Announcement

Collapse
No announcement yet.

Unreal Engine 4.23 Preview

Collapse
This topic is closed.
X
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    Originally posted by rdtg View Post
    Is anyone else experiencing an issue/memory leak after importing a large amount of 4K textures (say like 500mb+ in total) and then saving?
    Yep, since the first preview.
    George Rolfe.
    Technical Coordinator at Orbit Solutions Pty Ltd.

    Comment


      TAA with late update in SteamVR is still broken after that last fix (UE-74991). That fix improved the game thread latency, which maybe makes the bad velocity vectors from late update be less severe, but the problem remains.

      I just submitted a bug report, and made this as a test project to demonstrate the issue:

      https://drive.google.com/open?id=1O0...rbfhNpnJrHWFOq

      You can use the motion controller triggers to toggle late update on and off and see the problem.

      Comment


        Originally posted by decksounds13 View Post
        @Daddrendo I hear you on the marketing thing... I don't remember them having conference-level presentations on Light Propagation Volumes, because it is considered experimental so that's a fair critique.

        I personally jumped on the UE4 bandwagon a year ago because of raytracing but It is basically unusable unless you do simple interior/archviz type work. Forget it for out-of-doors because it can't be combined with ANY reasonable GI technology and the volumetric fog values when seen in raytraced reflections are too inaccurate (more fog in reflection than what nature would dictate for given distance.) not to mention the screwy translucency.I would be thrilled if we could even just have raytraced shadows with some other form of realtime GI. I have my doubts that RTGI will ever mature within the 2000 series timeline and will most likely be replaced by more efficient Nvidia tech like DDGI by that time. I personally think they should ditch RTGI for DDGI, especially if optimizations are the problem because DDGI only costs 1 ms at 4k and has NO NOISE. 1 ms!!! See link and video below.

        https://devblogs.nvidia.com/rtx-glob...nation-part-i/

        https://www.youtube.com/watch?time_c...&v=DOt28ooFMR4
        I also agree with you that RT Shadow and RT AO are already very good, both in terms of effectiveness and performance. RT reflection is also good, but the effect is not so complete. But RTGI is really terrible. At present, RT GI can only be used to capture and produce some non-real-time content. If a better performance of RTGI can be used for mixing effects, it is very desirable as a whole.

        LPV and SSGI actually don't look so good. Maybe the future direction will be better RTGI or other types of GI technology.
        Last edited by Leo Rakes; 08-02-2019, 08:27 PM.

        Comment


          My company bought expensive hardware to benefit from realtime raytracing, but it seems like a fraud. Since the first announcment almost 1.5 years ago, they marketed the hell out of the technology. It seems they have selling contracts with Nvidia, to promote their much to expensive hardware. By the way Epic Games made billions last year, i really cant understand why they dont increase the development team rapidly, and focus on this long promised feature? Chaos is nice, but why Epic Games is so often switching the focus and many of the cool engine features are not near to stable? For example PixelStreaming, ControlRig, Niagara and so on ? Its really frustrating. Please increase the development resources dramatically, or stop promoting features, that you couldnt deliver in the near future. It is okay, that development needs its time, but this marketing strategy is just a fraud

          Comment


            Originally posted by Lesuge View Post
            My company bought expensive hardware to benefit from realtime raytracing, but it seems like a fraud. Since the first announcment almost 1.5 years ago, they marketed the hell out of the technology. It seems they have selling contracts with Nvidia, to promote their much to expensive hardware. By the way Epic Games made billions last year, i really cant understand why they dont increase the development team rapidly, and focus on this long promised feature? Chaos is nice, but why Epic Games is so often switching the focus and many of the cool engine features are not near to stable? For example PixelStreaming, ControlRig, Niagara and so on ? Its really frustrating. Please increase the development resources dramatically, or stop promoting features, that you couldnt deliver in the near future. It is okay, that development needs its time, but this marketing strategy is just a fraud
            The current iteration of ray tracing is still in the early phase of things, mostly due to hardware restraints. If you want a fully complete version of it, that handles all of the effects, it's likely going to take a few years for the hardware to be powerful enough to handle it. The difference between the current implementation and some super complete version might be a frame time impact of 10 fold; which hardware isn't really ready for yet. Also, Epic has to do a lot of refactoring of the engine to make the rest of it work.

            As for niagara, it's going to take them a while to perfect it because it's a huge undertaking. If you aren't satisfied with their current progress on it, then use cascade...?

            The argument that they need to stick more people on the development is kind of pointless because you really don't want 50 million people working on the same section of the engine. That's a lot of toes to trip over with coding. A real world example that I can think of: Let's say you need to dig a 2x4 meter hole to fix a water main(just happened outside my house the other day). Which are you going to go with, 20 people with shovels or 1 person in a small backhoe? Sure, the 20 people with shovels could probably do it in roughly the same amount of time, but at the risk of them getting in each other's way and if anything, they will constantly be slowed down waiting their turn to scoop. That and you can't really easily fit 20 people in a 2x4 meter hole. The same goes with working on portions of a game engine.

            Comment


              Originally posted by IronicParadox View Post

              The current iteration of ray tracing is still in the early phase of things, mostly due to hardware restraints. If you want a fully complete version of it, that handles all of the effects, it's likely going to take a few years for the hardware to be powerful enough to handle it. The difference between the current implementation and some super complete version might be a frame time impact of 10 fold; which hardware isn't really ready for yet. Also, Epic has to do a lot of refactoring of the engine to make the rest of it work.

              As for niagara, it's going to take them a while to perfect it because it's a huge undertaking. If you aren't satisfied with their current progress on it, then use cascade...?

              The argument that they need to stick more people on the development is kind of pointless because you really don't want 50 million people working on the same section of the engine. That's a lot of toes to trip over with coding. A real world example that I can think of: Let's say you need to dig a 2x4 meter hole to fix a water main(just happened outside my house the other day). Which are you going to go with, 20 people with shovels or 1 person in a small backhoe? Sure, the 20 people with shovels could probably do it in roughly the same amount of time, but at the risk of them getting in each other's way and if anything, they will constantly be slowed down waiting their turn to scoop. That and you can't really easily fit 20 people in a 2x4 meter hole. The same goes with working on portions of a game engine.
              the point is simple: Epic Games makes raytracing and other features seem much better than they really are at the moment. And that's just misleading at a certain level. Nobody would have had a problem with that if they had taken a few more years, or at least made it clear that the technology isn't far enough at the moment. That would be honest and fair. But since you have done exactly the opposite, I think you can also be criticized for that. I think your comparison limps at 2 points:

              1.) you can code with a lot of people on a topic if the management is good and the work is divided into many small sections.

              2.) you can work with a lot of people faster and at the same time on several large projects and complete them. Epic develops new features like ControlRig, AR Implementation, Niagara, PixelSteaming, makes a huge breeze around these features, but then apparently changes focus and doesn't complete any of these features that are actually useful for developers. With a much bigger team, you might be able to make new features usable. And don't always open up new topics, increase people's anticipation extremely, and then sometimes not develop the features further for months or years.

              i'd rather have 3 stable features than 50 experimental /beta features that you can't use in productions for years
              Last edited by Lesuge; 08-03-2019, 06:18 AM.

              Comment


                Originally posted by Leo Rakes View Post
                LPV and SSGI actually don't look so good. Maybe the future direction will be better RTGI or other types of GI technology.
                Yes, because RTGI and SSGI are in the early phase of development. Not mentioned in any release notes yet, because these features aren't ready to use yet...
                The unfinished feature is included in the current release to avoid creating too many engine branches for rendering development.

                Sometimes I got feeling that we should have "unreleased notes" for unfinished features that many people try to use

                Originally posted by Lesuge View Post
                2.) you can work with a lot of people faster and at the same time on several large projects and complete them. Epic develops new features like ControlRig, AR Implementation, Niagara, PixelSteaming, makes a huge breeze around these features, but then apparently changes focus and doesn't complete any of these features that are actually useful for developers. With a much bigger team, you might be able to make new features usable. And don't always open up new topics, increase people's anticipation extremely, and then sometimes not develop the features further for months or years.
                Nope. 9 women can’t give birth to a baby in a month. This is a common misconception about software development. You can't infinitely hire developers and expect faster/better effects.
                The first article I found in Google after searching a phrase above: https://divante.co/blog/9-women-give-birth-baby-month/

                Epic is constantly hiring and you're behaving like giving them 101 lessons on software development.

                Epic didn't lose focus on Niagara. Every release brings a huge improvement and many developers working on it. It's just the entire next generation FX system created from scratch. You can check progress on the first page of this thread, Trello's roadmap, Github's branch for Niagara. Or simply try preview release.

                Yeah, Epic didn't develop ControlRig for a while, but they presented progress lately, definitely working on it.
                https://www.youtube.com/watch?v=TYSdxW2WHcI&t=988s

                -----

                I can't see where Epic over-promoted raytracing. Perhaps "Early access" label wasn't clear for users, now they started tagging such systems as "beta".
                There was only one release of Unreal Engine with raytracing - it required at least 1.5 years of work and large rendering refactors.

                A year ago, before announcing RTX cards, people didn't believe who could actually have realtime raytracing soon. They were like "yeah, maybe in the decade".
                It's already working for companies working on beefy GPUs for visualization purposes. It's huge progress since the last year.
                In case of video games, most probably raytracing will become common with the premiere of new consoles - coming the next holiday.

                It's not a fraud, it's a natural process of developing software.
                Last edited by Doctor Ergot; 08-03-2019, 07:58 AM.

                Comment


                  Originally posted by Doctor Ergot View Post
                  Nope. 9 women can’t give birth to a baby in a month. This is a common misconception about software development. You can't infinitely hire developers and expect faster/better effects.
                  The first article I found in Google after searching a phrase above: https://divante.co/blog/9-women-give-birth-baby-month/

                  Epic is constantly hiring and you're behaving like giving them 101 lessons on software development.
                  Exactly! Too many people, that have no real experience working with large teams on projects, think that throwing more people at a problem will solve it faster. It's not the case. I'm normally an electromechanical engineer, so I don't exactly have experience in working with large game dev teams, but based off my experience with engineering projects, throwing more people at a problem doesn't solve it faster. There's a serious diminishing returns effect to it and that starts to cost a company a lot of money, for a very marginal gain in output.

                  Comment


                    Lesuge There is a misconception on what is being doing on the current engine progress. There were different approaches and "states" for the realtime raytracing, the first time they showed it live to the public was the Star Wars demo, which required an absurd hardware (already obsolete), but it was an internal project and worked with a non-released experimental engine code. I am sure any company wanting to have access to that specific code would have cut a deal with Epic for this. In fact at any certain point in time, any company can have access to specific code Epic show in events, because it is unlikely that code will be made public, because it is not tweaked to all scenarios. Only companies with appropriate staff for dealing with that code would probably have access to it, like a big game studio, otherwise it will be a lot of trouble for Epic to diverge efforts, unless they are paid for it, just to support a version which will be completely different anyway.

                    What Epic is doing right now is making the engine ready for a public release in their regular releasing schedule and that everyone can have access to a universal code targeted for all industries: game, arch-viz and film.

                    At the 1st event, there were no promises on a specific release for it to be completed, but they were talking about its current state at every event they are participating. Is it a fraud? Absolutely no. It is a technology that needs time to mature and be ready for all usage scenarios, since they have different performance expectations.
                    Nilson Lima
                    Technical Director @ Rigel Studios Ltda - twitter: @RigelStudios
                    Join us at Discord: https://discord.gg/uFFSEXY

                    UE4 Marketplace: Cloudscape Seasons
                    supporting: Community FREE Ocean plugin

                    Comment


                      I noticed there are some errors when I'm trying to run Unreal Engine 4.22.3. As I remember I reported this bug. Fix it,please?
                      Last edited by Gamerz31w; 08-05-2019, 07:35 AM.

                      Comment


                        Originally posted by Doctor Ergot View Post

                        Yes, because RTGI and SSGI are in the early phase of development. Not mentioned in any release notes yet, because these features aren't ready to use yet...
                        The unfinished feature is included in the current release to avoid creating too many engine branches for rendering development.

                        Sometimes I got feeling that we should have "unreleased notes" for unfinished features that many people try to use



                        Nope. 9 women can’t give birth to a baby in a month. This is a common misconception about software development. You can't infinitely hire developers and expect faster/better effects.
                        The first article I found in Google after searching a phrase above: https://divante.co/blog/9-women-give-birth-baby-month/

                        Epic is constantly hiring and you're behaving like giving them 101 lessons on software development.

                        Epic didn't lose focus on Niagara. Every release brings a huge improvement and many developers working on it. It's just the entire next generation FX system created from scratch. You can check progress on the first page of this thread, Trello's roadmap, Github's branch for Niagara. Or simply try preview release.

                        Yeah, Epic didn't develop ControlRig for a while, but they presented progress lately, definitely working on it.
                        https://www.youtube.com/watch?v=TYSdxW2WHcI&t=988s

                        -----

                        I can't see where Epic over-promoted raytracing. Perhaps "Early access" label wasn't clear for users, now they started tagging such systems as "beta".
                        There was only one release of Unreal Engine with raytracing - it required at least 1.5 years of work and large rendering refactors.

                        A year ago, before announcing RTX cards, people didn't believe who could actually have realtime raytracing soon. They were like "yeah, maybe in the decade".
                        It's already working for companies working on beefy GPUs for visualization purposes. It's huge progress since the last year.
                        In case of video games, most probably raytracing will become common with the premiere of new consoles - coming the next holiday.

                        It's not a fraud, it's a natural process of developing software.
                        To be clear on this: Im in general a fan of Epic Games and the progress that the engine takes in the last years is really amazing.

                        Im not a newbie to software development, i work in a team of 10 people, sure thats not comparable to the UE development, but its a similar system. Its often not easy to work on the same Feature with many people and it requieres a good management and communication. But in general more developers can work on different features. And it seems that the core team at Epic shifts focus too often in my opinion. It seems that they have to do this bcause the team is too small. If they increase their team dramatically, there could be a core team of 5-10 people which concentrate on one specific feature and develop it until it can be used in production. But Epic opened tons of new features in the past and always shifted focus so that many features are not ready to use.

                        Its a general trend, that they open dozens of new features, but none of them is really usable. So there are 2. options in my opinion. Increase the team size, that they havent to shift focus from one topic to another or finish one feature before concentrating on the next one. Its not useful for users of the engine to have hundreds of early acess features, they can only experiment with.

                        and there are literally hundreds of topics they opened from rendering in general, to vfx to sound to VR/AR to multiplayer management, to scripting to modeling to animation and so on. Its nice that they work on so many new features, but for customers i think its a better way to finish one topic after another
                        Last edited by Lesuge; 08-03-2019, 10:45 AM.

                        Comment


                          Originally posted by Lesuge View Post

                          the point is simple: Epic Games makes raytracing and other features seem much better than they really are at the moment. And that's just misleading at a certain level. Nobody would have had a problem with that if they had taken a few more years, or at least made it clear that the technology isn't far enough at the moment. That would be honest and fair. But since you have done exactly the opposite, I think you can also be criticized for that. I think your comparison limps at 2 points:

                          1.) you can code with a lot of people on a topic if the management is good and the work is divided into many small sections.

                          2.) you can work with a lot of people faster and at the same time on several large projects and complete them. Epic develops new features like ControlRig, AR Implementation, Niagara, PixelSteaming, makes a huge breeze around these features, but then apparently changes focus and doesn't complete any of these features that are actually useful for developers. With a much bigger team, you might be able to make new features usable. And don't always open up new topics, increase people's anticipation extremely, and then sometimes not develop the features further for months or years.

                          i'd rather have 3 stable features than 50 experimental /beta features that you can't use in productions for years
                          Uh, so just dont' go all in developing with experimental features then?

                          if you're as experienced a UE dev as you say, you'll have probably realized by now that this is just how it works with UE.

                          1. Feature goes in (experimental) < - This is where you are apparently developing. (You should just be experimenting with the feature, or learning the concepts behind it)
                          2. It's broken or incomplete for a while
                          3. Its ready <- This is where you *should* be developing a shippable product with that feature.

                          Also, when it comes to Raytracing it seems pretty obvious that it's basically just a gimmick feature to sell Nvidia cards at this point. I actually laughed when I saw the "real time reflections" Star Wars demo , because the fidelity was only marginally better than what you'd get with SSR or other way cheaper effects but it was being hailed as the next coming.

                          tldr -

                          1. temper your expectations abit, and don't invest too much effort in developing for obviously experimental features.
                          2. realize that technology is and always has been more hype than substance.

                          Patience is a virtue
                          Last edited by UnexpectedSquirt; 08-03-2019, 02:05 PM.

                          Comment


                            Originally posted by Lesuge View Post
                            And it seems that the core team at Epic shifts focus too often in my opinion. It seems that they have to do this bcause the team is too small. If they increase their team dramatically, there could be a core team of 5-10 people which concentrate on one specific feature and develop it until it can be used in production. But Epic opened tons of new features in the past and always shifted focus so that many features are not ready to use.
                            Except Epic has over 1000 employees... I'd assume there are quite a few coders in that mix.

                            Its a general trend, that they open dozens of new features, but none of them is really usable. So there are 2. options in my opinion. Increase the team size, that they havent to shift focus from one topic to another or finish one feature before concentrating on the next one. Its not useful for users of the engine to have hundreds of early acess features, they can only experiment with.
                            Yeah, they experiment a lot and get feedback on them. A lot of them get shelved because there wasn't enough interest in them and it's not worth the time to finish developing. I mean look what happened with material layers, it was worked on for quite a while, but didn't quite pan out the way they wanted it to, so they shelved it.

                            I've never tested it, but in this day and age, I'd almost be certain that the editor occasionally sends telemetry data back to Epic for what's being used and what isn't. Sure, the editor can run just fine without a connection, but if it is connected, I'd bet it randomly sends information. It would be foolish if they didn't. Not just crash report data, I'm talking about from the actual editor itself. It probably wouldn't be specific code or anything and would be anonymous data, but it could check what features are being used and such. It would be an easy way to get a feel for what percentage of users are actually using the experimental features, to see if they should continue pumping resources into it or not.

                            Comment


                              Hi,

                              I'm posting this again, but now with more information about the problem:

                              I'm trying this tutorial, but it doesn't work on 4.22:
                              Some people say on the comments that it's not working on 4.22, and it's not working to me also.

                              I opened the old version of my project on 4.20 and tried the "Mask material only in early Z-pass". Project Settings > Engine - Rendering > Optimizations > Mask material only in early Z-pass.

                              On 4.20 it works and I have a lot more frames per second, but on 4.22 it doesn't work.

                              Someone know if it's fixed on 4.23?

                              If it's not fixed on 4.23, I think I will downgrade the project to the 4.20 version.
                              Last edited by Odenir; 08-03-2019, 12:33 PM.

                              Comment


                                Originally posted by Lesuge View Post
                                If they increase their team dramatically, there could be a core team of 5-10 people which concentrate on one specific feature and develop it until it can be used in production.
                                Last time I counted people submitting code and content for dev-niagara, it was around 15 people. If I remember correctly... Tried to count again now, but after merge from other branches, all engine commits are mixed...
                                They already do what you're suggesting. It just takes time, especially if have to make every feature working properly on all platforms. Figure out entirely new workflows, write a lot of custom UI, handle all edge cases...

                                Originally posted by IronicParadox View Post
                                I've never tested it, but in this day and age, I'd almost be certain that the editor occasionally sends telemetry data back to Epic for what's being used and what isn't. Sure, the editor can run just fine without a connection, but if it is connected, I'd bet it randomly sends information. It would be foolish if they didn't.
                                Of course. One example would be counting how many times Matinee will be converted to Sequencer in 4.23.
                                https://github.com/EpicGames/UnrealE...626bfb375f00fe

                                Comment

                                Working...
                                X