Jump to content

How "stable" is the game for you ?  

66 members have voted

  1. 1. How "stable" is the game for you ?

    • Getting worse for me
      11
    • Getting better for me
      2
    • 1/10 Extremely frustrating
      2
    • 2/10 Very bad
      1
    • 3/10 Bad
      3
    • 4/10 Sub par
      4
    • 5/10 Average, wish it would be better
      16
    • 6/10 Good
      9
    • 7/10 Very good
      3
    • 8/10 Happy
      2
    • 9/10 Very happy
      10
    • 10/10 Perfect
      4
    • Zero - I can`t launch or play the game
      0
    • I am still waiting for a certain crash fix
      0
    • I have a low game frame rate ( FPS ), below ~ 30 FPS average
      28
    • A re-occuring crash happens in my current save file, I can not progress with my save game
      4


Recommended Posts

Stability is decent, i get very seldomly black holes when i reload a long running game (either directly or shortly after in the game), so 9/10 here.


The performance is a complete desaster though, i've seen a lot of poorly optimized games, but ONI takes the cake here, 1/10.

I have a beefy computer (Ryzen 3950x, 64gb, Radeon 5700xt) that does extremely well in a lot of workloads (including gaming) -
But start ONI, get to cycle 1000+ and watch it judder around slightly below the 30fps mark in 3x game speed, no matter what you try.

Biggest joke is that my previous machine (10 years old, first generation Core i7 with 4 cores and 2.9Ghz) wasn't running ONI that much worse,
while everything else felt like a Tesla Model S making a traffic light race against a VW T1 Bully.

Which is a shame, i really love the game und would play it so much more if the performance in the endgame (where it's the most fun IMHO) wouldn't frustrate me so much.

Link to comment
Share on other sites

Taking into account size of the world and amount of calculations I think the game is doing great. I am around cycle 2400 with 5 duplicants and revealed world. Playing on speed 3x the game is smooth. But I guess number of duplicants is playing a role here.

The only thing that bothers me slightly is that usually after reload I am getting 0.1sec freezes every 3-5 sec that are not related to game speed. I remember discussion about some issues with clearing memory of previously loaded game.

Link to comment
Share on other sites

The one bad problem I have is that the game "eats clicks". This means I sometimes have to click on something 2 or even 3 times before the game registers my mouse click. This may be related to me clicking "blind", i.e. I usually do not wait for visual feedback, probably because I am also playing an MMO where I simply do not have time for that. But this used to work most of the time and it is not a problem with me being imprecise. The problem also showed up again late 2021. This might be micro-freezes or a problem with sometimes very fast click-release cycles not getting detected.

I do not really understand where this comes from. User-interaction like click detection should be real-time or at least run in its own thread. Mouse-clicks and keyboard-events should _never_ get lost.

I also see a crash every 250 cycles or so  when running unattended, sometimes hard, i.e. the game just die not run anymore when I check in the morning. 

And I often see small things getting corrupted, needing a game restart. For example I just found a blocked rail with absolutely no reason it was blocked. There were just a few "buckets" with nothing in them and so a conveyor termo-sensor had nothing to measure. There were buckets with stuff in them before and after. Cycling the sensor did clean that up, but this (empty buckets blocking a rail) should not be happening. In the same state, I also had debris for sweeping that did not have a visible priority. Could be selected manually and changed and that worked, but could not be selected via the prio-tool. I do observe other small things which seem to be partial object corruption for things. These may or may not be missing state-updates. This problem started when Klei got into performance optimization late last year, same as the "dropped click" problem.

Edit: Conveyor rails was apparently (or maybe not) my fault, but the observation of small inconsistencies popping up in many places after a while and rectifying themselves on save & reload remains the same.

Link to comment
Share on other sites

13 hours ago, LunaticLea said:

I have a beefy computer (Ryzen 3950x, 64gb, Radeon 5700xt)

No, when it comes to the work load of ONI, your "beefy" computer ist just nearly worthless. 

Only the first 2 cores of your 16 cores are useful at all. Your GPU is completely useless. You are just asuming your computer is good suited for this game, but you asume wrong. 

Link to comment
Share on other sites

@SharraShimada: *sigh* Of course people come running out of the woodworks to point out that my machine is "worthless" because it doesn't cater specifically to ONI :roll:

I'm very well aware that ONI only uses 2 cores - It's a joke on its own in this year and age, and the biggest reason ONI gets a 1/10 for performance from me. Using all available cores would be VERY low hanging fruit with a workload like in ONI, plenty of stuff that could easily be parallelized.

Link to comment
Share on other sites

1 hour ago, LunaticLea said:

@SharraShimada: *sigh* Of course people come running out of the woodworks to point out that my machine is "worthless" because it doesn't cater specifically to ONI :roll:

I'm very well aware that ONI only uses 2 cores - It's a joke on its own in this year and age, and the biggest reason ONI gets a 1/10 for performance from me. Using all available cores would be VERY low hanging fruit with a workload like in ONI, plenty of stuff that could easily be parallelized.

You system is entirely fine. I use a 3700X for ONI and performance is ok. Well, mostly. The last optimizations did break something. Calling your machine "nearly worthless" is just nonsense. Your computer is also pretty well suited for ONI, considering the available alternatives.

As to using more cores, that very likely is an engine limitation. And that makes it actually very hard to user more cores.

Also, at least on AMD, ONI is perfectly playable at 20FPS. On Intel you may need more because Intel has a lot more execution-jitter. 

 

Link to comment
Share on other sites

1 hour ago, LunaticLea said:

@SharraShimada: *sigh* Of course people come running out of the woodworks to point out that my machine is "worthless" because it doesn't cater specifically to ONI :roll:

I'm very well aware that ONI only uses 2 cores - It's a joke on its own in this year and age, and the biggest reason ONI gets a 1/10 for performance from me. Using all available cores would be VERY low hanging fruit with a workload like in ONI, plenty of stuff that could easily be parallelized.

Okay, you make a game, with Unity working like ONI and proof it can be heavily parallelized and THEN i may believe you are knowing what you are talking about. Until then... not. Sorry. In my oppinion you do not know anything about this matter.

Link to comment
Share on other sites

1 hour ago, SharraShimada said:

Okay, you make a game, with Unity working like ONI and proof it can be heavily parallelized and THEN i may believe you are knowing what you are talking about. Until then... not. Sorry. In my oppinion you do not know anything about this matter.

Of course *laughs*. I would even do that (just not with Unity) if i could go on an extended sabbatical in my current job and organize everything else to make a proper game out of it (would be quite the waste of time just programming the engine otherwise).

But you know what? I'm not gonna answer your posts anymore until you have at least some sort of argument, ok?

Link to comment
Share on other sites

ONI uses more than two cores. Saying it only uses two cores is either long-outdated information, or blind rumour-mongering.

I actually checked this information, and found it to be completely false.

Much of the work happens in a single main thread and this is almost always the effective bottleneck, however it does a clear majority (well over 50%) of its work in threads. Even in the extreme example of FJ's posted benchmark save, the majority of the work was still done in threads.

From my testing, the ideal number of cores to have would be 3 or 4, and definitely more than 2.

Link to comment
Share on other sites

I don't think that these threads are from ONI itself, they are more likely created by the unity engine.

Kerbal Space Program and 7 Days to Die (other unity based games that perform very poorly due to not using threading properly)
also create a lot of helper threads (with a naming scheme nearly identical to ONI and your screenshot) that basically
are completely idle most of the time.

Link to comment
Share on other sites

4 hours ago, LunaticLea said:

I don't think that these threads are from ONI itself, they are more likely created by the unity engine.

Kerbal Space Program and 7 Days to Die (other unity based games that perform very poorly due to not using threading properly)
also create a lot of helper threads (with a naming scheme nearly identical to ONI and your screenshot) that basically
are completely idle most of the time.

If one builds a lot in Kerbal Space or maxes out the game physics/simulation ( by playing the game ), then one or more of the Kerbal Space created gazillion threads hangs up, as the simulation is started or has to calculate out of range numbers, and either permanently freezes the game or instant crashes out to the OS.

Its a shame the Kerbal game was never really further developed, had no multicore support and was development dumped - Once they game got popular, sold a lot and the development team broke up. Mods for the single cpu core utilizing game are insane mess and make the unstable game even more unstable, if one does "everything" in the game.

IMHO Kerbal is the exact opposite of Factorio, in terms of development and game stability - Sadly. From my point and game stability opinion:

Kerbal: Zero - Broken

Factorio, and as another example Satisfactory ( 3D title ): 10/10 - Working

image.thumb.png.4f9ad2535bf1bbe0c5b37af210a81f25.png

Playing with 4 physical cpu cores, delidded i7 7700k with deactivated Hyperthreading, 1kg air cooler, no case. The game is passive cooled without running fans, the cpu would love to do more dear Klei !

Wishing everybody a good day, the coffee is good :beguiled:

image.thumb.png.fd83c96ebf13772f73128f3278ab3528.pngimage.png.da2aaf6de173e0094d70f0dbc1aba2b8.png 71 babies :cheerful:

Link to comment
Share on other sites

3 hours ago, LunaticLea said:

I don't think that these threads are from ONI itself, they are more likely created by the unity engine.

Kerbal Space Program and 7 Days to Die (other unity based games that perform very poorly due to not using threading properly)
also create a lot of helper threads (with a naming scheme nearly identical to ONI and your screenshot) that basically
are completely idle most of the time.

Yes there are another 80 or so useless threads created by Unity doing nothing. These threads however are (a) named according to the Klei naming sceme, and/or (b) actually doing work. Rather than quibbling, why not actually test it yourself? Or even just read the part where i stated that the amount of work being done by the other threads is two thirds of the total :roll:

Link to comment
Share on other sites

1 hour ago, babba said:

71 babies

I think this might be related to why your game has so much work done in the main thread :lol:

The usual bottleneck blame at the moment seems to be dupe pathfinding, which indeed seems to be all done in the main thread and a big amount of work. Probably the more dupes one has, and the more area dug out, the bigger the proportion of work done in the main thread.

For my 20-dupe bases with relatively minimal digging-out of areas, i have much less of the total CPU load assigned to the main thread. In fact for medium-sized mid-game bases it is comparible to factorio, which i found also does around 30% of its work in the main thread for a medium-sized mid-game base.

Link to comment
Share on other sites

Hehe :lol:

Your findings are interesting to read my dear friend. :cheerful:

My ONI maximum cpu load is the same ( maxed out at 50%, I have 4 physical cores ), with my current map save, when I play with 1 dupe. I had some big dying waves, where most of the dupes died in my played survival mode, as I play the game "Banished" style. In the past 4 ONI years the games cpu allocation had always been maximum 50% cpu load here on my system - Same with sandbox play in the first year of ONI, since then I don`t play the editor mode anymore.

I know the urge since 4 years that we would like to have more done by players available cpu cores, on the other hand its nice to play a somewhat stable game. I recall the new Klei guy working on multicore ladder optimizations and then the game was borked/crashy and unplayable for short time and they set stuff back to how it was before. Klei IMHO is regularly trying their best here and there, their optimizations ( fails or successes ) come in small portions.

Speed or memory optimizations can also have downsides, like "How well the game catches mouse clicks" ...as mentioned by @Gurgel in this thread, I can myself confirm the issue. There is so many things which can influence "catching" mouse clicks with all the software and hardware layers and the game itself, I am glad I am just a player and consumer of the game.

From my point of view the things which players can do and have in their own hands ( next to complaining :anonymous: ), is to update their cpu/board https://www.cpubenchmark.net/singleThread.html - If a player want to have more FPS in the game.

With the great new small $budget i3 12100f entry, I`m am hoping there will be great new high end cpu`s either this or next year at Christmas :razz: For babba it needs to at least have 5000 passmark single threading score, in a perfect ( unrealistic ) world it would be a dual core cpu with 128 mb level 3 cache. If the industry at least keeps 4 core cpu`s alive, I can live with that compromise.

Time for a new ONI/Factorio/Satisfactory/EvE/WoWs gaming system then for babba :ghost: Jingle Bells, Jingle Bells...

Link to comment
Share on other sites

@Yobbo There is a point in my current save game ( and 4 years of past save files of mine ), where the cpu load exceeds 50% ( I play with 4 physical cores ). In my current save game this starts at ~80 to 100 dupes. My theory in the last 4 years has been on this, that the cpu`s  l1 l2 l3 caches are then full and overloaded with ONI stuff.

While some cpu ( cores ) are on holiday and could actually be fed with lots of more ONI calculation stuff, the cpu % work load display in the task manager starts to increase beyond what the game or game engine can actually support and allocate on cpu resources.

In other words, the cpu starts to be overly busy with itself, sorting, swapping, dumping, rearranging and processing cpu caches data stuff.

Big cpu caches help improve the ONI game experience a lot, especially with a modern cpu. The smaller a players cpu caches are, the faster they start to overflow. These are my practical observations and assumptions ( I swapped a few cpu`s ), just as a regular player of 4 years ONi :razz:

Single threading cpu performance and benchmark results are affected a lot by available cpu caches, bus architecture and size. However, I`m just a gamer ...playing the game. :congratulatory:

The dupes errand system seems pretty complex and takes a good chunk of game performance, but results in great game play IMHO.

BTW I haven`t checked how big my current save file is, but loading it fresh it instantly allocates 11GB of RAM :lol: I currently have 1 colony. @.@

image.png.f2372a9337330c387ce63f7deb177a83.pngimage.png.2f2cf7876648db0fa8b2137e9b3ed600.pngimage.thumb.png.b8d83049da7772ecb1364d3e4bf1b6fa.png

image.thumb.png.7c4ed8ca5ebbfe46bb45057f3f04cc33.png Cycle 4577 - The Cold War rocket stacking installations have begun...

BTW The rockets are actually fully loaded and fueled, Ive got this rocket fuel/tank&oxidizer loading status display bug since rockets have been added to the game by Klei.

@sakura_sk :razz: wrote in the past that there is some practical form of rocket amount limit...something like 12, 14 or 16 rockets? I cant remember. If I understood it right its down to rocket interior-mapping , reserved interior map space and that building more rocket interiors could result in "map garbage" ? :confused: Please correct me if I am wrong.

That is one of the reasons why I would find it great, if Klei could make it possible that rockets can optionally be flown without interior and without getting inside. IMHO it would be great if DLC2 offers an autopilot module for rockets and if Klei removes a practical amount limit on how many "autopilot rockets" a player can operate. Also, please make it possible that we can build big fat Space Shuttles, so we can deploy satellites and deep space telescopes :bee:

Also please add Blueprints to the Survival Mode dear Klei :ghost::flustered::concern: I better stop now ``

P.P.P.S I want to add Rocket Side Boosters to rockets and Space Shuttles + new fuels ( Hydrazine etc. ), which can only be mixed in Space Stations...because reasons ! :afro:

Also need Alien Traders :lol:

P.P.P.P.P.P.P.P.S. Welding sparks, ignition, explosions, fires ! :ghost:

image.png.8a9c92c31a7139e6e3075249588c54d1.pngimage.png.76a4bc602f4141805b34ee256f91e805.pngimage.png.c3031507320d6dcbf1a233954a20de20.png

Link to comment
Share on other sites

19 hours ago, Yobbo said:

ONI uses more than two cores. Saying it only uses two cores is either long-outdated information, or blind rumour-mongering.

I actually checked this information, and found it to be completely false.

Much of the work happens in a single main thread and this is almost always the effective bottleneck, however it does a clear majority (well over 50%) of its work in threads. Even in the extreme example of FJ's posted benchmark save, the majority of the work was still done in threads.

From my testing, the ideal number of cores to have would be 3 or 4, and definitely more than 2.

Your posts prove absolutely nothing. You are just making baseless claims. Like KWorker means "Klei Worker" and runs game logic. But not something like "Kernel Worker" that is utility thread for working with Linux kernel, spawned once per core.

I didn't follow ONI development very closely. But from updates and discussions I know only about 3 game threads. 1 - Main thread in Unity. 2,3 - Native threads in SimDLL. One for temperature calculations, other for moving elements in pipes, if I'm not mistaken. There was attempt to move rails in separate thread, like pipes. But it caused bugs and I don't know how did it end.

For me 2 and 3 together can't utilize even half of the single core. If I filled the whole map with pipes, maybe my ONI would use more than 2 cores, because nothing stops 3 threads from using 3 cores. But that would not do anything with main thread being bottleneck already.
I have not seen ONI threads, that are real game threads, to use more that 2 cores, so all other cores are usefull only for Unity, OS and background programs that like to spawn threads to eat all available CPU and RAM.

In your linked post you showed: 59.8 + 20.6 + 12.6 + 8*5.3 + 5*5.0 + 3.0 + 2.3 + 9*2.0 + 3*1.7 + 0.3 = 189.1% = 1.891 core usage. Then you showed: 91 + 30.3 + 23 + 8*4.3 + 4*4.0 + 3.0 + 2.3 + 9*1.7 + 0.3 = 215.6%. And claimed that you have seen 2.3 core usage. But how about proving that all these threads run game logic? You say they "actually doing work", "doing meaningful computation", so you should know what they do, except thermodynamics work of heating CPU. High CPU usage is thermodynamics work, how about results? Their names tell me nothing. I wouldn't be surprised if their duty is to upload your files, browser history and pressed keys in internet to improve your user experience.

PS: Additional reasoning about worker threads. What I see is that their number is presumably the same as the number of cores and they utilize almost the same amount of CPU. My conclusion is whoever created them used data parallelism. Klei has used task parallelism so far: thread for temperature, thread for pipes, etc. In data parallelism exact same code is performed on different portions of data. This is some people, including me, were hoping to get in DLC: each asteroid gets personal thread with exact same code, so it can run on its own core. Only interaction points(rockets, teleports...) and global game time would need to be synchronized between threads.

Link to comment
Share on other sites

3 hours ago, SVV said:

...This is some people, including me, were hoping to get in DLC: each asteroid gets personal thread with exact same code, so it can run on its own core. Only interaction points(rockets, teleports...) and global game time would need to be synchronized between threads.

If Klei would rework the map system again...I will instantly disconnect from the internet and use this as old ONI computer, to have a stable game build for long term playing :lol:

In a perfect babba world, Klei will someday offer spreading Plutonium fires in the game - I`m praying :flustered: Down with the FPS, down !

image.png.c84d0d3b0ba9a673da6d19a4debab650.png RTX 3090 Ti Buying Experience

Link to comment
Share on other sites

5 hours ago, SVV said:

Your posts prove absolutely nothing. You are just making baseless claims. Like KWorker means "Klei Worker" and runs game logic. But not something like "Kernel Worker" that is utility thread for working with Linux kernel, spawned once per core.

These are all threads assigned by the Oxygen Not Included process on my system. Those KWorker threads are spawned only when a game is loaded. Obviously i do not know what the threads are doing. They are, however, quite clearly not doing nothing.

My post provides some actual experimental results, which is more than i have seen anybody else, including you, do.

In stead your post is simply a wall of baseless claims.

5 hours ago, SVV said:

I wouldn't be surprised if their duty is to upload your files, browser history and pressed keys in internet to improve your user experience.

If Oxygen Not Included is doing that (which i highly doubt), then it's certainly better that it does it in another thread than in the main one.

Link to comment
Share on other sites

6 hours ago, babba said:

If Klei would rework the map system again...I will instantly disconnect from the internet and use this as old ONI computer, to have a stable game build for long term playing :lol:

In a perfect babba world, Klei will someday offer spreading Plutonium fires in the game - I`m praying :flustered: Down with the FPS, down !

That would possible be another "The Big Merge" disaster so they might provide another branch for you.

Link to comment
Share on other sites

21 hours ago, babba said:

@Yobbo There is a point in my current save game ( and 4 years of past save files of mine ), where the cpu load exceeds 50% ( I play with 4 physical cores ). In my current save game this starts at ~80 to 100 dupes. My theory in the last 4 years has been on this, that the cpu`s  l1 l2 l3 caches are then full and overloaded with ONI stuff.

Maybe. If anybody gets one of the upcoming "big L3 cache" CPUs (Ryzen 7 5800X3D) from AMD that could provide a good data-point for this. Caches thrashing is certainly something that really puts a real limit on modern CPUs. I am tempted to get one, but we will see.

Sadly, AMD has not published any ONI benchmarks....

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

Please be aware that the content of this thread may be outdated and no longer applicable.

×
  • Create New...