip" button. But they still pump out new microtransactions stuffs and the people still buy said stuff...

As already mentioned, I have 4.000 hours on CKII but I won't ever touch CKIII - even if PDX would gave it to me for free!
(Let's be honest, that will never happen as I'm only a fan and not a fanboi and that particular straw, why I will never play it anyway, broke back when they "changed" the mediterran portrait pack, cut half of it and sold it as a different DLC)

As long as people, with more money than they actually deserve, blindly buy, politics of Publisher/Developer will never change!

In such cases I always like to give my buddy Steve 1:39 minutes of your time:

Also: I find the lack of Developer in this thread disturbing!
Click to expand...

Oh I completely agree. Really wouldn't kill to have a developer acknowledge their justifiably irritated customers who can't finish a game with a brand new cpu and a very good graphics card.

However let's see what they say next week and see if they can pull something off. It's very much in their interest to get this done as a priority as I certainly and you'd have to be very naive IMO to buy federations until we hear/see that the performance has been MASSIVELY improved. And DLC's at this stage are for players who know what the game was like and would actually like to finish a game.

Anyway let's see what they have to say but more importantly WHAT THEY DO as they've said some stuff in the past about improving performance which actually made things worse. Looking forward to being proved wrong by the silent development team.
 
Riftwalker said:
Windows 10 Pro, thanks for the help on this.



i get the impression that just 99% of performance threads aren't actually useful to the devs, they know it's an issue, but most people don't actually know what the problems are, while they can probably get statistics about what function calls are taking up the most time and thus where they need to focus their work.

it's just performance updates imo, are the hardest thing to fix, as you usually have to restructure code to get significant gains as otherwise you need to do assembly tweaking or find a way to cull the number of cpu cycles spent on certain things.
Click to expand...


You are right of course regarding performance being one of the hardest things to deal with. However I am not the most technically versed person but to put it simply.

It worked fine in the past and now it doesn't because of changes made, most obviously to the pop system. If it's broke, fix it.

It's especially frustrating as the game itself with all the paid DLC's is actually great to pay for the first few hours but it seems that an ill advised and I'd guess not properly tested design choice is why we are resigned to moaning on this thread.
 
Last edited:
Riftwalker said:
i get the impression that just 99% of performance threads aren't actually useful to the devs, they know it's an issue, but most people don't actually know what the problems are, while they can probably get statistics about what function calls are taking up the most time and thus where they need to focus their work.

it's just performance updates imo, are the hardest thing to fix, as you usually have to restructure code to get significant gains as otherwise you need to do assembly tweaking or find a way to cull the number of cpu cycles spent on certain things.
Click to expand...

You're almost certainly correct that most of the threads aren't particularly useful to DEVs trying to solve the issue. I'm reasonably confident that the DEVs are bright enough to understand what are the root causes of a lot of the issues we're seeing. What the threads MIGHT do is give feedback up the food chain that the customers are concerned about the issue and hopefully there will be a higher priority to resolve these issues at the cost of something else.

I used to do some optimization back in the day. While you can make nice gains by tweaking existing systems you are likely to only see huge gains by "revolution not evolution". In this vain I think pruning the calculation tree could hold some benefits. Changing systems in place so that they work better with the game may even yield greater benefits.


Shrike said:
You are right of course regarding performance being one of the hardest things to deal with. However I am not the most technically versed person but to put it simply.

It worked fine in the past and now it doesn't because of changes made, most obviously to the pop system. If it's broke, fix it.

It's especially frustrating as the game itself with all the paid DLC's is actually great to pay for the first few hours but it seems that an ill advised and I'd guess not properly tested design choice is why we are resigned to moaning on this thread.
Click to expand...


As a software engineer that had to do optimizations, sometimes even using assembly and "cheating like mad", I can say that performance isn't necessarily that hard of a nut to crack. It just takes RESOURCES [time, skull-sweat, etc.]. The problem is that if Paradox can make MORE money on using those RESOURCES to make a new DLC then Paradox "must" make that new DLC for their shareholders. Its up to us customers to let Paradox know it's in their BUSINESS INTERESTS to fix issues THEN they can reallocate resources to creating new revenue streams.
 
right, i get the impression to get a significant performance improvement they will need to redo how pops are handles... again.

something more like victoria 2's pop system and how factories work, would be much cleaner to process probably. each pop being an individual just gives a ton of overhead that doesn't do much.

they could do a system of job, species, planet, and then have a value on that of the number of pops, with political pie charts and consumption, etc.

there would be a bunch of generic jobs people could do, but your factories and the like make very efficient jobs to fill. idk, i doubt we'll see this, it's very unlikely a company would admit defeat on something like this.
 
Wonder why some of the calculus-stuff is not given to an agent in another instance (meaning another CPU-thread) and synced back into the game - something the community did in ARMA with adding another (headless) client taking over the AI. It's sad to have 23 of 24 cores just doing sitting there bored ...
 
My understanding is that calculus stuff works in a way that does not allow efficient distribution. Like if the game is doing filling the jobs thing first and calculating income from them second, if you have two processes for those 2nd will have to wait for the first to finish anyway. And if it does unnecessary attempts to place everypop on every vacant job, even if that pop already has a job, it does not matter if it is not required for income calculations for current day, if job placement takes 5000% more time than income calculation. It will be substantial work with substantial risk of creating new bugs for marginal benefit.
 
... if you look at it from the perspective "are there many parallel tasks to calculate / simulate" which are not neccessary in real time you could outsource a lot of it to relief the core engine. Especially something like props could be easyly distributed to agents (cpu threats) and synced in every day / 10th day or on an monthly basis (as well as triggered). It is of course some pain to implement it, but it would massively improve the performance (and could also extend the limiting factors of the 32bit envelope the engine currently runs on)... imagine the size of the galaxy can scale with the cores you have (depending on the mechanics even outsourcing somethething like the AI over the network) with a smooth running mid/lategame.

Hopefully something like that will at least find a way into Stellaris 2 ;-)

Imagine how I feel with plenty resources available and the game is going only slomo by only using 8-9% of the CPU power ...

bored-cpu.png
 
not enough parallelisation it should be parallelised more there is only so much that can be done due to the engine and nature of linear functioning needed oh then I guess I should wait for Stellaris 2 before they address it that way yes instead they should focus on optimising features reducing performance by other methods and as has been proven by modders there is some performance improvements to be gained there and they could not get at hard code which the dev team can which means it is likely they can do even better if only they would commit time to it they say they have let us wait until we see if that is the case rather than repeat the same discussion points over and over and maybe even read the thread before posting that again
heyperf.png

The meme seemed relevant.
 
Last edited:
  • 1Haha
Reactions:
Attitude Adjuster said:
Yeah I can't imagine they are gonna do a stream over it if there's nothing amazing to show, so hopefully its gonna be a "wow" moment.
Click to expand...

Either that or they won't actually take the streamed game to anything approaching a proper endgame.
I want to see what the game's like by 3200 with 1k stars, thirty empires, five FEs, 5x habitable worlds and primitives--y'know, the gameplay options they sold us.
 
Blayze said:
Either that or they won't actually take the streamed game to anything approaching a proper endgame.
I want to see what the game's like by 3200 with 1k stars, thirty empires, five FEs, 5x habitable worlds and primitives--y'know, the gameplay options they sold us.
Click to expand...

Honestly I would be happy with 10 real-time-second months in a full, large galaxy in 2550. I imagine that if they do finally fix the performance hit 2.2.x gave us then I'll reward Paradox and then start pushing for them to fix the next thing(s) on my list.
 
Oh sh*t, i am at work tomorrow when they stream it. I will absolutely watch the vod.

Blayze said:
Either that or they won't actually take the streamed game to anything approaching a proper endgame.
I want to see what the game's like by 3200 with 1k stars, thirty empires, five FEs, 5x habitable worlds and primitives--y'know, the gameplay options they sold us.
Click to expand...

They should show the how the game performs at the "recommended" specs with this settings, not on an modern cpu. I guess this would be a nice laugh in the evening for me... :confused:
 
You must log in or register to reply here.
Top Bottom
Back