I have an Intel Core I5-4460 @ 3.2 GHz, Stellaris is a great game but I am in mid game and I have noticed major decrease of my Frames, I have then seen that I am only using 40-50% of my GPU GTX 960 and 40-60% of RAM. I have software that read all needed information and I could see that the game only use 1 of the 4 cores that I have. I Really want to know if there will be an update for this problem and if there is any one that can help.
Use of one core of CPU Stellaris
- Thread starter TheBlueNight
- Start date
-
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.
- Status
After 300 years the game is unplayable. To 60 fps drop to 5.
Fx8350 and gtx970
Fx8350 and gtx970
- 2
- 1
As a developper not affiliated with Paradox, I doubt that more cores will be used any time soon.
Using multiple cores is a difficult to implement technical issue, it requires a lot of planning and if to be implemented, must be considered from the very start. It's not something you can just "tack on" down the line. Besides, it's not possible to just evenly use any of the cores; say 25% of program needs for each core. Data still has to be processed in a specific order, this limits multithreading heavily. Some tasks you may be able to process in parallel, but mostly, it needs to be done sequentially. Don't expect the game to run 4x as fast just because it uses 4x as many cores, if it ever comes to that.
Using multiple cores is a difficult to implement technical issue, it requires a lot of planning and if to be implemented, must be considered from the very start. It's not something you can just "tack on" down the line. Besides, it's not possible to just evenly use any of the cores; say 25% of program needs for each core. Data still has to be processed in a specific order, this limits multithreading heavily. Some tasks you may be able to process in parallel, but mostly, it needs to be done sequentially. Don't expect the game to run 4x as fast just because it uses 4x as many cores, if it ever comes to that.
- 18
- 3
It would already help, when they put everythink ship movement or combat related to the second core.
Core 1: POPS, diplomacy, map, etc
Core 2: Ship (movement, combat, graphic commands to GPU)
At the moment you can simply see how ships action overload the CPU, best way to counter this hard laggs, reduce gamespeed for every combat from very fast (maximum CPU useage) to a gamespeed lower than very fast until your CPU - Game handeling can catch up.
Core 1: POPS, diplomacy, map, etc
Core 2: Ship (movement, combat, graphic commands to GPU)
At the moment you can simply see how ships action overload the CPU, best way to counter this hard laggs, reduce gamespeed for every combat from very fast (maximum CPU useage) to a gamespeed lower than very fast until your CPU - Game handeling can catch up.
im fairly certain its not that straight forward to use multiple cores and i highly doubt its gonna happen after release
- 3
like what someone above said, it's not that easy.
programs tend to be(for very good reasons) be programmed sequentially. X>Y>Z ect. you can't do X and Y in parralel if the internal logic of the game relies on X being done at that point. and the vast overwhelming majority of applications work like that. because doing stuff in paralel is REALLY REALLY hard. if you don't do it properly you'll just slow everything down more likely than not. and if it does work, the amount of cycles youw in in doing so will liekly be fairly minimal.
rather than multithreading bigger parts of the game(as they already offloaded stuff that just makes sense to offload to other cores - the game acutally uses 100% of the CPU during loading if disk and ram allows IIRC for example), it's far better to code more effieciently and reduce the amount of cycles you waste, i.e. in CKII when greek characters were always checking all other characters to see if they should castrate or blind them.
programs tend to be(for very good reasons) be programmed sequentially. X>Y>Z ect. you can't do X and Y in parralel if the internal logic of the game relies on X being done at that point. and the vast overwhelming majority of applications work like that. because doing stuff in paralel is REALLY REALLY hard. if you don't do it properly you'll just slow everything down more likely than not. and if it does work, the amount of cycles youw in in doing so will liekly be fairly minimal.
rather than multithreading bigger parts of the game(as they already offloaded stuff that just makes sense to offload to other cores - the game acutally uses 100% of the CPU during loading if disk and ram allows IIRC for example), it's far better to code more effieciently and reduce the amount of cycles you waste, i.e. in CKII when greek characters were always checking all other characters to see if they should castrate or blind them.
Yeah why not. So you can have out of sync errors when people use different machines.
As lag problem shows it runs on 3+ thread. One for AI, one for everything graphic related and one for the game.
Ship combat use the same process handeling like rest of game....if your games starts heavy lagging, pause, and reduce speed setup to lowest possible your CPU can handle, without get scedule problems by the process handeling.
Do not forget, pardox use as speed setup a max limit counter in cpu process handeling and scale gamespeed to it.
Somethink like this:
Very slow - 100 ticks limit of cpu in 1s
slow - 200 limit ticks of cpu in 1s
normal - 400 limit ticks of cpu in 1s
fast - 800 limit ticks of cpu in 1s
very fast - unlimited ticks of cpu in 1s
So the moment you add fleetcombat, which adds 1000 ticks per second, even when your map runs smoth in very fast, in this moment you overload your CPU and your get the laggs, reducing speed is best way to counter the overlow in CPU handeling.
Do not forget, pardox use as speed setup a max limit counter in cpu process handeling and scale gamespeed to it.
Somethink like this:
Very slow - 100 ticks limit of cpu in 1s
slow - 200 limit ticks of cpu in 1s
normal - 400 limit ticks of cpu in 1s
fast - 800 limit ticks of cpu in 1s
very fast - unlimited ticks of cpu in 1s
So the moment you add fleetcombat, which adds 1000 ticks per second, even when your map runs smoth in very fast, in this moment you overload your CPU and your get the laggs, reducing speed is best way to counter the overlow in CPU handeling.
Whould it help if a person makes the star system smaller and use less empires like (600 stars and 10 empires and 1 fallen)
using a small galaxy currently means better performance until optimization has taken place to allow better performance for the larger ones
- 1
Sweet thanks I will keep my world 1000 stars but will start a new smaller one for now
Of course, that helps. Less stars = less colonies = less maximum fleet sizes and less AI sector handling, less AI empires = less checks and procedures in general.
Its basically the same as in round-based strategy games where you see which nation is currently processed, the difference is that real-time games have to appear like hey handle them all at once, which they don't, but of course the processing time is usually rather in the milliseconds so you don't notice. But if there is too much to handle, then it lags.
Stellaris won't be multi-threaded until the underlying engine is rewritten. And that's not likely to happen.
It's not impossible, just really hard, and Paradox may or may not want to spend the time/effort until other major bugs are fixed and all the low-hanging performance fruit have been harvested.
It's not impossible, just really hard, and Paradox may or may not want to spend the time/effort until other major bugs are fixed and all the low-hanging performance fruit have been harvested.
- 2
Indeed, iirc this has been a Clausewitz thing since very early on. Until Paradox develops a new engine (when that will be, as you point out, no one can even guess) I'm afraid we're stuck with the limitations (and advantages!) of Clausewitz.
stallaris like all paradox game demands far more from the CPU to calculate
Fleet - tick for movement for every fleet on map, AI, player, monster per ingame day (you can see it when you use a really old CPU, hperlane will drop your fame by star to star movement on day base for all moments of the road....so every ingame day, as long the fleet moves, it send a tick to the CPU, wormhole use only 1 tick for system to system moevemet. I tested this on a athlon 2 core 3,2GHz....with warp or hyperlane, the game laggs from day 1....wormhole runs smoth until you start a fleet fight....only way to reduce handeling load is reducing speed)
Fleet - tick for combat, for every single ship in fleet, together with follow up tick if dogde or not, and the tick which animation to use per ingame day
POP - tick for ethics for every POP on map per ingame day (if they loose or gain traits)
POP - tick for POP growth
Leader - tick for dying per ingame day
Leader - tick for XP gain per ingame day
Leader - tick for level up chance on trait
etc...
so smaller map, less action happens at the same time, far better CPU handeling, far better speed.
many players with high end systems run simply in lategame into the CPU handeling limits of there CPU...
Fleet - tick for movement for every fleet on map, AI, player, monster per ingame day (you can see it when you use a really old CPU, hperlane will drop your fame by star to star movement on day base for all moments of the road....so every ingame day, as long the fleet moves, it send a tick to the CPU, wormhole use only 1 tick for system to system moevemet. I tested this on a athlon 2 core 3,2GHz....with warp or hyperlane, the game laggs from day 1....wormhole runs smoth until you start a fleet fight....only way to reduce handeling load is reducing speed)
Fleet - tick for combat, for every single ship in fleet, together with follow up tick if dogde or not, and the tick which animation to use per ingame day
POP - tick for ethics for every POP on map per ingame day (if they loose or gain traits)
POP - tick for POP growth
Leader - tick for dying per ingame day
Leader - tick for XP gain per ingame day
Leader - tick for level up chance on trait
etc...
so smaller map, less action happens at the same time, far better CPU handeling, far better speed.
many players with high end systems run simply in lategame into the CPU handeling limits of there CPU...
We need more clock speed, that's the bottom line, and we haven't really gotten more clock speed for 10 years now.
Yeah, the power of the individual single CPU core has been a bottleneck in technology for a while which is not going to change anytime soon unfortunately. I wouldn't be surprised if eventually we end up with engines offloading CPU tasks to the GPU instead of the other way around like the olden
I dunno, from a few twitter snippets from the past it sounds like they may already be rewriting their engine for the new API's but thats just speculation.
- 1
whats the point? the problem is missing parallel processing. offloading anything is creating parallelization. im always wondering why developers dont play their own games.. when performance starts to tank realy realy bad in the lategame... shouldnt devs pick up on that shit?
I think we should let them iron out some of the bugs first. The performance may be fine on one core. I'm more and more convinced every time I play that a large part of the lag is caused by the AI constantly shuffling around pops and moving things and generally being indecisive. I'm only going to start evaluating the performance once these problems have been ironed out because before that I really doubt it matters how many cores it's using.
- Status