Please continue to optimize PA

The number was a fabrication, yes, not that my point hinged on it.

Problems with ubers servers have affected everybody, even me when I played on them.

So until we can play off of them, and we can, but triuly have proper our own MP servers, or hope that uber upgrades theirs, there isn't anything we can do.

Optimization isn't the issue, it's the bad servers.

But even then, people continue to play even if 2 out of 3 games are at 1 fame per second.

So I still assert, that the problem isn't be all end all as OP assumes it is.


So what if the exaggeration was a lie, that was never the point.

In all fairness, though, if the game were optimised to high heaven it could be played in the advertised fashion even on PlayFab.
Which, for me at least, acts as kind of a baseline: If the game does not even run on those specs, either the game should be optimised some more or it should be limited in scope more strongly - no use in advertising a whole solar system as a playing field when anything bigger than a small moon is bringing the game to its proverbial knees.

"The scale can go up" never meant "every computer will be able to handle every scale".
If you try to overload the servers, you will succeed. Congrats! How about trying a goal that is not that easy as stealing sweets from a baby next?

If you don't design your games/maps to avoid a stalemate, you will get a stalemate and WILL overlaod every computer:
https://forums.uberent.com/threads/lets-talk-system-design-or-how-to-avoid-stalemates.65331/

Near infinite maximum caps will hit finite processing power eventually. Nothing can be changed about that.

If the game was optimsied to oblivion and back people would just try games that would still overload every exisitng computer and would STILL complain about stawmans like "not optimsied enough".

We managed to overload a current day high end game computer with something as old as Warcraft 3. We just had to try hard enough (a mod without unit caps/conventional production).
PA has not unit caps, no sanity check against stalemate systems and a pretty high player cap. So you don't even have to try to overload a computer.

Managing expectations does help, though.

Your avatar reminds me of a time when I had to go over to someones house to play video games.... The smell of tile is still etched into my mind

We're not talking about overloading the game or pushing it to try and do spooky things. We're talking about playing the game as it was meant (and advertised) to be played. If that can't be done without making the game unplayable, then something is wrong.

In the spirit of the thread: Uber should multithread the sim as much as they possibly can, as that is the only way to really scale with bigger servers! On custom servers we can get 2-3 instead of 1 sim fps so far. Why? Because we can put all non-sim parts on an extra core, but the one core that works on the sim all alone is overloaded nevertheless.

This are the unique active users per week of PA Stats, so people who have installed the mod:


How many % of the player population use PA Stats I wonder. Really 25% already? I doubt that. If I let PA Stats list all players it has seen, even only "indirect", so this includes people who do not use PA Stats, but have been in a game with at least 1 person using it:

So even this is probably far from the actual number of players, as I am sure games are played where nobody has PA Stats enabled at all.

If anyone from Uber is still reading (I don't blame them if not - sidetracked discussion) my suggestion is this:

1) Put some logging on your servers to see what the standard sort of experience people are having when they play multiplayer games.

2) If it is indeed a overall terrible experience on average as I suspect, do one of two things (or both):
- optimize the game further
- beef up your servers

I see no reason though why you couldn't necessarily do both and get wins from both hardware and software.

And importantly, do this as a priority. Please.

/ end thread

@cola_colin " On custom servers we can get 2-3 instead of 1 sim fps so far"

So basically, you proved @igncom1 completely wrong, if I'm understanding you correctly. Even on a awesome server, the game still runs like trash.

Well it IS a question of how many units you throw at it. It doesn't run like trash really. Try to play a round of Supreme Commander with these 32 players and allow each of them a unit cap of 150. PA can handle that at a decent sim speed right now. Issue is: Players in PA tend to spam 10 times more units.

2 - 3 sim fps is pretty paltry man. Also, 32 players? Most of the games that have been running at 1fps that I've played have only had 3 - 4 players, lol.

To be honest, I think part of the fix really should also be in aiming to reduce the overall unit count too because that would help a lot. An easy way to do that would be to half the economy across the board. Overall, it's clear Uber bit off far more than they can chew when it comes to being able to handle as many units come up in everyday play on everyday maps.

And again, I repeat - if default, standard, normal, everyday, expected, as on the tin, shown in promotional videos, talked about on the website and kickstarter (how many other ways can I say this) gameplay causes unplayable lag -- something is very very wrong, somewhere.

I agree, I'd really love to see a significant improvement on the current simulation speeds in regular names. I just played 2 4vs4 AI games with 3 friends on around 4 planets, and each time at around the 25 minute mark and past that the simulation speed just got lower..and lower.. and lower.

Other than that it's truly an amazing game, this is just such a big issue for me. I'm not posting to irritate anyone just letting the devs know there's one more person suffering from this issue.

On a good server you can play a game at decent sim speed with 4000-5000 units basically. If you only have 4 players you'll have a nice game if the players understand that there is no point in area patrolling 10k dox on planets they own completely.

Everyone is suffering from this issue. That's the stupid thing and why I created this thread. We're all sitting around talking about cool upcoming feature updates and @jables is prioritizing based on our feedback but nobody is addressing the elephant in the room.

Seriously Uber. Do the logging. Ignore all the other BS in this thread. Do the logging. Test. At worst, you'll waste maybe 30 minutes writing a python script. At best, you'll discover something super critical driving negative reviews that should take almost top priority for fixing (how many reviews have you guys read that mention "unoptimized" and "performance issues" - lots on Steam) .

If it turns out I'm wrong, I'll happily back HR to the tune of $250.

Pretty much all the games I'm referring to had nowhere near 4000 units and were struggling badly. Or is this the part where you make use of the qualifier "on a good server" which, in practice, nobody is playing on.

Yep. Basically I am not looking at what Uber's servers can do. I am looking at what the best server I can get my hands on can do. I can tell you: More than ubers servers for sure. But not as much as I want them to do either, as the servers I have in my hands have a ton of cpu cores of which most end up unused so far.
So I am asking for Uber to really strongly consider multithreading in a way that scales at least per planet.
Now comes the "funny" part: This will sky rocket the sim performance on those "good servers" as they can have dozens of cores.
But if Uber does not change their official servers in a way that increases their server cost by a factor of 2x-4x they won't see any advantage on their servers, as those work with 1 pa-server per cpu-core from what we have been told.
So I guess they probably won't do much in this area until they actually have an official way to setup dedicated servers.
What I am asking for basically would make my "good servers" be super good, but you on the default servers would not notice much.
On that note, there are 2 such "good" servers that you can join right now via PA Stats.

Nice info, thanks cola.

Given the PA servers are run by PlayFab - which Uber has spun off and is pitching as an independent back-end service - it's surprising (and a bit sad) they're not running top of the line stuff. Shouldn't PA be essentially proof of how good PlayFab is as a service?

ign's reaction to this thread reminded me of how he welcomed me when I first started posting here. Passive-aggressive at first, that later turned to outright hostility. He has calmed down now in respect to me, but apparently he's turning on to other newbies now

Anyway, the OP is, of course, right. Everything that can be done to increase sim speed EXCEPT put unit caps (the lack of which is one of the game's three main attractions, next to spherical and multiple maps and large player numbers) should be done. This means:

1. Code optimization
2. Making the sim run on more than one core, as colin has been saying
3. Releasing the second stage of the server as soon as possible so that everyone can easily host custom servers and everyone can join those lobbies without PA Stats, thus taking some of the load off of Uber's servers

These steps take precedence over improved biomes. They take precedence over revamped Orbital and Naval. They take precedence over improved AI (which is extremely low on the list of priorities, since AI is said to be great already, and this is anyway at heart an MP game). And they certainly take precedence over custom map creation and whatever balance concerns people still have. The only current development that sim improvement DOESN'T take precedence over, as far as I can see, is the competitive update -- mainly a proper ladder. That's priority number 1 as far as I am concerned, so it's great that this is coming at the end of this month. I just wish it would be coming for team games too sooner rather than later, because that's all I and my friends play

They are running pretty decent servers, most of them seem to be hosted by amazon.
When hosting big dedicated servers I've used 16 core servers from amazon as well.
But from my (potentially outdated) information playfab kinda overload the servers with pa-servers. Like 8 games of PA on a single amazon-server that has 8 cpu cores. Is what we have been told earlier. This may no longer be the case. I don't know for sure. But I can tell you that even a 8 player game already profits from more than 1 core.

I would not say this takes precedence over everything else, @icycalm. Simply because Uber has more people working on PA, so they can work on many things in parallel. The server guys definitely should work on server optimization non-stop once the dedicated servers release is ready.