devtrackers.gg
7 months
ago -
Coffee Stain Studios
-
Direct link
Transcript (by Youtube)
| 5s | hey it's been a while remember me cuz I |
|---|---|
| 8s | barely do uh anyways welcome back to yet |
| 10s | another video it's been a while I missed |
| 12s | you you know it's it's a long time since |
| 15s | we since we got to hang out like this |
| 17s | you know so uh uh you know I constantly |
| 20s | get the question when we're going to |
| 21s | release optimization updates and the |
| 23s | truth is that almost every single patch |
| 25s | we make contains some form of |
| 27s | optimization it's sort of a never-ending |
| 29s | battle but some changing have a bigger |
| 31s | impact than others and today I'd like to |
| 33s | share some of the work that's gone into |
| 34s | improving the performance for the game |
| 37s | coming in |
| 39s | 1.0 before we dive into the topic at |
| 41s | hand there are two things that I want to |
| 43s | address first thing is that you may have |
| 45s | heard news that our parent company |
| 46s | embracer has sort of released a press |
| 48s | release where they're saying they're |
| 50s | planning to like transform into like |
| 52s | sort of three Standalone publicly listed |
| 54s | entities of which uh coffee stain and |
| 57s | Friends uh working name uh uh the parent |
| 61s | company for coffee s Studios is one of |
| 63s | those uh I'm not going to go into detail |
| 65s | about like this press release honestly |
| 67s | because it's a little bit above above my |
| 69s | pay grade uh you can find a link into |
| 71s | the description if you want to read more |
| 72s | on the on the press release uh if you're |
| 75s | interested in that kind of stuff uh but |
| 77s | there's been a bunch of speculation as |
| 78s | to how this affects satisfactory and I |
| 80s | just wanted to sort of come on here and |
| 81s | sort of set the record straight on these |
| 83s | things so so these changes to the |
| 85s | company structure will not affect our |
| 87s | plans of satisfactory at all we're still |
| 88s | going to release the game this year year |
| 91s | 2024 it hasn't changed the scope for 1.0 |
| 94s | at all and we also plan to keep working |
| 96s | on satisfactory after 1.0 is out whether |
| 99s | that's in the form of DLC or you know |
| 101s | free updates to the game is yet to be |
| 102s | decided uh it might be either it might |
| 104s | be both actually so so we'll sort of see |
| 106s | but development for satisfactory is |
| 108s | still going strong and actually speaking |
| 110s | of which uh that's sort of the second |
| 112s | thing I wanted to address here uh and |
| 114s | it's sort of you know give you a little |
| 115s | bit of a state of development because |
| 116s | it's been a while since we published a |
| 118s | video here on YouTube channel so um yeah |
| 120s | we've been very busy behind the scenes |
| 122s | managing the Clos beta for the select |
| 124s | few that managed to get picked around a |
| 126s | thousand is people have been playing the |
| 128s | Clos beta now for a couple of weeks and |
| 130s | the general feedback that we've gotten |
| 132s | has been pretty positive actually and uh |
| 134s | it's been a it's been really refreshing |
| 136s | to hear all the nice things that people |
| 138s | have said there's obviously a few |
| 139s | technical things we all already knew |
| 141s | that and there's going to be more of |
| 142s | that down the line but so far |
| 145s | everything's looking peachy actually so |
| 146s | with this last bit of feedback we're now |
| 148s | gearing up for the final push and you |
| 151s | know hopefully it won't be long until we |
| 152s | can maybe announce sort of a release de |
| 154s | or something like that who knows |
| 156s | exciting times anyways with that out of |
| 159s | the way let's get to the main course and |
| 161s | talk about |
| 162s | optimization my favorite thing one big |
| 164s | disclaimer is that performance is sort |
| 166s | of tied directly to your hardware and |
| 168s | configurations so some of these changes |
| 170s | may be significantly more notable for |
| 172s | some people rather than others so yeah |
| 175s | just bear that in mind that performance |
| 177s | is very varies a lot for from person to |
| 180s | person uh also worth noting is that I'll |
| 183s | put uh as much of this into layman's |
| 185s | terms but there might be some technical |
| 187s | jargon going on here and there as we go |
| 189s | into details so just bear with me let's |
| 191s | start talking about foliage my favorite |
| 194s | subject everybody loves foliage so right |
| 197s | now there are colliders for the vehicles |
| 199s | which allow them to sort of detect |
| 201s | contact with with any foliage and sort |
| 203s | of destroy it as you're driving this |
| 205s | results in pretty huge performances as |
| 208s | like when you're driving through |
| 209s | especially f dense areas but also in |
| 211s | places where foliage is nearby but not |
| 213s | necessarily in your path thanks to like |
| 215s | wonky hitbox detection that we have uh |
| 218s | in 1.0 most folage colliders are going |
| 220s | to be completely removed from Vehicles |
| 222s | actually uh which means that going |
| 223s | forwards you can't destroy smaller |
| 225s | foliage by driving it will simply just |
| 227s | persist in the world uh which honestly |
| 230s | also looks a bit better than when it |
| 232s | just like vanishes into thin air when |
| 233s | like the first pixel of your vehicle |
| 235s | hits it so we we thought at first this |
| 238s | was going to be a strange change but |
| 239s | actually testing it out we felt like |
| 241s | this kind of actually makes sense most |
| 243s | noteworthy for dedicated servers folage |
| 245s | has been like a huge culprit of Ram uh |
| 247s | where sometimes it reached over like 12 |
| 249s | gigs on like just a fresh safe file so |
| 252s | not having Collision on foliage will |
| 254s | reduce it down to like somewhere around |
| 256s | like 5 and 1/2 gigs which is a huge |
| 258s | Improvement relaxing folage Collision |
| 260s | will also help greatly with level |
| 261s | streaming and reduce overall stuttering |
| 264s | because uh you're not trying to like |
| 266s | demolish like dozens of objects within a |
| 268s | second many times in a row all the time |
| 271s | constantly so that's great but Vehicles |
| 273s | will still be able to destroy almost all |
| 275s | folage that requires at least the |
| 277s | chainsaw to get rid of so like it's not |
| 279s | gone totally but it's gone for like the |
| 281s | majority of like the smaller folage |
| 283s | types where the fine line goes between |
| 284s | like exactly what folage is destructible |
| 286s | or not is something that we're still |
| 288s | sort of like figuring out and fine |
| 289s | tuning so if you ever use belts then |
| 291s | you'll be happy to know that we've |
| 293s | optimized those significantly belts and |
| 295s | pipes currently exist within like sort |
| 297s | of unique instance groups in updated |
| 298s | meaning that like each length of belts |
| 301s | between connection points was sort of |
| 302s | managed by themselves but in 1.0 they're |
| 304s | going to be pulled together instead this |
| 306s | is a big computational relief for your |
| 308s | CPU as it won't have to work through as |
| 311s | many instances of belts and pipes as it |
| 313s | currently does so now instead belts and |
| 315s | pipes uh belong to like fewer but bigger |
| 317s | pools and now instead we're relying on |
| 319s | the GPU to do the heavy lifting gpus can |
| 322s | work extremely fast even when |
| 324s | computations are chunky in size which is |
| 326s | why for settled for this like fewer but |
| 328s | bigger approach this helps stuff like |
| 330s | Shadow rendering a lot since shadows |
| 331s | will also belong to massive pools being |
| 333s | calculated and drawn rather than like |
| 335s | many individual objects or like smaller |
| 338s | groups it's also great for instances |
| 340s | where many small belts occur frequently |
| 342s | which tends to happen a lot a prime |
| 344s | example of this is manifolds where you |
| 346s | know belts currently have to be computed |
| 349s | per belt no matter how small they are uh |
| 352s | which they tend to be in manifolds uh |
| 354s | after this change again they're pulled |
| 356s | together and the DPO will just like show |
| 359s | it all through one thing that a lot of |
| 360s | people ask us about is is whether we're |
| 362s | going to use nanite for conveyor belts |
| 364s | and we did play around a bit with nanite |
| 366s | splines to try and like optimize the |
| 368s | belt but in the end we sort of ended up |
| 370s | settling on our own instancing solution |
| 371s | because we found that it worked better |
| 373s | for our game speaking of objects that |
| 374s | may occur frequently everywhere we've |
| 376s | now introduced a system that sort of |
| 377s | converts several architectural |
| 379s | buildables that players don't interact |
| 381s | with uh or you know don't produce |
| 383s | anything into Data only objects uh |
| 386s | currently these buildables are what's |
| 387s | referred to as heavy actors in a real |
| 390s | like buildings that have like logic and |
| 393s | actually do stuff uh the most common |
| 395s | examples going data only would be you |
| 397s | know foundations walls or architectural |
| 399s | buildables like railings or stairs and |
| 401s | such what changing the fancy term means |
| 404s | is that we're basically cutting a lot of |
| 406s | fat from a massive amount of stuff |
| 407s | present in almost all factories so now |
| 410s | we only store The crucial data for these |
| 412s | things like locations the mesh that's |
| 414s | using customization data such as like |
| 416s | the colors and the recipe that we built |
| 418s | with and the cool thing with all this |
| 419s | like light data stored in a pre-prepared |
| 421s | location in memory is that writing it to |
| 424s | save data will be much faster so |
| 426s | hopefully this will reduce the time the |
| 428s | game stalls uh when saving when you have |
| 430s | like a big old factory we're also |
| 432s | skipping a bunch of steps when Gathering |
| 434s | save data for these lightwe buildables |
| 436s | uh related to production which also |
| 438s | saves a bunch of uh time on that note uh |
| 440s | lad times could also be a lot faster |
| 442s | than 1.0 and and actually quite |
| 444s | substantially faster uh because when all |
| 447s | objects were heavy actors spawning them |
| 449s | was costly because because we had a |
| 450s | bunch of code that was like running |
| 451s | through everything uh loading it all no |
| 454s | matter if it was a foundation piece or |
| 455s | if it was an assembler being spawned so |
| 457s | having nonproduction build be |
| 459s | lightweight allows us to skip a bunch of |
| 461s | this code and you know set your files up |
| 463s | faster and the actual safe files will |
| 465s | also be a lot smaller non-production |
| 466s | buildables being lightweight |
| 469s | lightweight uh by the way I just |
| 471s | realized I probably haven't pronounced |
| 472s | lightweight properly but it's too late |
| 474s | we're too deep into we got to keep going |
| 476s | non- production buildable being |
| 477s | lightweight also helps decrease runtime |
| 479s | memory consumption you know actors are |
| 482s | huge in unreal and with so many heavy |
| 484s | actors turned lightweight uh enjoy not |
| 487s | having like every inch of your RAM being |
| 489s | you know used up by satisfactory this |
| 490s | also connects to multiplayer because |
| 492s | when actors are replicated to clients in |
| 494s | Unreal Engine it requires that those |
| 496s | clients allocate memory to track the |
| 498s | state of the actors lightweight |
| 499s | buildables means that this extra data is |
| 501s | no longer needed so yeah tldr |
| 503s | lightweight system makes stuff that |
| 505s | don't have logic a lot smaller and |
| 508s | easier to load and good and on the |
| 511s | client side when playing multiplayer |
| 513s | joining a game is also much faster now |
| 515s | because processing those lightweights is |
| 516s | so much faster than plowing through all |
| 518s | these like heavy actors that existed |
| 520s | before there's also a ton of General |
| 522s | tweaks here and there that by themselves |
| 524s | don't really amount to much but |
| 525s | altogether add up stuff like reducing |
| 528s | unnecessarily large texture sizes uh |
| 530s | sorting stuff that was thrown together |
| 532s | quickly for testing and then forgotten |
| 535s | uh and fixing you know straight up |
| 536s | broken in the world and and this |
| 538s | means that vrm require IR ments are now |
| 540s | a lot lower across the board scalability |
| 543s | settings have also been looked at and |
| 544s | adjusted so that people can expect you |
| 546s | know a more reasonable result from their |
| 547s | settings uh for example the performance |
| 549s | hit going from medium to high in |
| 552s | specifically sh Shadow quality settings |
| 554s | was pretty weird so we've adjusted that |
| 556s | and since the last update we've also |
| 558s | upgraded Onre engine to 5.3 uh this will |
| 561s | give the game a general sort of |
| 562s | performance boost since you know epic |
| 564s | has fixed a bunch of stuff on their end |
| 566s | uh it also resolves a couple of crashes |
| 568s | some people were running into and look |
| 570s | okay I know what you're thinking okay 5 |
| 572s | four just came out are you are you gonna |
| 574s | upgrade to that one uh why aren't you |
| 576s | upgrading uh well okay so there's a |
| 579s | couple reasons why we're not going to |
| 581s | upgrade to 5.4 uh it just isn't viable |
| 583s | for 1.0 uh we might do it in the future |
| 586s | but this is the version we're going to |
| 588s | ship 1.0 with and for all you upscaler |
| 590s | enjoyers out there you'll be happy to |
| 592s | learn that we've updated dlss and FSR to |
| 595s | their respective latest versions so if |
| 597s | you're using those upscaling options you |
| 600s | should also see a bit of an improvement |
| 602s | there and finally in the last video we |
| 604s | talked about new features coming to |
| 606s | Dedicated servers well in the land of |
| 608s | optimization dedicated service I've also |
| 610s | seen some pretty massive improvements |
| 612s | thanks to stuff like the foliage Collis |
| 614s | thing and the pooling of belts and pipes |
| 616s | however we're also experimenting with |
| 618s | dedicated service to enable streaming up |
| 620s | until now the way dedicated service has |
| 622s | worked is that like it loads everything |
| 624s | in the game so it loads the entire map |
| 626s | and everything on it um and by enabling |
| 629s | streaming for dedicated servers this |
| 630s | means that everything on your server |
| 632s | doesn't have to be loaded all the time |
| 633s | so if you have like a bunch of people |
| 635s | hanging out in like the same area uh you |
| 637s | know loading that space should be enough |
| 639s | until you start using the map more which |
| 642s | obviously will sort of require much less |
| 644s | memory when you do start exploring and |
| 646s | building far away from each other uh |
| 648s | then the stress of the server will sort |
| 649s | of scale appropriately instead so those |
| 651s | are some of the improvements coming in |
| 653s | satisfactory 1.0 I'm sure there's going |
| 654s | to be even further optimizations between |
| 656s | now and final release but this is where |
| 658s | we are right right now you may have |
| 660s | noticed that there was a bit of a gap |
| 662s | between this video and the last one you |
| 663s | know there's like I said before there's |
| 665s | a ton of stuff going on my scenes uh it |
| 667s | just makes it really hard to push out |
| 668s | these videos continuously right now uh |
| 670s | but we will be back on track soon enough |
| 672s | so please bear with us and until then I |
| 675s | hope you have a lovely day stay |
| 677s | optimized and take care yay optimization |
| 681s | hooray |
| 690s | oh I forgot to put on the beanie oh |
| 692s | no did I record the entire video without |
| 694s | the beanie oh no oh will people even |
| 698s | recognize me oh |
| 700s | man a bummer |