Jump to content

Will DU ever be optimized? Can you please fix the LAG?


Sethioz

Recommended Posts

I'm just wondering, will dual universe ever get optimized? my pc specs are not important, what IS important, is that my pc can run games like dying light 2, forza horizon 5, resident evil village, cyberpunk 2077 ..etc in 1440p or even in 4k @ 60fps, but DU i have to run in 1080p with lowest possible settings and i get only 15 - 40fps at best. if i look straight into the sky, i get 50-60fps, but even then it sometimes lags below, it's extremely unstable and i don't understand WHY?

 

this is clearly serious optimization issues, why is it running so bad? also i dont remember having such bad lag in early access, i remember getting 60fps in most areas, i see NO visual improvement to graphics, so what is going on?

 

if i'm correct, DU has been in development for about 8 years, so technically it's 8 years old game and my PC used to be best possible about 5-6 years ago, so DU should run like a dream on my pc, but instead it lags like a broken bag.

i mean just look at the graphics, du graphics are nothing amazing, it's very basic. trees, grass and generic 3D models are very basic. there are other games with similar graphics and i can run them in native 4k maxed out settings in 60fps. dirt rally 2 is one of them, it has much better graphics and runs like a dream in 4k. snowrunner has quite complex terrain physics and amazing graphics and it also runs like dream in 4k, so why is DU lagging so bad even in areas with very few constructs? even my M core ship lags down to 20-30fps over the ocean, WHY?

 

first thing i noticed about DU, is that graphics settings lack any actual settings, for example you're not able to change shadow distance and shadows are real FPS killer in any game, i did turn shadows off, but it doesn't really improve fps, so i keep them on low as without shadows it looks horrible. some said try changing CPU threads, but that won't change anything, because my CPU is not under 100% use, my CPU barely gets 30-50% use while playing and that's with other stuff running on background, so CPU is not the issue, while my GPU is under constant 100% use, so clearly the issue is with graphics.

 

PlanetSide 2 has about similar graphics as dual universe and i can run ps2 in native 4k maxed out settings, so what's the deal with DU? why can't i run DU in 4k with such basic graphics?

will you (the developers) ever going to optimize it so you dont need 2000eur graphics card to run it?

 

there must be some SERIOUS optimization issues, because DU literally runs worse than any other game i have ever played and i've played over 1000 games for sure. ok ok .. not entirely true, i've seen some cheap indie games that have even worse graphics and lag, but DU is definetely in the top 5% of worst optimized games.

 

can you please work on optimization? re-work how graphics are rendered, because it's so demotivating having constant 20-30fps.

 

Suggestions:

+ add AMD fidelity upscaler, because not all graphics cards support DLSS. DLSS is only for latest nvidia cards, so AMD and lower nvidia cards can't use that. AMD fidelity however, also works on nvidia cards, i've used it in other games and it works quite nice without ruining quality too much.

 

+ add option to reduce shadow distance and resolution.

 

+ add other graphics settings that other games have, such as anti-aliasing.

 

+ fix vsync (I noticed that turning vsync OFF increases me fps by a good 5-15 fps, depending on areas and NO, i'm not talking about frame lock, i'm PC expert and i know what vsync is and how it works. my monitor is 60hz, so i know going over 60fps won't have any effect as my monitor is only able to show 60, i'm talking about increasing fps from like 40 (vsync on) to like 50 (vsync off). so instead of using vsync, i used nvidia settings and locked du framerate to 60 and turned vsync off. vsync should not cause lag, it should actually make things smoother and better.

 

+ optimized the way graphics are drawn, because clearly there's something wrong if basic graphics require 10x more power than they should require. such basic graphics should run 60fps even on GTX 680. in fact i have GTX 680 in one other PC and that can still run most games on acceptable level.

 

 

If anyone has any other suggestions how to make DU run better, let me know (i'm not talking in-game settings, cuz obviously i have tried them all as i'm pc expert, i'm talking if there's some tricks with some config files somewhere, like some games make some config files and you're able to mess with them to change things that cannot be changed inside game and they can greatly improve performance, so maybe DU has something similar?)

 

thanks and i really hope DU devs will optimize graphics!

Link to comment
Share on other sites

Not for games. Determining that your fancy new algorithms and ideas for a game can realistically run smoothly, is something that is done at a very early stage since it will have HUGE ramifications on the entire game. And if you don't... Well you risk getting another Dual Universe.

Link to comment
Share on other sites

Most of those assets in other games are optimised. A vehicle/building in planetside 2 has been optimized to run better. Either by stretching polygons so less rendering is needed. Moving lights so less shadows are created in one area.  In DU very few players are optimising their ships and buildings. Basically in a voxel based game where players can create any shape/size place lights and shadows anywhere you will never get the same performance. 

 

Also I just cannot recommend running in 4k. especially running a 1080. Heck I plan on upgrading to a 4090 and will still not upgrade to 4k. 

Link to comment
Share on other sites

10 hours ago, CptLoRes said:

Not for games. Determining that your fancy new algorithms and ideas for a game can realistically run smoothly, is something that is done at a very early stage since it will have HUGE ramifications on the entire game. And if you don't... Well you risk getting another Dual Universe.

true, games should be optimized before adding any content, make a base concept and then optimize until it runs smooth as tetris!

 

13 hours ago, Zeddrick said:

Usually programmers start optimising software once they've finished writing it.

completely wrong, it's other way around. think of game engine as foundation for house, if foundation is weak, entire house will eventually collapse. this can be seen on some large buildings, where engineers didn't do proper research, ground started sinking and as result building started sinking. just couple months ago i watched a video on some huge skyscraper that has been sinking for years and now they trying to patch it up by digging into the foundation and adding more support pillars and concrete, but it's big waste of time and resources .. if they would have done it right in first place, it would never have that issue.

 

same goes for games, you gotta optimize game as early as possible, you make basics work and then you optimize and make sure it runs smooth. adding content doesn't make it lag again. for example if 3D models are rendered in optimized way, then it won't change this base algorithm when you're adding more 3D models.

 

i don't write my own code out of scratch, but i know how programming works and i can easily optimize somebody else's code if i see it, i can point out what needs to be changed ..etc.

 

7 hours ago, RugesV said:

Most of those assets in other games are optimised. A vehicle/building in planetside 2 has been optimized to run better. Either by stretching polygons so less rendering is needed. Moving lights so less shadows are created in one area.  In DU very few players are optimising their ships and buildings. Basically in a voxel based game where players can create any shape/size place lights and shadows anywhere you will never get the same performance. 

 

Also I just cannot recommend running in 4k. especially running a 1080. Heck I plan on upgrading to a 4090 and will still not upgrade to 4k. 

 

it has nothing to do with optimize the build itself, even when looking at nothing with no constructs nearby, i still lag, it's the base game itself that lags. basic environment is already laggy for me. it doesn't matter how big the build is, there's only so much that can fit on your screen, that's the part that GPU has to render, everything inside or other side of ship should never render, until you're about to look at them. this is how optimization works, things that are not visible, should never render in. so having ship with 10000 lights and elements, shouldn't cause any lag as GPU should only load what's visible on your screen. regardless of your POV, there's still only limited amount of stuff that fits on your screen, as soon as something is out of your view, it should no longer put load on GPU.

 

which is one of the problems i noticed fast, DU renders in stuff that is not visible to you, which is WORST thing a developer can do. for example lights go thru voxels, it means graphics card has to do a lot more work if there are lights behind a wall. in fact i see this kind of lazyness in many games nowdays. i remember playing old games such as tomb raider 1 2 3 4 5, doom 2 ..etc and i don't remember any area where lights or shadows were glitching thru walls, that's because back then hardware was very limited and devs really had to optimize every bit, but as result games ran smooth and didn't have many visual glitches that nowday games do have.

 

i never said i run it in 4k, i said it SHOULD run smooth even in 4k. i have to run it in 1080p and it still lags really bad, my average fps is probably 35, it's quite cry-worthy, considering that all other games run quite fine in 1440p (not talking about some extremely badly optimized indie games that lag even in 720p).

 

 

 

SIGNS OF BAD OPTIMIZATION:

1: if i look straight down at floor that has only voxels or terrain, it STILL lags, why? there's NOTHING to render, so why does GPU work so hard to render something that even 15 year old computer would be able to render in 4k 60fps? that's one of the biggest issues, with almost NOTHING on screen or even nearby, it still puts heavy load on your GPU.

 

2: shadows have insanely high distance for no reason, in fact i think shadow range is infinite. i often see shadows from like 10km+ away, altho turning shadows completely off doesn't help much .. so i wouldn't be surprised if shadows are still calcualted and rendered, but they just hide them from view. that's only explanation why messing with shadow quality or turning them completely off has almost no increase/decerase in framerate.

 

3: lights glitch thru walls/objects. this is another big issue in many games, if you're behind a wall and looking at light source, then that light shouldn't even render in, it should be invisible and only known to game/server that there is light with xx properties, only when you move so that light is about to become visible, should it be rendered by graphics card. by doing that, GPU doesn't have to do any work until up to very last moment before object is about to be rendered in.

I would rather have some stuff not rendering in-time, than have lag. then they could add option to adjust the hidden objects draw distance, so if you have issue that objects don't render in when you look at them, you can increase that option a bit or vice versa. this can find balance between rendering objects and framerate.

 

4: impossible to have DU on HDD, it has severe stuttering and screen freeze all the time, because game constantly reads/writes from/to HDD, which is clearly bad optimization, that's what RAM and pagefile is for. my pagefile is on SSD, yet DU still uses HDD over RAM or pagefile. Dying Light 2 and Cyberpunk 2077 require massive amounts of data to be loaded too, but they have made it work with HDD. in fact dying light 2 is on my HDD and when i start the game, i have to wait about 2 minutes before i can move, otherwise it will stutter and lag, but after that wait time, it runs smooth with no further stuttering or lagging, same goes for cyberpunk 2077, sometimes it does have issues catching up, but then it just loads very low quality models, like traffic lights or NPC cars are very low quality until HDD can catch up, but it doesn't stutter ... but DU? it just straight up stutters and stops me in my tracks, entire screen just freezes until it can do whatever it's doing.

 

5: turning vsync on causes me to lose about 10fps in most areas. turning vsync on/off shouldn't affect performance in any way, vsync is only designed to prevent GPU from rendering more frames than your monitor can display, which in my case is 60fps (60hz monitor). so with vsync ON, i get about 35fps and with vsync off i get 45fps in SAME exact area. i literally didn't move at all, just pressed ESC, went into menu and changed vsync. i've seen this in some other games too, that having vsync either on/off causes lag. in some games its other way around, that vsync off causes lag, while vsync on works fine.

 

 

IDEA about terrain rendering:

terrain is currently loaded in "default" state and then player changes are loaded, why is this needed? DU is persistent universe, just save the changes permanently. i've had this idea for a long time "dynamic game engine" that re-writes itself as you play, which would be perfect fit for DU.

right now entire DU world is kept in some kind of database, but that's already a big lag maker. instead of keeping things in database, make it render into permanent model. so if someone does work on terrain, then those changes will become permanent part of game, even if they hard wipe, those changes would never get wiped. this method would remove any and all lag that is related to terrain that has been altered.

 

 

but bottom line here, is that early access didn't lag so bad, i'm not talking about in extremely populated areas, i'm talking about like on sanctuary or haven. i was able to play in 1440p 60fps on same PC, no issues. it only lagged near markets cuz they were FULL of builds, way more than now as there was no 7 day park timer, so everything was just abandonded all over, yet it still ran better than now. so what have they done?

 

and don't forget that i had it on HDD before, but full release doesn't work on HDD at all, it stutters so bad that it's not playable. i see no visual improvements at all.

wish i could buy new GPU, but my income is below cry-worthy, i can't even afford proper food. i think its quite important to have optimized games.

 

new GPUs might be powerful, but that's not an excuse to have badly optimized game or just because SSDs are very common nowdays, doesn't mean that game should be unplayable on HDD.

also i wanted to mention that my GPU has 8gb VRAM and game barely uses like 3-4gb, yet i see some textures unloading, like if i switch between 3rd/1st person view, i see some textures unrender, if there's enough VRAM, why not just keep them in VRAM and re-use when needed? that's again more work for GPU and SSD/HDD, it has to reload those textures. if you're flying your ship, it should keep textures in VRAM and re-use them when needed, no need to unload them if there's enough VRAM to go around.

 

If any developers are reading, please focus on optimizing the game. with these basic graphics game should run smooth as silk on GTX 1080.

 

 

Link to comment
Share on other sites

12 hours ago, RugesV said:

Most of those assets in other games are optimised. A vehicle/building in planetside 2 has been optimized to run better. Either by stretching polygons so less rendering is needed. Moving lights so less shadows are created in one area.  In DU very few players are optimising their ships and buildings. Basically in a voxel based game where players can create any shape/size place lights and shadows anywhere you will never get the same performance. 

 

Also I just cannot recommend running in 4k. especially running a 1080. Heck I plan on upgrading to a 4090 and will still not upgrade to 4k. 

On of the things NQ did to try and improve performance was that voxels are converted into a normal polygon model when you leave the build mode. This allows them to use the built in traditional polygon optimization tricks and LOD rendering methods that come with the 3D engine.

 

But there must be something fundamentally wrong with how NQ does rendering. because excluding a visit to Market 6, there is nothing show on the screen in DU that justify the slow rendering (FPS) performance. And streaming content from slow servers should not impact FPS performance on the client side, since loading new data and rendering existing data are two completely separate tasks.

 

So the argument is actually that slow servers should lead to higher FPS, since there will be less data to show when it has not loaded yet.

Which leads back to my point that it appears to be something fundamentally wrong with how NQ does 3D rendering. Probably because they for a LONG time had a "just make it work somewhat, and we will improve it later" attitude. And as all experienced developers knows, the fixing later time never comes around. And sometimes it also turns out that stuff can't be fixed and has to be completely changed. Which then becomes a HUGE problem when you have used this as the foundation to built an entire game on top of.

Link to comment
Share on other sites

3 hours ago, Sethioz said:

completely wrong, it's other way around. think of game engine as foundation for house, if foundation is weak, entire house will eventually collapse. this can be seen on some large buildings, where engineers didn't do proper research, ground started sinking and as result building started sinking. just couple months ago i watched a video on some huge skyscraper that has been sinking for years and now they trying to patch it up by digging into the foundation and adding more support pillars and concrete, but it's big waste of time and resources .. if they would have done it right in first place, it would never have that issue.

 

same goes for games, you gotta optimize game as early as possible, you make basics work and then you optimize and make sure it runs smooth. adding content doesn't make it lag again. for example if 3D models are rendered in optimized way, then it won't change this base algorithm when you're adding more 3D models.

 

i don't write my own code out of scratch, but i know how programming works and i can easily optimize somebody else's code if i see it, i can point out what needs to be changed ..etc.

 

Actually I was being a little sarcastic there -- I would rather have a finished game with more actual actual well designed game in it (gameplay loops, things like missions actually being part of the in-game economy, reasons for territory wars, etc) even if it's not optimised than a fully optimised shiny, but empty, game so I can walk around my pointless city construct at 60 FPS and look at all the other players wandering about wondering what to do now they finished building.

 

But I do write code from scratch from time to time and for this sort of very complex code it's really silly to try to optimise it right at the start.  Optimisation is hard and painstaking.  You spend hours looking at small pieces of code in the NVIDIA profiler (or whatever) trying to see where your code is memory bound, where it's bound by the available instruction pipelines, where it's bound by the speed of accessing global vs local memory, can you make the code use a register instead of a local memory reference, etc.  It takes a long time and what you really don't want to do after you spent all that time is start refactoring the code and/or changing the structure or logic of the code.  Because when you do that you have to start doing a lot of the optimisation all over again.
 

One would hope that, at this stage, al the underlying logic of the game was settled but we are still talking about redesigning a lot of the features of the game and, IMO, some of the fundemental design decisions of the game (using the client to do the physics simulation for an object, for example) are going to need re-thinking in order to make it a scalable and sustainable game without wierd behaviours (like ships freezing 4km up).  If it were me I'd finish all of that and get it right before I started to optimise the dead-ends.

Link to comment
Share on other sites

14 hours ago, CptLoRes said:

On of the things NQ did to try and improve performance was that voxels are converted into a normal polygon model when you leave the build mode. This allows them to use the built in traditional polygon optimization tricks and LOD rendering methods that come with the 3D engine.

 

But there must be something fundamentally wrong with how NQ does rendering. because excluding a visit to Market 6, there is nothing show on the screen in DU that justify the slow rendering (FPS) performance. And streaming content from slow servers should not impact FPS performance on the client side, since loading new data and rendering existing data are two completely separate tasks.

 

So the argument is actually that slow servers should lead to higher FPS, since there will be less data to show when it has not loaded yet.

Which leads back to my point that it appears to be something fundamentally wrong with how NQ does 3D rendering. Probably because they for a LONG time had a "just make it work somewhat, and we will improve it later" attitude. And as all experienced developers knows, the fixing later time never comes around. And sometimes it also turns out that stuff can't be fixed and has to be completely changed. Which then becomes a HUGE problem when you have used this as the foundation to built an entire game on top of.

very true, i've seen this nonsense since the very first multiplayer games, where people are like "you are lagging me with your high ping" or like "you're lagging the entire server with your high ping".

i have noticed this too, that there's something very wrong on base level, because DU often freezes or lags my screen when it can't keep up with downloading data from server, like WHY does it cause my FPS to drop? FPS = frames per second, which is directly related to GPU (graphics processing unit), so having download issues should never cause your framerate to drop as it has nothing to do with GPU.

 

14 hours ago, Zeddrick said:

If it were me I'd finish all of that and get it right before I started to optimise the dead-ends.

no offense, but that's the problem .. lot of people think this way and that's why games are so badly optimized nowdays. this is not how you program. you make the very basics and make it run smooth, then and ONLY then, you start adding more content and fixing other bugs and glitches.

like i said earlier, if you build a house on bad / weak foundation, no matter how well you build rest of the house, it will never stand and it will start sinking and breaking due the bad foundation, same exact concept goes for games -- FIRST you optimize the game engines itself, then you add and fix content.

 

For example right now in DU, i'm working on L core size puzzle maze, i didn't just make 100 rooms and then start optimizing and realize that things are all wrong and don't connect up, i started with 1 room at a time. I currently have about 6-7 rooms that are fully working, with traps and hidden areas ..etc. I made sure each room works exacatly as intended before moving on to next one.

only thing i'm going to do later, is decorating and replacing materials to make it more immersive. i even asked few friends to test it out and make sure it works, before moving onto next room. i had some issues with lasers, seems like they're like engines, even if side is blocked, they sometimes don't register player being in the zone, so i had to mess a lot with them to get it right. i've tested all rooms at least like 20 times (each).

this concept goes for almost everything, start small and make sure entire system works on base level and is solid, only then start expanding / adding content.

 

 

I really hope devs are taking notes from this topic. i don't remember exactly which game it was as i've played so many over the past 20 years, but i think it was Wreckfest, where they changed entire game engine as game wasn't performing as well as they wanted.

 

unfortunetely from experience, there really aren't many devs who are willing to go the extra mile to fix poor performance, because devs are always rich or at least rich enough to buy best possible hardware and they don't get any lag issues, so they don't understand how frustrating it is for those who can't afford to buy new GPU every 2-3 years.

at least DU uses 100% of GPU, i've seen some games that lag down the toilet real bad and i only see like 20-30% usage on my GPU, those are the worst.

Link to comment
Share on other sites

On 1/22/2023 at 8:45 AM, CptLoRes said:

Which leads back to my point that it appears to be something fundamentally wrong with how NQ does 3D rendering. Probably because they for a LONG time had a "just make it work somewhat, and we will improve it later" attitude. And as all experienced developers knows, the fixing later time never comes around.

 

Which is also one of the reasons cited for FF14's infamous mulligan. They had a pervasive attitude of "we will fix it after launch", and this is a horrible idea for a subscription-based MMO...or any product in general, really. 

 

10 hours ago, Sethioz said:

I really hope devs are taking notes from this topic. i don't remember exactly which game it was as i've played so many over the past 20 years, but i think it was Wreckfest, where they changed entire game engine as game wasn't performing as well as they wanted.

 

Let's be realistic about that: NQ picked an obscure engine that isn't typically used for gaming because they insisted that they must have a 64-bit coordinate system (which is not true at all) and were willing to sacrifice anything to get that.

 

Unigen2 was picked because of laziness and ignorance -- the founder and early prototyper hadn't worked a day in game dev, he had no clue how critical it is to pick a robust engine...so he picked one that would get closer to their apparent goals right away without any real study or experience. 

 

I agree that DU's performance flaws start at the engine level...but for a studio that has taken 8 years just to realize they needed at least 1 PvE mission...they won't be re-platforming, lol (never mind writing their own engine, as is typical for most MMOs)

Edited by blundertwink
Link to comment
Share on other sites

On 1/23/2023 at 9:27 AM, sHuRuLuNi said:

I like your videos. And your accent.
Especially when you say stuff like "look at that MANIAC!" :D

 

Which, BTW, currently is the only interaction I have with DU.

ah ah offtopic, but thanks xD

 

On 1/23/2023 at 3:50 PM, blundertwink said:

 

Which is also one of the reasons cited for FF14's infamous mulligan. They had a pervasive attitude of "we will fix it after launch", and this is a horrible idea for a subscription-based MMO...or any product in general, really. 

 

 

Let's be realistic about that: NQ picked an obscure engine that isn't typically used for gaming because they insisted that they must have a 64-bit coordinate system (which is not true at all) and were willing to sacrifice anything to get that.

 

Unigen2 was picked because of laziness and ignorance -- the founder and early prototyper hadn't worked a day in game dev, he had no clue how critical it is to pick a robust engine...so he picked one that would get closer to their apparent goals right away without any real study or experience. 

 

I agree that DU's performance flaws start at the engine level...but for a studio that has taken 8 years just to realize they needed at least 1 PvE mission...they won't be re-platforming, lol (never mind writing their own engine, as is typical for most MMOs)

 

very true, i didn't know about their past decisions tho, i haven't kept my eye on it at all. in fact i heard about the game only about 3 years ago when i was invited to check it out. i dont know about the game engine, but i'm sure it's possible to fix the lag.

 

for example, they already have DLSS, why not add AMD fidelity? i have seen lot of games lately that use AMD fidelity upscaler and it makes it bareable and playable on cards that don't support DLSS. from what i've seen, it can be added to any game engine.

Link to comment
Share on other sites

After reading the various posts, there are many very good arguments and ideas for resolution.
 

I think, like you, that the optimization should be done as the development goes on. Each part should be optimized at the time of release. 
It's not just a matter of releasing a feature, it has to work, and the way it works is also through optimization and usability.

Of course, some optimization must be done independently of the development of the features because they affect several modules.


But for example, the fact that when leaving an industry unit, we sometimes have a freeze that can last 5secs is not normal. It doesn't matter that the communication with the server is not yet finished. The user is not "informed" but just blocked for x seconds. In a normal world, he should have at worst a loading information or simply that the task is done in background.

But should never be stuck.
So in that case, the dev should have take in consideration that the communication with the server is not instant and have handle this. Not just freezing the player.

Link to comment
Share on other sites

On 1/22/2023 at 9:51 AM, Sethioz said:

completely wrong, it's other way around. think of game engine as foundation for house, if foundation is weak, entire house will eventually collapse. this can be seen on some large buildings, where engineers didn't do proper research, ground started sinking and as result building started sinking. just couple months ago i watched a video on some huge skyscraper that has been sinking for years and now they trying to patch it up by digging into the foundation and adding more support pillars and concrete, but it's big waste of time and resources .. if they would have done it right in first place, it would never have that issue.

 

same goes for games, you gotta optimize game as early as possible, you make basics work and then you optimize and make sure it runs smooth. adding content doesn't make it lag again. for example if 3D models are rendered in optimized way, then it won't change this base algorithm when you're adding more 3D models.

 

i don't write my own code out of scratch, but i know how programming works and i can easily optimize somebody else's code if i see it, i can point out what needs to be changed ..etc.

 

 

 


No, it's not completely wrong.  Although your example does work in your simple scenario it's just not relevant for a big, grown up software project where there are managers who have a giant list of things for developers to spend time on and need to select the ones which generate the most value.  Say you can write a piece of code in a week and optimise it in 3 weeks.  That 3 week task is going to be on the list.  Now say the code is only taking up 0.1% of the CPU for users running the game.  The absolute most that 3 weeks of developer time is going to get is a tiny speedup even if the sub-optimal version is 100 times slower (or whatever).  If you try to optimise *every single piece of code* you will spend 4 times as long developing everything and end up with the sort of product where development is really slow and the product is brittle because devs are constantly making tiny changes to the logic for no reason other than premature optimisation.  Actually that does sound a bit familiar.....

What you need to do in order to do proper optimisation is look at the whole stack running and target the optimisation where it actually matters.  You use stuff like call stack sample profiling and flame graphs to identify the areas where optimisation has a real, tangible benefit and that's where you allocate the developers time.  Perhaps also measure memory allocation patterns and network requests to see what can be combined/reused/cached/etc.  But you can't do that until you have everything there.  Something might be taking 20% of resources now but later when another feature is added or something changes then that might not be true any more.  An example would be optimising the voxel->mesh rendering and then adding the mesh server later, meaning now the optimised code only runs when you enter build mode instead of every time you see a new construct so the time spent optimising it bought you nothing.  Once you have the big picture you pick your targets, cost them and 'We could get a 20% game engine speedup with 3 person-weeks' goes onto the backlog alongside 'fix XXX bug', 'add YYY feature' and 'make dancing avatars' and then the team selects based on whatever goals it has been set.

As it happens I don't think DU has a massive problem with lower level optimisations.  My PC has a 2070 in it (I think) which is not bad but also not bleeding edge and it rarely gets up to above 50% utilisation even when the game is laggy and frame dropping.  Likewise I have about 10 gen10 intel vcores which don't often get thrashed very hard either.  What I see is memory use crawls up over time and CPU use/lag goes up to match, which is a sign that the game has memory leaks.  Also when there are a lot of network events (I go to market 6 and lots of things load) then it can get really laggy.  That's a sign of bad server-side or network code usually and the game does thrash the network sometimes. 

For me the game started to get worse in performance terms when there was a project to move from a dynamo-db based storage tier to a hand-crafted backend.  At that point I started getting lag, poor framerates and problems loading constructs in various places (flying over the ODY base in beta at a 3k height, for example).  I don't find this too surprising really as the decision to spend an 'innovation token' re-inventing distributed operational store tech rather than using one someone else already made seems to have been motivated by lowering operational costs rather than making things better.  Generally you get what you pay for with this sort of thing.

Link to comment
Share on other sites

10 hours ago, Leniver said:

After reading the various posts, there are many very good arguments and ideas for resolution.
 

I think, like you, that the optimization should be done as the development goes on. Each part should be optimized at the time of release. 
It's not just a matter of releasing a feature, it has to work, and the way it works is also through optimization and usability.

Of course, some optimization must be done independently of the development of the features because they affect several modules.


But for example, the fact that when leaving an industry unit, we sometimes have a freeze that can last 5secs is not normal. It doesn't matter that the communication with the server is not yet finished. The user is not "informed" but just blocked for x seconds. In a normal world, he should have at worst a loading information or simply that the task is done in background.

But should never be stuck.
So in that case, the dev should have take in consideration that the communication with the server is not instant and have handle this. Not just freezing the player.

 

for me, it has become unbareable near market 06 to leave my ship, the action just never completes ... i need to open a ticket about it, because this is not playable anymore. just yesterday i found abandoned core with some stuff on it, i landed and exited my ship .. but then all my controls froze, while screen was still going, i saw people use chat and then i saw another player landing and just taking everything right under my nose and i was just sitting there and couldn't even press ESC or ALT+F4, nothing worked, while game was running fine.

so i had to minimize, which produces 1fps desktop, then spend like 3-4 minutes trying to open process explorer so i can kill dual.exe.

after that, i logged back in and i was standing in my ship outside the seat, i was instantly able to enter my ship and fly away.

 

another HUGE issue i see, is that animations seem to be server sided aswell, because when you're near market 06, you often see animations lag behind by a good 1 minute or more, i put thrust to 100%, but there are no sound nor visual effects, then after i get like 10km away from market, then engine sounds turn on and i can see visual effects on engines. like what the actual heck? i have NEVER seen any other game make sounds and visuals server sided 😕

 

5 hours ago, Zeddrick said:

As it happens I don't think DU has a massive problem with lower level optimisations.  My PC has a 2070 in it (I think) which is not bad but also not bleeding edge and it rarely gets up to above 50% utilisation even when the game is laggy and frame dropping.  Likewise I have about 10 gen10 intel vcores which don't often get thrashed very hard either.  What I see is memory use crawls up over time and CPU use/lag goes up to match, which is a sign that the game has memory leaks.  Also when there are a lot of network events (I go to market 6 and lots of things load) then it can get really laggy.  That's a sign of bad server-side or network code usually and the game does thrash the network sometimes. 

For me the game started to get worse in performance terms when there was a project to move from a dynamo-db based storage tier to a hand-crafted backend.  At that point I started getting lag, poor framerates and problems loading constructs in various places (flying over the ODY base in beta at a 3k height, for example).  I don't find this too surprising really as the decision to spend an 'innovation token' re-inventing distributed operational store tech rather than using one someone else already made seems to have been motivated by lowering operational costs rather than making things better.  Generally you get what you pay for with this sort of thing.

 

what program and how do you monitor your GPU usage? i don't believe this for a second that 2070 can run DU with only 50% GPU use, no way. 2070 is worse than 1080 and mine is constantly under 100% use, even when looking in the sky or ground. altho i don't remember it being that bad in early access, so are you talking early access?

 

also i have never seen memory leaks. i have 16gb 3400mhz DDR4 and when i play DU, it's usually 50-80% full, never seen it creep over 80% and i have some background stuff running too, like firefox, fb messenger, steam, winamp, msi afterburner ..etc. i think the memory leak is specific to some PCs, because i have seen people talk about it in the game chat too, but i never had memory leak issues myself.

could also be that the more RAM you have, the more game uses, because windows is designed this way, if you have more RAM, it uses more before clearing older stuff out of RAM. so it could be that if you have over 16gb, then DU won't function properly anymore. just a speculation tho.

 

that last bit "generally you get what you pay for" is no longer true, it used to be true up to early 2000s maybe, but after that came the era of monetization and vertical videos, where optimization and quality is no longer important. EVERYTHING nowdays is made to bare minimum sellable level. i remember GTA 5 launch on PC, i was waiting on teamspeak with like 10 friends to stream multiplayer .. and multiplayer didn't even work for first 2-3 days LOL. and GTA 5 made like 1 BILLION in just about 3 days when it was released on consoles... i think that's world record holder for highest profit game ever on launch .. and even up to this day, gta online is total garbage, their net-code is complete mess. there are 3rd party  modders who have made it work far better for FREE and BILLION dollar company can't bother fixing any of it.

so paying is not the issue.

 

i have also seen some really small indie companies that consist of just few devs and they do better work, one of them was DRONE the game, but unfortunetely they have abandoned the project i think. haven't heard anything from them for about 2 years by now. they made amazing concept, game has quite nice graphics, drone building and basic multiplayer and also level editor concept, it ran like a dream and it was just 5 developers. i still have 2 of them in my steam friends, altho they have gone inactive, so maybe they sold out or something.

but my point here, is that their entire project made maybe like 300k dollars total over 3-4 years, which is very small amount because they had to pay for licenses and servers and to get game on steam ..etc, yet the initial game concept ran like a dream without any bugs. they had right approach, but unfortunetely it didn't get popular. if i remember correct, one of the devs told me that their idea of making a game is same as what i described here -- first optimize what you got, then add more content.

Link to comment
Share on other sites

10 hours ago, Sethioz said:

that last bit "generally you get what you pay for" is no longer true, it used to be true up to early 2000s maybe, but after that came the era of monetization and vertical videos, where optimization and quality is no longer important. EVERYTHING nowdays is made to bare minimum sellable level. i remember GTA 5 launch on PC, i was waiting on teamspeak with like 10 friends to stream multiplayer .. and multiplayer didn't even work for first 2-3 days LOL. and GTA 5 made like 1 BILLION in just about 3 days when it was released on consoles... i think that's world record holder for highest profit game ever on launch .. and even up to this day, gta online is total garbage, their net-code is complete mess. there are 3rd party  modders who have made it work far better for FREE and BILLION dollar company can't bother fixing any of it.

so paying is not the issue.

 

I really can't agree with this idea. 

 

First, GTA5 is not merely one of the most successful games. With over 6 billion in revenue, it's among the most successful entertainment products in history, period. By comparison, every "Harry Potter" movie combined is estimated to have made ~$7.7 billion. 

 

So...it's clearly an outlier not only in video games, but in the massive world of entertainment media in general.

 

There's a lot more examples of games struggling to sell because of bad performance than examples of games that were successful despite bad performance. Let's not give into GTA5's survivor bias. 

 

Games like Cyberpunk and NMS infamously had rocky releases because of performance struggles -- it's silly to suggest that studios simply "don't care" and only want to release an "MVP" because that's what makes money. 

 

What actually makes money is releasing a good game that works as it should across every platform, even in the world of Early Access. 

 

Let's take the example of an atypically successful game's PC multiplayer port not working well with a grain of salt...I personally do not see evidence that the industry in general only wants to push MVPs because "that's what makes money". 

 

I also can't agree that "paying isn't the issue" as if money has no relation to what it takes to build a successful game...especially in the context of an MMO.

 

The performance of servers is absolutely proportional to NQ's budget, especially running on AWS. The quality of dev you can afford is 100% driven by budget. NQ's CTO is someone who's previous job was literally as an intern, with NQ being their only game dev role. There's a big difference between someone like that and someone that's spent 30 years making MMOs leading the tech. 

 

Further, most modern MMOs secure at least $200 million in funding -- NQ is trying to do this with 10% of the funding of a typical "theme park" MMO. It's silly to suggest that hasn't hurt DU's chances. 

 

This example of a tiny, discontinued game made by a few people is a moot point and greatly misunderstands the real struggles with optimization and performance between complex titles like DU and simple, incomplete titles made by indies...as someone that spent several years working on an indie title that was also on Steam for a period, I can't fathom the comparison between a not-finished indie game and a released MMO like DU. 

 

Even GTA5 isn't a fair comparison -- there's a world of difference in complexity between 3rd party instanced servers and a true persistent MMO. Granted GTA5's client side optimization and graphics is leagues beyond DU's...but my point is that none of this is so simple. 

Link to comment
Share on other sites

26 minutes ago, blundertwink said:

 

First, GTA5 is not merely one of the most successful games. With over 6 billion in revenue, it's among the most successful entertainment products in history, period.

 

'tis a shite game.

Link to comment
Share on other sites

16 hours ago, blundertwink said:

 

I really can't agree with this idea. 

 

First, GTA5 is not merely one of the most successful games. With over 6 billion in revenue, it's among the most successful entertainment products in history, period. By comparison, every "Harry Potter" movie combined is estimated to have made ~$7.7 billion. 

 

So...it's clearly an outlier not only in video games, but in the massive world of entertainment media in general.

 

There's a lot more examples of games struggling to sell because of bad performance than examples of games that were successful despite bad performance. Let's not give into GTA5's survivor bias. 

 

Games like Cyberpunk and NMS infamously had rocky releases because of performance struggles -- it's silly to suggest that studios simply "don't care" and only want to release an "MVP" because that's what makes money. 

 

What actually makes money is releasing a good game that works as it should across every platform, even in the world of Early Access. 

 

Let's take the example of an atypically successful game's PC multiplayer port not working well with a grain of salt...I personally do not see evidence that the industry in general only wants to push MVPs because "that's what makes money". 

 

I also can't agree that "paying isn't the issue" as if money has no relation to what it takes to build a successful game...especially in the context of an MMO.

 

The performance of servers is absolutely proportional to NQ's budget, especially running on AWS. The quality of dev you can afford is 100% driven by budget. NQ's CTO is someone who's previous job was literally as an intern, with NQ being their only game dev role. There's a big difference between someone like that and someone that's spent 30 years making MMOs leading the tech. 

 

Further, most modern MMOs secure at least $200 million in funding -- NQ is trying to do this with 10% of the funding of a typical "theme park" MMO. It's silly to suggest that hasn't hurt DU's chances. 

 

This example of a tiny, discontinued game made by a few people is a moot point and greatly misunderstands the real struggles with optimization and performance between complex titles like DU and simple, incomplete titles made by indies...as someone that spent several years working on an indie title that was also on Steam for a period, I can't fathom the comparison between a not-finished indie game and a released MMO like DU. 

 

Even GTA5 isn't a fair comparison -- there's a world of difference in complexity between 3rd party instanced servers and a true persistent MMO. Granted GTA5's client side optimization and graphics is leagues beyond DU's...but my point is that none of this is so simple. 

 

my point about gta5, is that their netcode is complete garbage, this has been proven and shown by MANY programmers and youtubers, they show how easy it would be to fix their code, but they don't care, because they made their BILLIONS from it.

my point is that just because something is expensive, doesn't make it good.

 

another good example is Ferrari and Lamborghini. I've had chance to work on both. one friend/landlord i knew, bought broken ferrari as a small side project and i was working on it with him. it caught fire and back end had some fire damage. so when we were taking it apart ... i was literally like --- FACEPALM 1000x.

i instantly noticed WHY it catches fire so easily, it's not just this case, look it up. lot of ferraries catch fire for no reason at all, even from normal driving. there are lots of youtube videos on them too. so the problem is that they those donuts have made fuel lines out of RUBBER and they run basically on top of engine's most hottest areas.

you know what happens to rubber when it heats up and cools down all the time? it starts to crack and loses its rubbery properties. this happens a lot in countries that have colder weather, so with colder weather, it heats up and when you stop, it rapidly cools down.

once it cracks and leaks, fuel drips on very hot parts and will catch fire, which easily melts rest of the rubber fuel line and there you go .. up in flames it goes.

Lamborghinies are about same, they're designed so badly .. they only got famous back in the day for their unique designs and high performance and now everyone thinks that ferrari and lamborghini are good, when really they'er just expensive junk.

if i had money to buy one, i'd rather get me a 69' charger and swap out some parts for modern ones. always liked dodge m4s too.

 

cyberpunk and NMS sold well actually, but NMS was scam. reason why it got lot of pre-orders, is because they LIED. i stepped into that bucket too. they showed AMAZING graphics and scenes, but out came total trash with horrible graphics and no multiplayer. it's known as "one man's lie". lot of people quoted him on many occasions, where they asked "does it have multiplayer, can you play together" and he was like "ummmmm ..... yes". you could clearly tell he lied, he was just thinking what would make him more money, lying on the spot to earn more pre-orders or saying "no" and risk losing many sales, so he decided to lie.

there were also videos that showed comparison of pre-release footage to actual game footage, but NMS publishers quickly copyrighted them all, you can still find them off-youtube.

 

cyberpunk got lot of sales because of Keanu Reeves being in the game and advertising it and also because they advertised nudity in the pre-release footage, which in fact was a very small part of the game. if i remember right, just one mission had a nude girl in it, otherwise you couldn't even take clothes off of your own character. that was their sales trick to pull in lot of teenagers.

cyberpunk wasn't too bad tho, i still have it installed. it's fun to drive around sometimes and smash stuff, but it gets boring quite fast. it does have huge and detailed map, but still lacks the actual gameplay part. story was also very disappointing.

so that was just a sales trick to pull in people, altho i wouldn't say it had bad performance, it ran rather ok, seen much worse.

 

 

i'm not saying anything is simple, the points i was making, is that expensive doesn't always mean it's good. it USED to be true up to until early 2000s, where only reason why products (digital and real) costed a lot, is because their creators actually put lot of money and effort into making them, so they had to sell it at a higher price as it was WORTH IT!

 

but nowdays, people sell stuff expensive only because they are greedy, they put minimal effort into things and sell as high as possible. .. and as long as they get away with low quality -- high price, they keep on doing it. why would they want to work hard, if they can do only half the work, but get 10 times the money?

 

I say that BEST games are made by those who are passionate about their game, those who enjoy playing their own game and are proud to have made something like that. those who see money only as a positive side effect and develop the game because they want to see it work! as soon as devs lose interest in playing their own game -- game over!

without naming the game, i spent over 7000h in one grindy game, it started out GREAT, devs used to play the game, they replied to comments and reviews on steam ..etc, but after about a year, they lost interest, they never played the game themselves, so everything fell out of balance, they started making changes only based on log files and what bad players complained about, like if some bad players kept dying to certain weapon (because they were BAD and didn't know how to counter it), then devs just nerfed the weapon, even tho there was nothing wrong with it. that ruined the game completely.

 

right now in DU, optimization is biggest issue, so they gotta fix this. i honestly can't see anything else wrong with it, sure we could have more stuff, like wheeled vehicles and planetary combat, but it's good as it is. i'd take DU optimization over DU content at any time. would love to play solid 60fps and able to leave my vehicle instantly, without getting huge lag spikes or getting stuck.

Link to comment
Share on other sites

On 1/26/2023 at 4:29 AM, Sethioz said:

 

what program and how do you monitor your GPU usage? i don't believe this for a second that 2070 can run DU with only 50% GPU use, no way. 2070 is worse than 1080 and mine is constantly under 100% use, even when looking in the sky or ground. altho i don't remember it being that bad in early access, so are you talking early access?

 

also i have never seen memory leaks. i have 16gb 3400mhz DDR4 and when i play DU, it's usually 50-80% full, never seen it creep over 80% and i have some background stuff running too, like firefox, fb messenger, steam, winamp, msi afterburner ..etc. i think the memory leak is specific to some PCs, because i have seen people talk about it in the game chat too, but i never had memory leak issues myself.

could also be that the more RAM you have, the more game uses, because windows is designed this way, if you have more RAM, it uses more before clearing older stuff out of RAM. so it could be that if you have over 16gb, then DU won't function properly anymore. just a speculation tho.

 

 

I'm just monitoring using the standard tools that come with windows.  It turns out my card is a 2060 and not a 2070.  It might have been tuned a little with MSI Afterburner a year ago to optimise for low power etherium mining, I can't remember.  I just went over to the D6 market and ran around for a while.  I didn't get any lag, but the GPU didn't go over 75% the whole time (it does seem to use more than it used to) and sat at 60% for most of it.  The CPUs were in the high 80s most of the time.

What was more interesting is when I stopped moving and just stood still the CPU use went down to around 35% but the GPU use crept up to 100% (just rendering almost the same image over and over).  As soon as I start moving the GPU utilisation drops off again, which is a sign that the CPUs are the bottleneck and can only drive the GPU at full speed when they don't have many calculations to do in order to render the scene.

I have seen people complaining before somewhere else about DU performance and saying they were using a 1080.  In terms of 'better' and 'worse', that's a very subjective thing.  a 1080 is an old card now (they're in the 4000s now I think?) and in my experience with programming for NVIDIA cards (admittedly doing business software not games) is that the generation does matter with these things.  Usually the higher generations have faster memory, better handling of memory caches, memory read combines, etc and, most importantly, better and faster access to CPU main memory.  So if you have a 1080 and you're running some optimised piece of software (a benchmark, for example) then the 1080 is going to really shine because it has a lot of SMs.  But if the code is less optimised then the newer generations of chipset can add more value by accelerating the less well optimised operations.  It might very well be that the 1080 is better at running well optimised code but a 2060 is better at running bad GPU code?

Memory leaks do happen for me (with 32G of RAM).  I've seen the memory used by DU creep up over time (I usually have about 10G free but when DU creeps up it can take it all) and a lot of others report the same.  In particular I noticed it when mission running, I would have to re-log after the 5 hour run through the PvP zone, even though the machine had done very little during that time, or I get bad framerates when trying to land.  Also after PvP battles I sometimes need to re-log to get the framerate back to where it should be.

Link to comment
Share on other sites

On 1/31/2023 at 5:01 AM, Zeddrick said:

I'm just monitoring using the standard tools that come with windows.  It turns out my card is a 2060 and not a 2070.  It might have been tuned a little with MSI Afterburner a year ago to optimise for low power etherium mining, I can't remember.  I just went over to the D6 market and ran around for a while.  I didn't get any lag, but the GPU didn't go over 75% the whole time (it does seem to use more than it used to) and sat at 60% for most of it.  The CPUs were in the high 80s most of the time.

What was more interesting is when I stopped moving and just stood still the CPU use went down to around 35% but the GPU use crept up to 100% (just rendering almost the same image over and over).  As soon as I start moving the GPU utilisation drops off again, which is a sign that the CPUs are the bottleneck and can only drive the GPU at full speed when they don't have many calculations to do in order to render the scene.

I have seen people complaining before somewhere else about DU performance and saying they were using a 1080.  In terms of 'better' and 'worse', that's a very subjective thing.  a 1080 is an old card now (they're in the 4000s now I think?) and in my experience with programming for NVIDIA cards (admittedly doing business software not games) is that the generation does matter with these things.  Usually the higher generations have faster memory, better handling of memory caches, memory read combines, etc and, most importantly, better and faster access to CPU main memory.  So if you have a 1080 and you're running some optimised piece of software (a benchmark, for example) then the 1080 is going to really shine because it has a lot of SMs.  But if the code is less optimised then the newer generations of chipset can add more value by accelerating the less well optimised operations.  It might very well be that the 1080 is better at running well optimised code but a 2060 is better at running bad GPU code?

Memory leaks do happen for me (with 32G of RAM).  I've seen the memory used by DU creep up over time (I usually have about 10G free but when DU creeps up it can take it all) and a lot of others report the same.  In particular I noticed it when mission running, I would have to re-log after the 5 hour run through the PvP zone, even though the machine had done very little during that time, or I get bad framerates when trying to land.  Also after PvP battles I sometimes need to re-log to get the framerate back to where it should be.

 

when i'm near 06, my 1080 doesn't go 100% use either, i didn't notice it before, but i looked at it while ago and it was sitting around 50% or so. CPU use wasn't too high either, so in my case, nothing gets maxed out at 06, it's just laggy due bad optimization.

 

i can see CPU getting more use when you're using more powerful GPU, because regardless what GPU you have, CPU still has to "feed" your GPU the information that it needs to calculate, therefore with bad CPU, your GPU will never reach its full potential (unless mining or doing raw calculations on GPU).

 

as about RAM, i think it really is issue with higher RAM amounts. you said you have 10gb free, so 22gb used, that's more than i even have. i have 16gb and i never see it over 80%, but that's normal, because as i mentioned before, windows is built in a way that it consumes more RAM if you have more to spare.

 

there's no such thing as running good or bad code better, bad code runs bad and good runs good. reason why 2060 gives better performance, is due the architecture probably. which again points to bad / lazy developing. devs never bothered to test on multiple architectures, so even tho 2060 is a lot weaker than 1080, it might actually run DU better as DU devs only wrote it for newer architecture.

 

also don't forget DLSS. i don't know if you use it, but dlss was designed to use less computing power with same visuals. it renders game at lower resolution, which means you need a lot less GPU power to render it. for example it can render game in like 480p and then upscales it to 1080p with same visual quality as if you'd render it in 1080p. dlss uses some A.I algorithm that fills in missing pixels, it can look bit blurry in some areas, but generally it's quite good.

 

this is another sign that devs didn't properly test nor care, because there's also AMD fidelity upscaler that does similar thing, only that it works on every GPU, but they never bothered to include it. and ofcourse as i mentioned before, they made it in a way that its not possible to run DU on HDD.

 

way i see DU's bad optimization, is that if some country has silk smooth roads and then they make cars that can only drive on those roads, but if you drive over any rougher area, then your car starts shaking, speed drops and it consumes 10x more fuel, because devs only tested it in their own country and don't care that it can't drive on any other roads.

 

this goes bit offtopic, but we have same problem with websites ... they make websites for specific browsers, it's like racism coming back all over. in order to filter out web browsers, they have to add specific code for each browser, which is why nowday websites take 10x more space and are 10x slower to load, because they need to load 10x more code than a simple non-racist website that works exactly same in every browser.

i've been working on websites since i was like 14 or so, i know how optimized and fast websites used to be, compared to this horror they call "modern" nowdays.

HOWEVER with games its other way around, you need to make sure it DOES run on different hardware, which can be more time taking indeed, but that's what optimization means, you make sure your game runs smooth on older and newer hardware.

 

i'd like to see DU on UE5, it's exactly the type of game that should be on UE5. it has lot of repeat-textures that are rendered seperate ..

 

there's just NO EXCUSE to have GPU under 100% use when looking at floor that consists of same voxels, it should use less than 5% of GPU to render non-moving texture. i only get 60fps when looking at sky and even then sometimes it creeps up to 100% use, so there's something very wrong with how game renders textures and that's what i'm complaining about, they need to fix this. looking at wall or ground with nothing on them, should produce 60fps even on some old GTX 670 or even 660. specially cuz those cards were released around the time they started developing the game.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...