Stellaris Dev Diary #181 : Threading and Loading Times

[Coconut_Cookie]

Will there be more attention for the technical limitations of the game when designing features for future expansions? The game is still a lot slower then it was before the Megacorp dlc. New features can be nice and all but if the late game stays as slow as it currently is, I wouldn't consider it a lot of fun.

 

[correctamundo]

Douze points!

And your speed
Is all you'll ever need
All you'll ever need to know.

 

[MatRopert]

Shame I was wondering too how the Linux version would benefit from this change. Have you done any comparison between the native linux build and the new DX11 renderer running under proton with DXVK?

Would we ever be likely to see a Vulkan rendered? Seeing as this could benefit multiple platforms.

I was also wondering how DX11 will affect Stellaris on Linux. And thanks for the update

DX11 will obviously not change anything outside of windows.
Our engine team has been experimenting with Vulkan and DX12 for the new games but there are no plans for Stellaris as far as I know.

Will there be more attention for the technical limitations of the game when designing features for future expansions? The game is still a lot slower then it was before the Megacorp dlc. New features can be nice and all but if the late game stays as slow as it currently is, I wouldn't consider it a lot of fun.

We are aware about concerns of game speed. I can take a moment to clarify a bit.
Features are usually not the reason why the game performance varies. The big impact from Megacorp is due to the population system rework in the free patch. Galaxies nowadays have at least 4-5 times more population to simulate now, which comes with a cost.
As for why we looked at startup speed in the first place, first we developers restart the game a _lot_ during development (think something like 20+ times a day) so that's a huge productivity improvement for us. Second as I mentioned this could be shared with other titles, making it valuable for most of PDS and not just the Stellaris team.

 

[Mastikator]

It's good to see a Dev Diary. That performance boost looks really cool. Will we get the same benefit in game or just during launch.

Also will the next Dev Diary have information about content for 2.8, maybe a DLC? If so please say it's an espionage/intelligence update

 

[StJimmyRocks92]

Welcome back!

 

[GuildenSpur]

DX11 will obviously not change anything outside of windows.

Well, it might when using steam with proton to run the windows version where the DX11 instructions are translated to Vulkan.
I could run Bannerlord on Linux from the start.

 

[Silyus]

Thanks for the update, looking forward to the next dev diaries. I hope there will be more updates on the late game performances (and maybe the AI).

 

[Coconut_Cookie]

Features are usually not the reason why the game performance varies. The big impact from Megacorp is due to the population system rework in the free patch. Galaxies nowadays have at least 4-5 times more population to simulate now, which comes with a cost.
As for why we looked at startup speed in the first place, first we developers restart the game a _lot_ during development (think something like 20+ times a day) so that's a huge productivity improvement for us. Second as I mentioned this could be shared with other titles, making it valuable for most of PDS and not just the Stellaris team.

Thanks that is good to know but I should have been more clear with the question. With features I really meant all changes that are introduced including all changes the free patches introduce. My hope is that since Megacorp the impact on performance will have more attention when designing any new change to the game.

You don't have to answer any of this but because of Megacorp I have strong reservations about any new change that might be added. Adding something on top of a system that doesn't perform as well as it should or could might end up making the problem worse. We have a lot more pops now but what does it really add? You barely interact with them, or interact with factions, or leaders. I would completely agree with cutting down the total number of pops if that means that we can have far more interesting interactions with them.

 

[MatRopert]

It's good to see a Dev Diary. That performance boost looks really cool. Will we get the same benefit in game or just during launch.

Those changes are purely about loading times. As I mentioned the in-game speed is limited by completely different factors.

Also will the next Dev Diary have information about content for 2.8, maybe a DLC?

You will hear more about the content of the 2.8 update in the upcoming dev diaries.
Although it'll mostly be about Jeff...

 

[Inny]

I dare you to try that loading time check again on a non-SSD hard drive and a hundred mods.
This is my life playing Stellaris at the moment. You can guess I'm over-hyped at this announcement.

 

[AJGP]

Our engine team has been experimenting with Vulkan and DX12 for the new games but there are no plans for Stellaris as far as I know.

Thanks, good to here you have been looking at the newer gfx pipelines, obviously I'm biased and I would hope that you go Vulkan over dx12 if only because I would love you to continue supporting linux natively.

 

[arosenberger14]

Nice to hear that some under the hood performance is still being worked on even after the lag issues were alleviated (though not solved entirely). Hopefully there's a similar level of effort going into balance, gameplay, and bugs for the next version too.

 

[DiscoRay]

Features are usually not the reason why the game performance varies. The big impact from Megacorp is due to the population system rework in the free patch. Galaxies nowadays have at least 4-5 times more population to simulate now, which comes with a cost.

To at least partly address this issue, if there isn't anything else viable to address it, would the team consider having pops that instead produce a good chunk more, but grow slower realtive to that increased producion?

 

[Dsingis]

So, let me get this straight, see if I understood this correctly:

The programmers on Stellaris have worked all summer just to increase the start-up speed of the game, instead of the speed of the gameplay itself? Or will we see another optimization dev diary explaining in detail the optimizations made to the gameplay itself?

Excuse me for being so blunt, but starting up the game faster, doesn't make the game itself faster.

 

[hadaev]

Why can't you calculate everything on host's machine and stream data to clients?

 

[Pyzayt]

My potato battery laptop thanks you for these optimisations.

Also, I knew something was up during 2.5 when my old laptop just refused to boot up the game!

 

[Coconut_Cookie]

To at least partly address this issue, if there isn't anything else viable to address it, would the team consider having pops that instead produce a good chunk more, but grow slower realtive to that increased producion?

Exactly. I would gladly cut the number of starting pops in half and decrease their growth rates if it means better performance and using that computing power to make the game more interesting. The logic pops use to switch job, their migration logic, how they support factions or how crime works could all be improved upon and could be tied to interesting decisions for the player.

 

[MatRopert]

So, let me get this straight, see if I understood this correctly:

The programmers on Stellaris have worked all summer just to increase the start-up speed of the game, instead of the speed of the gameplay itself? Or will we see another optimization dev diary explaining in detail the optimizations made to the gameplay itself?

Excuse me for being so blunt, but starting up the game faster, doesn't make the game itself faster.

This isn't the only we worked on, no. This is merely what I toyed with during the slow summer month with a bit of outside help.

Why can't you calculate everything on host's machine and stream data to clients?

This is an idea that comes back from time to time. Thing is, it requires an efficient way of discerning what data changed during an update and then another efficient way to transmit that over the wire. So far we haven't seen a lot of empirical evidence that this would prove more efficient.

 

[Janx14]

It is nice to see the first dev diary back be about performance and not about what we should be buying.

Thank you.

 

[WolfAleron]

So all this applies just to loading game and not to actually running it? Cuz Stellaris actually runs the worst for me while Imperator Runs the best and the smoothest. Personally don't mind loading times as much as how much time it takes for one tick to happen.

Anyways, idk these things well, but am quite curious.