CPU performance bottleneck !

[LunawormAntassinia]

My 13700K simulation speed is about 40 minutes in the game = 30 seconds in reality under normal circumstances, and it will be reduced to 1 minute in the game = 1 second in reality when the population is about 200,000. To 250,000 people, the game has made all 13700K threads and core usage have reached more than 95%, and occasionally there will be crashes and flash back. It seems that this level has reached the limit of the 13700K smooth play of City Skyline 2. Should I replace it with a better CPU like 14900KS or 9950X?

[V_Cred]

Even swapping for a newer one, there will be a limit and it will be not that far from your current one. Maybe is better to check if your CPU is being thermally throttled first, and then do some changes to see if you can obtain a little more before spending money on a new one.

But keep in mind that the game will crush and push to the thermal limit any CPU that you throw at it, there's little we can do about it.

[Cosmic Sea]

No. The game is CPU bound because of poorly optimized code. A better CPU will only help marginally. Not worth upgrading, it's something they need to fix in code.

[V_Cred]

I currently have a very common CPU (ryzen r5 5600), but since I live in a hot location I do have an appropriate thermal solution and also some bios tweaks to handle thermal. No crashes, black screen, or anything else, the system runs stable, at the cost of not pushing it to the limit all the time.

It's up to ou to find the "best" spot for your system and get the most of it without sacrificing your hardware.

[V_Cred]

Originally posted by Cosmic Sea:

No. The game is CPU bound because of poorly optimized code. A better CPU will only help marginally. Not worth upgrading, it's something they need to fix in code.

I do believe that the game could do better, after all any code can be improved.
But in this case is just how the game is, I don't think the issue is code optimization, is just that this kind of simulation requires lots of processing, it is not an ordinary game.

My point is, I don't think it can be improved a lot because is badly optimized, is just how this kind of simulation game works.

[General T.Montana Tropic Thunder]

A skilled dev team probably would utilize 1% of the used cpu load with this game.

[mosshunt]

Its badly optimised, got a city recently to 664k which is now grindingly slow; down to optimising all my pedestrian crossings, car parking, public transport (number of routes seems to impact performance). however this only gets you so far. It needs a fix seems a huge amount is pathfinding related.

[Cosmic Sea]

Originally posted by V_Cred:

Originally posted by Cosmic Sea:

No. The game is CPU bound because of poorly optimized code. A better CPU will only help marginally. Not worth upgrading, it's something they need to fix in code.

I do believe that the game could do better, after all any code can be improved.
But in this case is just how the game is, I don't think the issue is code optimization, is just that this kind of simulation requires lots of processing, it is not an ordinary game.

My point is, I don't think it can be improved a lot because is badly optimized, is just how this kind of simulation game works.

There are lots of calculations, sure, but not so much that it would need 100% of a top tier CPU to handle more than 250k agents. Its poorly optimized.

[V_Cred]

Originally posted by Cosmic Sea:

Originally posted by V_Cred:

I do believe that the game could do better, after all any code can be improved.
But in this case is just how the game is, I don't think the issue is code optimization, is just that this kind of simulation requires lots of processing, it is not an ordinary game.

My point is, I don't think it can be improved a lot because is badly optimized, is just how this kind of simulation game works.

There are lots of calculations, sure, but not so much that it would need 100% of a top tier CPU to handle more than 250k agents. Its poorly optimized.

Maybe one way to "optimize" is to cut unnecessary or not-so-relevant queries, or query some aspect less often, is hard to know, but a tiny change in code can escalate very differently based on agent numbers. I'm a dev myself and I deal with similar questions every day.

For sure there is room for improvements, but sometimes it requires cutting content or adjust it, delivering less simulation or fidelity, which is a complex equation.

Note that I'm not siding with C.O., I'm just looking at the situation with my professional experience, and it looks a huge challenge.

[Ry4n]

7700x and no real complaints with performance in CS2, simulation speed could improve with larger pops but CS1 is more limited for me.

[LunawormAntassinia]

Originally posted by V_Cred:

I currently have a very common CPU (ryzen r5 5600), but since I live in a hot location I do have an appropriate thermal solution and also some bios tweaks to handle thermal. No crashes, black screen, or anything else, the system runs stable, at the cost of not pushing it to the limit all the time.

It's up to ou to find the "best" spot for your system and get the most of it without sacrificing your hardware.

In fact, my CPU cooler is pretty well, fully loaded does not exceed 85 °C. My crashes and black screen problems generally occur when automatic saving. I suspect it has something to do with the 13 and 14th-gen Intel cpus or the temperature of the SSD, but there is no evidence.

[sternenstaub70]

Originally posted by LunawormAntassinia:

In fact, my CPU cooler is pretty well, fully loaded does not exceed 85 °C. My crashes and black screen problems generally occur when automatic saving. I suspect it has something to do with the 13 and 14th-gen Intel cpus or the temperature of the SSD, but there is no evidence.

To find out any evidence the best thing you could do is to try all known tricks to rule those suspected components out.
Intel CPU: Do the undervoltage tricks
SSD: Open case of your computer , put a fan next to the computer and blow wind into the computer.
If the problems disappear you may have found your culprit. Otherwise, blame the game..

[AlSemz™]

Its the engine. its not going to matter. They just need to move to unreal 5 for better performance already. The game would also have better visual fidelity, but overall unreal 5 is capable of more power, its just harder to learn and design with. Maybe they need new coders too. Honestly they should focus on delivering a few month of free DLC to say sorry, then move on to Cities 3 on Unreal 5, hoping for a release in 2026-27.

[Cosmic Sea]

Originally posted by V_Cred:

Originally posted by Cosmic Sea:

There are lots of calculations, sure, but not so much that it would need 100% of a top tier CPU to handle more than 250k agents. Its poorly optimized.

Maybe one way to "optimize" is to cut unnecessary or not-so-relevant queries, or query some aspect less often, is hard to know, but a tiny change in code can escalate very differently based on agent numbers. I'm a dev myself and I deal with similar questions every day.

For sure there is room for improvements, but sometimes it requires cutting content or adjust it, delivering less simulation or fidelity, which is a complex equation.

Note that I'm not siding with C.O., I'm just looking at the situation with my professional experience, and it looks a huge challenge.

Impossible to know how "huge" of a challenge it is without access to their code base. It could be relatively simple if they have tools to identify what is sucking up all the processing power. But itis probably not that huge of a challenge and can be optimized without losing the quality of the simulation.

[mosshunt]

Originally posted by AlSemz™:

Its the engine. its not going to matter. They just need to move to unreal 5 for better performance already. The game would also have better visual fidelity, but overall unreal 5 is capable of more power, its just harder to learn and design with. Maybe they need new coders too. Honestly they should focus on delivering a few month of free DLC to say sorry, then move on to Cities 3 on Unreal 5, hoping for a release in 2026-27.

Better still they can stick by CS2, improve it to UE5 (this isn't impossible I play a few games where this is taking place FOC - that is the apology itself), then they can sell full cost DLCs like CS1. This should also recover CO and PDX reputations somewhat and people will be more tolerant of a bad release again.

[Trouter]

Originally posted by AlSemz™:

Its the engine. its not going to matter. They just need to move to unreal 5 for better performance already. The game would also have better visual fidelity, but overall unreal 5 is capable of more power, its just harder to learn and design with. Maybe they need new coders too. Honestly they should focus on delivering a few month of free DLC to say sorry, then move on to Cities 3 on Unreal 5, hoping for a release in 2026-27.

Unreal provides absolutely nothing that makes it better for simulation games. Unity has DOTS while Unreal has nothing comparable at the moment. Unreal is notorious for having issues with stuttering so I'm not sure anyone can state that Unreal is magically better for performance.