Performance optimization

[EustaceCS]

On 12/26/2021 at 9:28 PM, gabberworld said:

? you probaply not know what i talk.

As a matter of fact, I do, better... Let's quote it again.

On 12/20/2021 at 7:02 PM, gabberworld said:

convert can be done when game starts for example

Dumping all conversions to session start are unlikely to please anyone. And resulting positive mid-session impact will be incomparable with... let's say, getting rid of that pesky cycle start lag...

Besides, mid-session conversions are kinda inevitable. Because hats, suits and whatever.

But oh, I don't know what I'm talking about... sheesh...

[gabberworld]

7 hours ago, EustaceCS said:

As a matter of fact, I do, better... Let's quote it again.

Dumping all conversions to session start are unlikely to please anyone. And resulting positive mid-session impact will be incomparable with... let's say, getting rid of that pesky cycle start lag...

Besides, mid-session conversions are kinda inevitable. Because hats, suits and whatever.

But oh, I don't know what I'm talking about... sheesh...

i not know that critters use hats and suits, also you probaply not know how is handled factorio game what have indeed much better performance

also does steam turbine have hats and suits? even if it not looks good for dupe animation, that still can be done allot other ones

[Smife]

ok let's be real. Most who buy ONI play it for like max 50 cycles then leave it to play smth else. Or they like the game, leave a positive review, but don't want to commit. So they don't get to experience lag at all. Klei can continue to sell new content or try to appease the 5% who go to space by focusing on performance, hmmmmm...

[gabberworld]

30 minutes ago, Smife said:

ok let's be real. Most who buy ONI play it for like max 50 cycles then leave it to play smth else. Or they like the game, leave a positive review, but don't want to commit. So they don't get to experience lag at all. Klei can continue to sell new content or try to appease the 5% who go to space by focusing on performance, hmmmmm...

hmmmm, it does sound evil

[tscharf]

I think we could also help showing to the developers what causes most lag from our perspective, maybe we can get further performance improvements as happened a couple times.

This discussion is also nice that shows what we need to be able to do big bases (where avoiding lags is one more challenge).

(I am in a trip and I miss playing ONI, while I cannot play, I am trying to help in other way )

 

For example does anyone have issues with large liquid/gas pools, or # of critters, or # of steam turbines?

- For bases that have large chunks of vacuum, is the performance better? (or compare vs. cases with oxygen/co2 mix in amost entire map)

- What about # of debris scattered in the base (vs. quantum storage)?

 

Did someone do any tests comparing performance vs. # of each feature of the game (10 steam turbines vs. 100 steam turbines, 10 hatches vs. 100 hatches, nb of debris scattered across the bases, etc)?

[Krishot7]

It does thousands of calculations per second and the worst of it is pathfinding all of this puts the whole load on cpu

[misotoma]

I bought a 12600k literally only for this game. I’m glad my other games will work very well but it’s crazy I have to build a PC with this game in mind lol. One point I’d like to make… before the DLC was given to us the base game was very good actually I could run a game twice as long on my old build. Something with the DLC ruined the game in my opinion (performance wise) 

it’s so bad that even deactivating the DLC won’t bring back the base game as it used to be. It now does have performance issues as well. Something happened to the game and I am truly upset. 

On 2/19/2022 at 11:14 PM, Krishot7 said:

It does thousands of calculations per second and the worst of it is pathfinding all of this puts the whole load on cpu

If someone can confirm this in a proper way would be cool. Because last map I was at 1,000 cycles and decided to turn on sandbox. I boxed in my dupes in such a way that it really only left one path for them to travel and it had the same amount of lag. I honestly think it’s the amount of information that gets added each time the game saves. The lag happens over time and doesn’t go away. Just my two cents. 

[Sasza22]

22 hours ago, misotoma said:

The lag happens over time and doesn’t go away. Just my two cents.

 

22 hours ago, misotoma said:

before the DLC was given to us the base game was very good actually I could run a game twice as long on my old build.

I honestly wonder what in the dlc made such a difference.It can`t be the usual things like pathing, thermal calc or gas movement since none of that changed. They added one overlay for radiation but that shouldn`t impact the base game. My guess is that all the diagnostic things did that. The game now remembers a lot more stuff to make graphs and tracks things it didn`t track before. Maybe that is the reason.

[babba]

On 3/3/2022 at 9:32 PM, misotoma said:

I bought a 12600k literally only for this game.

...

Your system is now "compute ready" for Klei to implement rapid spreading fires in to the game with DLC2

Once a cpu is on the market with 100mb level3 cache, then babba will be full in !

[calibayzone]

Performance is a big problem and not one solved by high end computers b/c it is single-core. Although having different planetoids is begging to be made multicore!

The game encourages accumulating large numbers of starving critters, pacu in particular as they do not need to be fed at all so are free high-tier food, lime, and even heat deletion.

I once suggested forming fish schools (or hatch tranches, pip estates, shine stars, morb mashes, slug cornucopias, etc) when enough critters end up in one space outside of an active ranch. They would path-find and heat-transfer as single entities, vastly reducing the CPU cost as these calculations need to be done far more often than reproduction and starvation. Any hungry members in the herd would move the whole heard to food, and critter traps would randomly remove one from the heard.

Performance is ultimately about perception. If we decouple the GUI from the core game logic, pressing F would be painless even if the simulation itself is slow. On a related note, adding incremental blueprints (i.e. build these ladders first, then these tiles, then destroy the ladders, then sweep up debris, and finally seal the entryway) would allow us to make "fire-and-forget" complex structures without the risk of suffocation. This means less painful waiting while they complete each step and less micro in general. Maybe making the game feel smoother is a more feasible goal than actually making the simulation faster?

[gabberworld]

50 minutes ago, calibayzone said:

Performance is a big problem and not one solved by high end computers b/c it is single-core. Although having different planetoids is begging to be made multicore!

The game encourages accumulating large numbers of starving critters, pacu in particular as they do not need to be fed at all so are free high-tier food, lime, and even heat deletion.

I once suggested forming fish schools (or hatch tranches, pip estates, shine stars, morb mashes, slug cornucopias, etc) when enough critters end up in one space outside of an active ranch. They would path-find and heat-transfer as single entities, vastly reducing the CPU cost as these calculations need to be done far more often than reproduction and starvation. Any hungry members in the herd would move the whole heard to food, and critter traps would randomly remove one from the heard.

Performance is ultimately about perception. If we decouple the GUI from the core game logic, pressing F would be painless even if the simulation itself is slow. On a related note, adding incremental blueprints (i.e. build these ladders first, then these tiles, then destroy the ladders, then sweep up debris, and finally seal the entryway) would allow us to make "fire-and-forget" complex structures without the risk of suffocation. This means less painful waiting while they complete each step and less micro in general. Maybe making the game feel smoother is a more feasible goal than actually making the simulation faster?

there is one solution for multi-core but not sure if its possible for game, make game work same way like firefox doing this by now. this is guarantee that every window does have own core.

this kind job needs rewrite the code tho so im not sure that we see this

also in theologically you can overlap multi windows by using the transparent, this way you could make also game layers at different cores