Bad performance ruins long term experience

[Buzz2005]

Ok thanks, not much of an improvement
approximately what I get in my busiest sector landed on a station

[Memnoch]

Hi,

I didn't want to start yet another "performance" thread so searched for one instead and found this one!

When X4 came out I bought it and refunded it shortly after. I was getting very poor performance but was using an old CPU, and as other have stated, and I have read, this game can be heavily CPU bound rather than GPU.

I recently upgraded my system to a water cooled 9900K, 32GB Corsair Dominator Platinum, along with an EVGA 1080 FTW GPU.

Does anyone here have a comparable spec. and able to offer some perforamance feedback. I am temtped to pick the game up again but I'm not sure I want to go down the road of refunding again if the game still performs terribly.

Howard

[Skeeter]

The only bad performance really is when ur on a big station platform like a wharfs even if its empty theres fps issues which could be from a bug i found that has a ship stuck and not moving because really theres like 50 ships in that one spot which tanks fps (u pop one shot at the stuck ship and boom 50 show up). Might be one off but others have reported similar thing. I mean it might not be related to hangers but if there are those type of buggy ships (the little ai drone ships) that is bugged like that but u dont notice it then it could be.

Anyhow. Like i said ive only had bad performance there, if ur flying in space u can max everything and play at say 75fps if ur monitors 75hz with freesync on in my case.

[AquilaRossa]

I am back to X after being away since Terran Conflict and building those spaghetti and meatballs mega stations. The new stations are much nicer. I am liking what I am seeing as far as the concept and ambitions for the game goes, but this thread is exactly my experience and those stations seem to be where the worst happens. After a few game days FPS tanks near stations. Must be resolved because getting blueprints is a key art of the plot and trying to do that at 20 FPS in a space suit is a nightmare.


I still believe the game is not making best use of CPU resources available to it though. I did monitoring on the CPU and when low FPS and stuttering was going on half the cores would get maxed for a moment and then not, so the average for each core that the game is using was below 50%. it was the same vanilla later in a play through, but just low FPS and not the stutter so much. I also noticed it using one thread of a core and not the other, yet on another core it would be using both -- meanwhile four other cores on my 2700X are sitting idle not getting used it seems to me (so chucking a 12 core Ryzen in will help due to better IPC, but I doubt the game will use all those extra cores). GPU is just a RX590 and is barely breaking a sweat. I am running the game on medium with AA and anisostropic filtering turned off, so I could confirm the CPU was the mill stone.

Long time since X3. Longer than I realized. No Egosoft emails in my gmail so i could find my old user name here, so I had to make a new account.

[Memnoch]

I may hold of on that for now then. Despite the assurances of the developers, or perhaps their explanations for what is going on, I don't thing Egosoft have done a very good job of optimising their engine and it makes me wonder how much hardware you would need to throw at it to run at decent frame rates, no matter what is going on.

I read what CBJ wrote about external tools not being a very accurate way to measure performance i.e. looking at your task manager and looking at your 16 cores, then looking back at the game and thinking "Huh?".

Perhaps ES has access to tools that we don't for this very thing? But it does make me wonder, from a software QA point of view, how their internal testing goes, what hardware they use, and what situations they would test, where someone would make the call "Yes, this is acceptable for the public to buy". 20fps for me is not acceptable.

Howard

[Xenon_Slayer]

We are aware of several aspects of the late game which may lower performance. Some of these are side-effects of issues such as factions building too many ships/stations, so in fixing (and patching) these, there'll hopefully be a noticable improvement. Of course, it's all situational and doesn't help if the cause is player objects. A sector with several mega-complexes is always going to tank performance, but we are looking at general optimisations for cases such as these.

[Memnoch]

I guess what users have reported, at least these with very high end systems, is that the games perfornmance drops at some point, and they look at the CPU and GPU load and don't see anything that would inidcate either is being maxed out in any way, and you have to conclude that its not your system at fault.

For instance, until a few weeks back I had been running a very old, original, Nehalem chip. The i7 975X. I could play almost any game with that except most Unity games. I could play at 80+fps in Ghost Recon:Wildlands, with everything turned to max, yet a game like Satellite Rain had terrible performance. Similarly, the original Division was flawless, and yet The Division 2 was bad. Changing my CPU out for a 9900K made all those problems go away.

But there is only so much we as users can do, and for everything else we need to wait for the developers to fix things for us. The long running convcern for me is that there is this drive to add content over time, instead of stopping and fixing what is already there, and only then moving forward.

What would be nice to see would be the equivalent of Quakes timedemos. The recent shooter World War Z had something like this. It would at least allow users to compare performance results on something approximating an even keel. Perhaps a save game state in a busy sector would do, from a load point for view, but you would need some scripted viewingpoint changes so that everyone does exactly the same thing.

From what I have read on Reddit recently, perhaps we just have to accept that any game that has large scale simulations going on under the hood will end up being a performance hog, but even that implies that throwing hardware at it can mitigate things. Perhaps more optimal use of multi-core systems is key? Early days for Egosoft on that front perhaps, based on their previous games, and previous conversation over the years where that wasn't even a thing.

[SPiDER]

I hope that patch comes soon even in completely empty Xenon sectors you cant do much.in Trinity Sanctum III I'm Nearly upto the magical 500 stations...

My game was a new 2.5 start with no 10x10x10 complex's

[Player]

I'm waiting for this patch too. Now after almost 300 hours and especially after 2.50 the game is really getting slower and slower with low fps. I7 8700/16 gb ram e GTX 1070.

[DaMuncha]

To Egosoft: When your game runs at 10 fps on stations, your doing something wrong.

[Player]

To Egosoft: When your game runs at 10 fps on stations, your doing something wrong.

Agree. Waiting for the patch. Any date for release?

[plchung3]

We are aware of several aspects of the late game which may lower performance. Some of these are side-effects of issues such as factions building too many ships/stations, so in fixing (and patching) these, there'll hopefully be a noticable improvement. Of course, it's all situational and doesn't help if the cause is player objects. A sector with several mega-complexes is always going to tank performance, but we are looking at general optimisations for cases such as these.

Thanks for making good games and keeping up the good work.
I am using a dual Xeon CPU machine (12 cores 24 threads) with 72GB ram and a RTX 2070.

I found that when the fps drops actually many of my CPU cores are idle and GPU also pretty light loaded. I am a programmer myself and understand it is very hard and maybe too late to re-engineer the game engine code drastically now, but I still recommend considering utilizing more cores whenever possible if the simulation is CPU bounded. Nowadays sticking with single core wont get the speed up to any where, you need to parallelize the simulations.

[pref]

I am using a dual Xeon CPU machine (12 cores 24 threads) with 72GB ram and a RTX 2070.

Those 12 cores are mostly useless in X4, as you will sooner be limited by core speed. Xeon isn't exactly for gaming, i can imagine you getting more fps with a 4700k even.

X4 utilises CPU much more then your usual AAA titles, but not all tasks can be parallelized. Guess the render loop is one of those.

[Lord Crc]

Those 12 cores are mostly useless in X4, as you will sooner be limited by core speed. Xeon isn't exactly for gaming, i can imagine you getting more fps with a 4700k even.

Indeed. I just upgraded my desktop from a Threadripper 1920X to a Ryzen 3800X, and typical FPS in non-empty scenes in X4 went from low 60s to mid 80s. It's only actively using two or three cores, and then some minor usage on the remaining.

edit: the Threadripper is now an awesome NAS/virtualization server

[ezra-r]

Of course we could improve map performance by taking away functionality, but I'm not sure people would be as happy with that as you seem to imply. Being able to make that choice, as a player, is exactly why there are filters, and exactly why some of those filters have a significant effect on performance.

yes, please do not touch the map if its not for improvements.

[Imperial Good]

I am using a dual Xeon CPU machine (12 cores 24 threads) with 72GB ram and a RTX 2070.

Multiple NUMA domains is probably not good for gaming. Games are almost certainly not NUMA aware and so would incur heavy communication overhead penalties if they tried to use more than 1 of the CPU sockets (NUMA domains) and associated memory. At best the game would be smart enough to use just the fastest processor and allocate all memory to it.

You could try explicitly forcing X4 to use a single NUMA node using task manager. Hopefully the OS would then implicitly allocate all memory for X4 to that node. This might result in improved performance.

Making a game fully NUMA aware is not easy. One would have to fragment game state into separate parts, each on a separate NUMA nodes with mostly threads of that NUMA node manipulating that state. All game data and required assets would need to be duplicated between NUMA nodes so that each NUMA node can operate independently from each other. The only communication going between NUMA nodes would then be cases where state is influenced outside the NUMA node and should occur fairly rarely. In theory this would deliver very good scaling due to the large increase in memory bandwidth utilization and low communication overhead.

Between consumer hardware only offering 1 NUMA node (multiple is HEDT or server) and that games struggling to scale with thread count, there really is no reason for developers to make NUMA aware games. At best they could try limiting execution and memory allocation to only 1 node, since such nodes usually have 6-32 cores which is likely beyond the thread count the game will scale to anyway.

Guess the render loop is one of those.

Game state usually cannot be updated at the same time draw preparation is running due to its dependency on game state. This is why a little improvement in CPU performance can result in a large improvement in frame rate since the fraction of time that game state update is running is proportionally less allow for many more draw preparations to be scheduled without decreasing simulation speed and hence more frames to be drawn.

[phrozen1]

I tested playing with 1440p on my not-so-new machine and the differences in performance depending on the surroundings became very obvious.
And it seems like different types of fog are a real issue here. In some areas it's quite ok but in others i only get like 1/5 of usual fps. (Grand Exchange IV is a extreme example)
Also on stations it seems like there is too much details beeing rendered which you don't see or would even notice.

I really hope for optimisations there and/or more graphics-options for lod of fog, asteroids and station-interiors.

[alexthespaniard]

I really like the 3D map looks awesome. Myb can be implemented as an option.

[adeine]

I tested playing with 1440p on my not-so-new machine and the differences in performance depending on the surroundings became very obvious.
And it seems like different types of fog are a real issue here. In some areas it's quite ok but in others i only get like 1/5 of usual fps. (Grand Exchange IV is a extreme example)
Also on stations it seems like there is too much details beeing rendered which you don't see or would even notice.

I really hope for optimisations there and/or more graphics-options for lod of fog, asteroids and station-interiors.

Yeah, I've found the same thing. Some types of the fog/asteroids mix can really cripple framerates.

Hopefully something that can be looked at, optimisation-wise, once more critical issues with the game are sorted out. For the time being, best to build elsewhere

[pref]

Game state usually cannot be updated at the same time draw preparation is running due to its dependency on game state.

I'd imagine these run in parallel, and rendering just takes a sample of the universe sim at every frame. Less things need to be synced the better probably.