Regular significant stuttering/skipping with every game tick.

[Kingman]

I feel like something is looping several times to make this happen, I used to play up to the 3000s with major mods (that would increase galaxy entities many times) without much lag (I mean I did have small lag but its expected and it was at the 3000s!).

Now playing vanilla, at 50 years in already see stuttering....

 

[Aliothkn]

After playing new beta patch i can confirm that there is minor(yet noticeable) performance increase at the start, but at the same time mid-late game become unplayable.

 

[KingAlamar]

Dumb question: At the start of the game, on fastest speed, how long does it take in real time to run through a whole month [or year]??

If I recall correctly with base settings it takes 250 years [min] before an end-game crisis event occurs. With the speeds I'm seeing I could literally leave my computer running for a whole day and I'd still have time before the crisis shows up. Whether or not that's normal or intended that feels wrong.

Then again what I'm seeing doesn't seem to have the same footprint of what everyone else seems to be seeing so I wanted to know what's normal and what's "abbie-normal".

 

[ShaTiK]

Dumb question: At the start of the game, on fastest speed, how long does it take in real time to run through a whole month [or year]??

If I recall correctly with base settings it takes 250 years [min] before an end-game crisis event occurs. With the speeds I'm seeing I could literally leave my computer running for a whole day and I'd still have time before the crisis shows up. Whether or not that's normal or intended that feels wrong.

Then again what I'm seeing doesn't seem to have the same footprint of what everyone else seems to be seeing so I wanted to know what's normal and what's "abbie-normal".

That's an interesting question. I made a quick test myself, and results are, for the default Large galaxy with only UI mods with ironman enabled:
(All times are in the ballpark, not exact) 22 seconds for normal speed, 12 for fast and 6 for fastest.
Here is where things gets funny. In my proper game, 2320, same default Large galaxy, AI mod on, I had the following results: 25 seconds for normal, 19 for fast and 17 for fastest. But it wasn't faster - the game just was stuttering halfway into the month on fast and was just overall choppy on fastest. What's funny is the fact that game still runs faster in, well, faster speeds. I thought that baseline was 'normal' and no matter what - it would be the same. Seems I was wrong and lag affects even normal speed as well.
That's for my i5 3570, GTX 1070, Intel SSD 530 series. I saw a few streamers when they where doing pre-release of 2.2 - and I think they had similar performance in the beginning. Don't know about mid to endgame tho.
All and all, 70% slowdown in fast and 200% in fastest around mid to endgame is bad (2320 could be considered the peak of the midgame, even by game logic). And, hey, eu4 and hoi4 are not capable of running for the entire game on their fastest speeds either. And that's fine, 'cause you just dial speed back one notch and all is fine, more or less (not saying there is no performance issues with these games). But Stellaris right now, ouch.. I feel bad for devs, but cmon, please fix this shite, preferably before February.

 

[KingAlamar]

That's an interesting question. I made a quick test myself, and results are, for the default Large galaxy with only UI mods with ironman enabled:
(All times are in the ballpark, not exact) 22 seconds for normal speed, 12 for fast and 6 for fastest.

I guess I have a different issue. On my "old" rig on a brand new game it takes about 22 seconds per month on fastest. On a game with the "lag array" [Sentinel] built I'm at roughly 42 seconds per month.

I may have found a culprit though. System misconfiguration showing in CPUID ... will be back to see if we have a fix

 

[TheFunMachine]

I may have found a culprit though. System misconfiguration showing in CPUID ... will be back to see if we have a fix

How do you figure out the misconfiguration? I'm curious, wouldn't mind looking into this on my own PC.

 

[KingAlamar]

How do you figure out the misconfiguration? I'm curious, wouldn't mind looking into this on my own PC.

Honestly the mistake was really bad so I'm actually embarrassed to say ... however to add to the pile of "don't do what I did" type of mistakes:

In CPUID I was watching CPU utilization, video card utilization, heat, "clocking" information, etc. When I was gathering this information to submit a bug report I noticed that the WRONG VIDEO CARD was getting utilized. Sure enough when I checked I see I plugged HDMI [tv cable] into the wrong port on my computer!! I shut down, plugged things in correctly, powered up, disabled the "onboard" [slow] video in Windows, rebooted, turned off anti-virus, and reran testing:

The times I'm seeing with my configuration corrected + Anti-Virus off:
New Game: About 5-6 seconds per month with no mods. [Was 20+ seconds per month with wrong config]
Game in year 2500 with the Sentinel Array built [no mods]: About 22 seconds per month. [Was 40+ seconds per month with wrong config]

Now the question(s) become:

• Is 5+ seconds per month on a brand new game about right?
• How far off, if at all, is 20+ seconds in year 2500+ with the Sentinel Array??

Specs:
Core i5 3570K ; 16 GB Ram ; AMD HD 7850 2GB ; Win 7 x64 ; Anti-Virus off ; 7200 RPM HDD
Single thread CPU utilization when paused and game in background: 50-60%
GPU utilization when paused and game in background: 70%

Based on Steam's listing for recommended H/W & OS I'm scraping the bottom of the CPU & GPU area for "recommended".

 

[mightnight]

I posted the most likely problem and fix for this whole thing here: https://forum.paradoxplaza.com/forum/index.php?threads/daily-tick-lag-fix.1143153/ . Please go there and drop likes so the devs can read and implement those changes.

 

[ShaTiK]

Now the question(s) become:

• Is 5+ seconds per month on a brand new game about right?
• How far off, if at all, is 20+ seconds in year 2500+ with the Sentinel Array??

About right, from what I've seen in streamers and in my games on Large galaxies. The latest Glavius AI mod had some performance fixed in it, that makes stuff better in mid to endgame. But still, 3 to 5 time increase in fastest speed as game progresses is a usual thing. Curious is that small galaxy, or zero wormholes, or even vanilla AI (it is basically braindead, so galaxy-wide POP count is way lower) doesn't seem to make that much of a difference. Mid to end game - fastest goes from around 6 seconds per month to 15-20. And it even slows down normal speed too. So, as it was stated a lot before - something REALLY wonky going on with game calculations. Stuff happening way more frequently, like many times the needed numbers, plus, I suspect, some of these calculations are done very inefficiently. I know that logistics and all these 'traveling salesman' problems are among the most difficult and calculation-heavy tasks. But still, there are ways around that

 

[Guraan]

It should be noted that I am indeed frustrated as this is not the first, second or third major issues I myself have endured after spending hundreds of dollars on Paradox games and DLC. So please take my rant as that a rant at frustration of not being able to play this awesome game because of the code issues you guys have. Sorry for being upset and I wish nothing but the best from this game and company. Also unbelievable how many people respectfully disagree with the facts I have stated. Are you just that much fanboys that you argue against FACT?

What i meant is, this is the tech forum not the customer complaints... I agree with you to 100% but this is the wrong part of the forum, and in this part of the forum we do not give a rats ass about the voting system tbh O

Preach it brother!

I try to O once you get to know it you never go back O;DD

I am wondering if this issue is linked to the one I have during MP. When I "hot join" my friend's game, I automatically get 2 in-game days behind and the whole game becomes laggy. An individual on the Paradox IRC channel over at Rizon informed me that it was an issue he always had, specifically blaming windows/linux crossplay.

I always had stuttering with stellaris, especially in the late game, but here it's mostly some kind of latency with the host. It would be nice to have an in-game diagnostic tool, to see how good the link is between players (that good old "ping").

Never heard of, plz send me a link or something to that issue.

In general, the performance fixes where never merged into 2.2.3 beta before christmas... The performance improvement @Jamor is referring to is the trade thru wormhole edge case.
I will be back from vacation on thursday and depending on who is working then we might be able to push a new beta with the promised performance changes.

//Guraan

 

[xrogaan]

Never heard of, plz send me a link or something to that issue.

The reference I have from IRC is just that. I did fire my google-fu and found similar issues, but it doesn't seem to be a crossplay issue: https://steamcommunity.com/app/281990/discussions/0/1368380934254862466/

It really is just that, the game is fine for several hours after a new game but then the interface gets a latency which seems to increase with time. You would order a fleet around, but have to wait several seconds to see the effects. The host seems unaffected. What I suspect is going on is that the host is subject to the small stuttering, and as the host slows down the client has to wait for it for each action.

If it were only regular slowness on my end I wouldn't mind, but right now the interface itself makes me wait. Hiring a new leader then be required to wait for it to appear is kind of annoying.

So not a linux/windows issue per say, but at this point I am ready to believe anything

 

[Zaiku]

Hi,

I tried Stellaris on a new iMac 5k, 27”, core i7, 32GB of memory and even if it’s better the stuttering are still there.

We look forward to the upcoming beta performance patch.
Enjoy the end of your vacation Guraan!

 

[Guraan]

The reference I have from IRC is just that. I did fire my google-fu and found similar issues, but it doesn't seem to be a crossplay issue: https://steamcommunity.com/app/281990/discussions/0/1368380934254862466/

Thank you very much!
Also, gonna shoot down that the issue is 32bit as one of the posts in the steam forum suggests before we get yet another 32bit vs 64bit rant here O

It really is just that, the game is fine for several hours after a new game but then the interface gets a latency which seems to increase with time. You would order a fleet around, but have to wait several seconds to see the effects. The host seems unaffected. What I suspect is going on is that the host is subject to the small stuttering, and as the host slows down the client has to wait for it for each action.

I believe you are correct here or at least that is one of more potential culprits.
The MP model used for stellaris and all our other games from that generation or before is in some degree lock step based, so when you hire a leader as a client you will send a "command" to the server about doing so, the server will then add this command to an execution queue thus handling the ordering of all user actions from all players, then at a given moment in the game loop the host will pop commands from this queue sending them back to its clients for execution.
So as you said, depending on if the server is experiencing performance issues the commands may be waiting in the queue for some time before being processed and/or when it sends the commands back there may be computational expensive commands executed before the command you are waiting for.
But this also includes the player on the host machine as well so either this is not the case or it is just that the host player experiences it as unaffected due to timing of sending its own command.

Another thing that crossed my mind is that i recall us bundling these commands into payload chunks before actually sending them cross the network, so basically commands with small amount of actual data will wait until buffer size reaches current MTU size or ~200ms has passed (Nagle algorithm).
MTU is confed per network interface so this may vary between machines even that it usually is somewhere ~1500 Bytes, but what hit me is that in worst case for a single command it may take `2x ~200ms + 2x network transfer time + host processing time` before it is executed. Not that this would give several seconds of delay i believe that it might be noticeable.

 

[xrogaan]

Not that this would give several seconds of delay i believe that it might be noticeable.

From your post, I can only reasonably conclude that more testing is required to properly pin point the issue. From my end, I can only wait for the next optimization patch and see if it improves anything. I don't really know about the performances on the host machine, I can only say that my friend doesn't mind the game being slow and is one of those mega galaxy / mega space battles junkie playing as a driven assimilator. Always such a joy to play with

I'm not quite sure what I can do to provide further assistance/information with this matter though.

 

[Sayakus]

Had to drop Huge galaxy game around 2350. It was unplayable.

 

[Killeraoc]

Play any game with 3dstats on and by 2220 on a large galaxy it starts to look like an EKG heart monitor. Small pulses every day (10 ticks) with a massive spike at the end of the month. I can hear the game recalculating over and over again. Gradually maxing out that one core it runs on as time ticks by. It’s like a memory leak. The game just accumulates worthless calcs and gets slower and slower to the point that it’s unplayable. Even using Fast_Forward command and taking rendering out of the picture just kicks the can down the road...the core reason of the slow down is still there. Ditto playing on a smaller galaxy. You’re just delaying the inevitable...and not by much. You can see the ticks slow down as time goes by even without graphics.

Glavinus AI mod has achieved AIs that are at least pseudo-competent at building an approximation of planetary economies...but he’s not a miracle worker. The root cause of the slowdown is something that’s outside the reach of modders. Whatever the cause it has to be addressed.

Fleet pathfinding? Trade routes? Piracy? Core AI logic? At least open up the daily tick calcs to modders and crowdsource the solution. Someone will minimize em in a way that gets things playable. As it stands it’s not possible to finish a game.

 

[Sayakus]

Actually latest Glavius AI works pretty well, game works bit better thanks to new workaround with pop jobs check, also AI can actually wage war a bit and do not commit suicide by breaking economy.

 

[Killeraoc]

I’m using it - his latest update gets the speed of the game back up to par with the base AI (which doesn’t deserve the name) with much improved behavior.

But the core slowdown is still there and still grows worse with time. With his mod or without it.

 

[Yandersen]

Dear Devs, it's Jan 6, when shall we expect a performance patch coming?

Literally, I am waiting since the last year to be able to finish the game. The UI having it's own life and the slideshow makes me so dizzy that I just can't.

 

[Guraan]

From your post, I can only reasonably conclude that more testing is required to properly pin point the issue. From my end, I can only wait for the next optimization patch and see if it improves anything.

Cannot tell when we will get to pure MP optimizations since the SP ones are a bit more critical, than ofc SP will reflect on MP O

I'm not quite sure what I can do to provide further assistance/information with this matter though.

Just continue give us feedback on future patches regarding performance, remember to read the patch notes about what changes has been done.

Dear Devs, it's Jan 6, when shall we expect a performance patch coming?

Dear customer(s), it's Jan 6 afternoon O;P
Last week there where no QA available due to vacation so right now i do not know their verdict about the current performance changes.
But right now there is a `performance branch` merge request pending acceptance that contains changes to the following major areas:
* Reduce "stuttering" by the cost of overall game speed performance.
* Rebalancing threading to favor mid/late game performance, sacrificing early game performance. (early game performance drop is barley noticeable if not comparing raw numbers)
* Various fixes to job/pop calculations for improving overall performance.

I hope this will be approved and merged the upcoming week(s).
Meanwhile i and rest of the team is continuing with general bug fixing, QoL and next batch of performance fixes that most likely will be aimed towards general game speed performance.

//Guraan