Stellaris End-Game Unplayable Even on High-End Computers

[Drowe]

I've reached the endgame crisis with the new patch and so far no significant slowdown for the default settings on a 1000 star galaxy. And my system is by far not the best one around, given that I built it 6 years ago and the only thing I've changed since then is the cooling of the CPU. Not even Windows reinstalled.

 

[LCS73]

I guess I'll have to try to give Stellaris another go then. A few months ago, Stellaris brought my PC to a screeching halt.

i7-5960X
32GB Ram
2x Titan X Pascal

 

[Soraki]

Lets just look at this statement a little bit closer shall we.
CPU minimum AMD Athlon II X4 640 @ 3.0 Ghz / or Intel Core 2 Quad 9400 @ 2.66 Ghz

nCPUs valid = 124 (intel) + 54 (AMD) = 168

GPU minimum AMD HD 5770 / or Nvidia GTX 460, with 1024MB VRAM. Latest available WHQL drivers from both manufacturers.
nGPUs valid = 34 (nVidia) + 46 (AMD) = 80

However each GPU has multiple versions which means you're probably looking at ~240 cards that are valid.

Then you have the number of RAM options which is easily in the hundreds (speed/capacity/timings/size/number of sticks). ~ 250 (and this is low)

Lets multiply these out shall we, 168*240*250 = 10,080,000 . Yes 10 million combinations. This is before we look at drivers, operating systems, antivirus programs.

There are more possible combinations of computers than there are people on the planet. It is impossible to test all of them or even most, or even a small fraction of them. Any other claim is frankly delusional.

Yes I would like the game to run better. The people saying PDX have done a bad job are free to found their own company to do a better one. Those who suggest that the engine is not suited to multi-core have been told this is not the case by the developers and those who think going to 64 bit memory addresses makes things faster are the ones completely missing the point.

So all other game companies can plan for these "endless" options...but paradox can't...

 

[RevanOn3r]

Not even going to quote the original authors uneducated complaint and lack of programming knowledge, yet alone optimization technique implementation though 32b vs 64b. When you compile your own source code, build an SQL database, write event scripts, and create an installation application (I made game when I was 17), maybe then you'll have some merit to your statements, that is, if you can provide supporting evidence.

But, I'm not here to be a prick, my son can do that when he's using the computer.

I will say this, what do you expect on a 1000 galaxy simultaneously simulating THOUSANDS of unique ships, planets, stations, populations, events, the list goes on. The depth of this game is amazing, and I have a lot of respect for the development of it and the optimization progress I have seen thus far.

Try playing 600 systems or 400, just increase the habitation rate for planets. I actually just won the game for the first time in Ironman mode the other day, ended with 2000 ships (value not total) and a fleet of 360k, which I could have easily doubled. Ran 400 systems, 200% habitation, running a basic AMD 4x core @ 2.6ghz with only 4gb ram. Had no lag whatsoever in the year 2550.

 

[DanaDark]

I suppose I can chime in on this rapidly degrading thread...

There is ONE thing that causes slow down via CPU: number of calculations made. That's it. This game looks nice but it is by no means some graphics juggernaut so we can safely assume very few people are experiencing graphics lag.

If calculations HAVE TO be made sequentially, then 600,000 cores and threads won't mean a thing. And in many cases, things have to be done sequentially. Imagine if chess players had to take their turns at the same time.

So, what can YOU do to alleviate the stress on the CPU? Fewer empires helps (this doesn't mean killing them off, it means starting with fewer). Smaller naval capacity. Fewer stars. Fewer bases. Fewer of anything, something, everything. Everything in the game results in calculations. Two identical rigs playing the exact same map setup can yield vastly different late game speeds due to the differences in the number of THINGS (ships, outposts, planets, species, etc.).

This doesn't even scratch the idea of what each person considers 'slow' or 'laggy'.

As for technology...
32bit vs 64bit, for calculations 32bit is faster. Anyone that really remembers their stuff from Computer Architecture classes knows this. The benefit of 64bit is the allowance for referencing a significantly larger number of RAM addresses over 32bit. It will yield no benefit if you do not use over 4gb ram.

Multi-Thread optimization, the OS does most of all of this. Developers can generate threads for parallel processing when it can be done (remember, not everything can be parallel) . If your OS says "F U I love core 1! Core 2 sucks and will be ignored!" then you're screwed.

Stress testing most configurations, ... ... ... ... No. Anyone that has any experience in the field will immediately look at you like you just murdered a litter of kittens.

-------

Currently I enjoy playing with 1,250 star systems, and 26 species. Regular habitability. Haven't had any lag problems at all yet. But that's for me, and my experiences are independent from others'.

 

[grommile]

So all other game companies can plan for these "endless" options...but paradox can't...

No, they don't 'plan for these "endless" options'. They pick a set of test configurations they think will be sufficient to give them confidence that it'll run on anything exceeding their published minimum spec, and test their games on those. The bigger the QA budget, the more scope they've got to increase the number of test configurations - but in practice, they're still going to be testing a tiny slice of the possible configurations.

The really striking thing in this thread is that you've got people with high-end PCs saying "Stellaris laaaaaaaags like a laggy thing in late game" and people with mid-range PCs saying "works fine for me". As such, it's not a simple case of "too many calculations". Something else is going on, and this thread is not going to get the bottom of what.