UE5

Hello there,

My daughter and I have just heard UE5 is ready for preview. Has anyone here tested it yet and what are your thoughts so far?

 

...and so it begins.

brb microwaving popcorn

 

My thoughts are disbelief at the size of the demo - apparently 100 GB. I cannot imagine how large an actual game project would be, as well as the final build size.

Two posts in literally 6 years. I'm always fascinated by this. Do you come here often, just never find anything interesting to comment on? If not, what brought you here today?

 

Nope. They're aiming at 2022 release, so there's not much point. The only interesting piece of tech there for me is Lumen (becaue I'll be the person who will take Lumen and see if it works with low poly.), and animation improvements.

However, following piece of information might be interesting for those eyeing high fidelity capabilities.

 

This thread reminds me of another thread a few weeks/month back talking about the state of unity and unreal. It was deleted off the face of this forum. Maybe this one will magically disappear as well.

 

The popcorn.

You mean this one?
https://forum.unity.com/threads/unreal-engine-5-game-changer.889501/

 

No, a different one -- but it's deleted

 

I don't remember seeing any other one, now I'm feeling excluded... I'm always up for a flamewar.

 

As interested as I am to give it a try, I can't see how they're like "yeah it works magically really easy, no extra perf impact" while those system requirements are basically very high end.

I guess they were aiming for worst case scenario and getting it to run on top end hardware, but I wonder what can be done with something less. If not much, then it's not yet useful.

Honestly it's getting really hard to ignore how Unreal solves so many problems with built in tools (seriously, like half of the popular asset store tooling is UE's built-in kit) while we get mostly broken things like SRP and DOTS for the last 5 years. Unity sticks because of where it works best, but sometimes it's as if it's trying really hard to not even do that.

 

Apparently the demo is 24 GB when built: https://www.resetera.com/threads/unreal-engine-5-dev-workflow-stream-today-5-26.431627/post-65781908

How much would an actual game be at that fidelity?

 

Let's hope Unity can narrow the gap in the decades to come.

 

Few hundred gigabytes, most likely.

I believe when chasing high fidelity eventually you'll reach a point where it will be impossible to hire enough artists to produce media assets OR you hit storage/bandwidth limits.

--edit--

Imagine what would happen if somebody actually tried to create a life-sized GTA, except it would be 1:1 scale, with every interior present, every npc voiced and unique looking.

That will likely dwarf ISS construction budget and will be terabytes of space and decade in the making.

 

I believe in case of the other thread certain user showed up started talking about "Premium". (-_-) And that led to thread deletion.

 

To run the demo at 30fps a 12 core CPU is recommended?

I was more impressed with collaboration and terrain loading features than the visuals. That "film quality" character did not look like it would belong in a recent CGI movie.

 

I was expecting stuff like UE4's early scalability issues from early on in its release cycle but the poor performance of this even in EA is kinda staggering. Unsurprising to see that the performance of things like Lumen and Nanite are both dramatically less than they said they'd be outside of very controlled situations. Lumen far more usable than Nanite, however.

That said, I in no way want to try and develop anything NPR using either of these systems so this engine is right out the window for me.

 

Sure, I'm on a peasant 970 - but thats the third person example map .
I hovered my mouse cursor over the desert demo, and my computer immediately caught on fire.

 

I'm on a 1070 that's showings its age and I'm getting about 5fps more than that. It may not be a new card but this is ridiculous. I'm not about to try and develop any game in an editor that's running at an unstable sub-30fps framerate.

 

*** will solve. Most things are in Unreal 5 (2022), half of the plugins are in beta, not recommended for production.

 

Ah, thanks for the clarification. Still, I doubt Unity will have any progress toward a retort in 2022 anyway.

 

Its also limited to rigid meshes.

To be fair, I'm sure they will by the time this is released (or at least mostly ). Don't think anyone is going to pick this up now unless they have some special deal with Epic, nor is anyone expected to.

It is pretty cool to have a gander at though.

 

I like three things in this demo: Lumen and Data Layers are awesome and scene partitioning is way more superior than subscene shenanigans.

 

I don't know if anyone else noticed but they're using their own upscaling solution to achieve near 4K from 1080p and it's enabled by default. Search the link below for "Temporal Super Resolution". Here's hoping they improve performance or most of this will not be touched until the next console generation.

https://docs.unrealengine.com/5.0/en-US/ReleaseNotes/

I have only briefly played around with it but my favorites are: a far less cluttered UI and ray tracing being referred to as deprecated.

 

I too was looking at the pretty pictures....I actually wonder if maybe there is a use-case for Nanite, where maybe we don't necessarily author meshes with "millions" of triangles, but we DO author them simply without worrying about the count, and we take advantage of the system to not have to do normal baking, etc... If Nanite will work work in those situations, I see it being useful for the "average indie" that isn't going to have access to all these "movie quality assets."

The Lumen lighting system is also shiny, but as mentioned above, some of this stuff is requiring quite powerful hardware....and many indies either don't have that level of hardware(I don't, I'm on my GTX1070 still too ), or if nothing else their target customers may not have such hardware. So until either time passes where this hardware is more commonplace, or they fix some of the performance issues(or make it scalable, and viable for in editor use as well), I'm not sure how deep of an adoption they are going to get with us "average" users.

I didn't see anyone mention that new shiny sound system. That is actually quite interesting as well...making sounds similar to how you make materials using nodes.....don't know how viable it will be until I actually see it in practice, but it is certainly interesting.

 

Interestingly though, I tried converting the “Medieval village” to UE5. Now this map already ran like hot garbage in UE4 on my system … and it still does – but not all that much worse.

 

I haven't played with it and I already don't like that one. They made the choice to put it in the form of blueprints. Which is a dumb choice (from Unity too for that matter). Just take a look at any notable audio middleware or 3rd-party software. There is a reason the majority of them aren't noodles between boxes...
Until then... FMOD it is...

 

There are some interesting tidbits in the documentation. UE5 is still really really early. Stuff like blueprint nativization is not supported (not sure if epic is going in a different direction or just not ready yet). Also things like translucency and reflections still needing work.

The new input system looks similar to Unity's new input system.

The builtin world partitioning and what epic is calling data layers is nice.

As for nanite I am curious what would happen with performance if we put all our unoptimized geometry AND then compared with optimized geometry.

The virtual shadow map is interesting.

 

This is weird. Im on a 1060 and I get 120fps in the same project.

I have been learning to make prototypes in unreal using blueprints past few weeks and I have to say, editor performance wise it is night and day difference between unity. In unity with an empty project, entering playmode takes up to five seconds. Opening big heavy open world nature scenes in unreal are nearly instant. In unity I need (expensive) plugins to even create similar scenes and these plugins have lots of problems with the render pipeline nightmare going on right now.

Opening projects is faster is unreal. Rendering lots of stuff is faster. Lots of QOL things in editor that keep the RSI stresses down as well as reduce general mental fatigue. In Unreal I can sort and filter items in the browser. In fact, i can do that multiple ways. In unity I cant do it at all. In Unreal I can flip normal maps. Unity requires me to open photoshop, another program that takes up too much RAM and takes too long to open.

In unity even after several years of using it and a published game, I still dont really understand how lighting works and I struggle to make things look decent. In unreal its as easy as marmoset toolbag. The art tools were designed for artist. They are easy to find and they work as you would expect. All things are designed for fast iteration. You dont have to have some esoteric knowledge - to be among those elites who "know what they are doing". You just see the buttons and the names make sense and when you push them it all just works.

All these thousands of little things add up. The result is that in a few days I have a great time making a functional and good looking shooter prototype in unreal, whereas making a ball roll in unity is still a struggle for me. (I'm not a programmer...yet)

Overall, in Unreal, I am enjoying myself a lot more. I don't really care about the latest tech. But if the moment to moment developing is easier/funner/more efficient in one, thats where my recommendation goes.

I dont see myself using unity anymore unless I am making mobile games.

The old reason I used to think Unity was a better choice for me was because of more learning content and more active community. But these things arent true anymore. I am finding more and better quality instruction from unreal. And a ton of it is totally free.

I wouldn't recommend unity right now to noobies. I would recommend unreal. Everybody use to say unreal was more difficult, but to me that seems opposite from the truth.

There is also a clear culture difference. In the emails I get from unity it always seems like someone is after my cash. With unreal I get excited about the possibilities of game development. In teh documentation unreal is clear about current limitations. Unity is like the insecure guy at a party who talks to loud about himself. With unity I dont know where to go but everyone wants to tell me for a fee.

Its obvious that unreal has a lot more money. I wouldnt expect unity to have as many shiny bells and whistles. But what they (unity) are putting out lacks focus and makes it seem like some get-rich-quick-minded execs are in charge of the business.

Unity doesn't need to have more money to do well and be useful. But they have to scope down and focus on doing fewer things well.

 

Probably bandwidth difference. In Nanite, bandwidth probably will be the final bottleneck, no matter what. And the later the card the more bandwidth it usually gets (due to PCIE improvements). At least that's my immediate guess.

 

what determines bandwith?

 

Hm, just checked. it's probably not it in this case. It gets more interesting. I was probably wrong.

1060 / 970

Memory Clock 2002 MHz 8008 MHz effective / 1753 MHz 7012 MHz effective
Bandwidth 192.2 GB/s / 224.4 GB/s
Memory Bus 192 bit / 256 bit
Memory Size 6144 MB / 4096 MB
FP32 (float) performance 4,375 GFLOPS / 3,920 GFLOPS
Pixel Rate 82.03 GPixel/s / 65.97 GPixel/s
Texture Rate 136.7 GTexel/s / 122.5 GTexel/s

 

That's interesting.

One other note, not really related but interesting anyways, I have Unity running on my SSD, but Unreal is going from my HD. So I'd expect it to be slower. At least for booting up, but in every way I can tell it runs significantly faster.

Anyway, clearly I dont have anything nice to say, just pouring gas on the fire here but I been testing things out on my own and they seem to go counter to conventional wisdom so i feel compelled to say something.

 

With lumen and virtual shadows enabled? Seems to be whats killing performance.

This really is my biggest issue with Unity now. When they brought up the “500ms” goal years back, I was happy. Too bad its basically gone the other way. I cannot imagine working on a large-scale project in this engine.

Fun thing is that I don’t even mind slow compile times … if I can continue working. In Unreal is basically a fire and forget, oh my compile is done – cool. With near instant blueprints, there’s rarely time spend waiting.

 

I dont even know what Lumen and virtual shadows are so I couldnt say . I just opened the default project.

I actually started on unreal because I was trying to do a large project in unity and it just became too many development stopping issues all related to rendering and performance issues that basically cant be solved without a really expensive specialist.

Right off the bat with in-house gear in unreal those are non-issues. So all I got to do is be able to do the basic programming for a walking sim and the project I could not finish in unity is looking finish able in unreal. We'll see. Maybe something I don't know yet but based on testing so far it seems like basically switching engines solves the issue I was having.

I shouldnt have even tried the project in unity, it is better suited for UE's strengths, but people always say, "ah there is no difference. UE just looks nicer because they enabled some post process by default." But I think people just want to feel good about the habits they are accustomed to. For myself, I think the sort of games I want to make fit better in UE and also, looking forward into the future, I see one engine making big strides and the other doesnt seem to know where it wants to go. More importantly, one engine is developed by developers who do the work of making games, while the other "communicates closely with developers." Anybody who has worked on any team ever knows that 99% of problems are caused by the error-prone nature of human communication. If you want to really botch a job, you let it depend on convoluted communication (like not having boots on the ground help make big decisions).

 

No doubt UE5 has some impressive tech and I can imagine that companies that specialize in visualization that this will be a game changer for them.

For deving games we will have to wait and see. For indie devs while the technical limitations of incredible detailed worlds may be removed, the budget and time constraints of making them still exist. For pro devs if the world looks amazing people will expect the same detail in other facets of the game. To paraphrase Arnold Schwarzenegger in Pumping Iron, if you greatly increase in one area, you need to then increase in all other areas.

I'm curious to see how Nanite integrates with physics. Because if the tech decides how to render on the fly I can imagine there are limitations on handling physics outside the camera frustum.

I'm more interested in making interactive worlds - procedural character animations, next gen physics etc as I think they enhance gameplay more than just pure eye candy. Maybe more will be revealed at Siggraph.

One thing is there is no such thing as a magic silver bullet. Everything has strengths and weaknesses.

 

?

Frustum culling accounting for this sort of thing is a solved problem and doesn't have any notable issues transitioning to nanite. Also, why would physics specifically be an issue at all here?

 

The ratio of strengths to weaknesses is what matters and those vary wildly and are very important.

Dismissing everything and painting them as similar is just to reinforce being complacent.

"Nothing is perfect, so I may as well continue using the current imperfect tools I'm already using" <- Nah.

I would also like to point out that of course Unity users see an impressive feature and immediately think "I bet there's something really basic broken about it", Unity has trained us well over the years.

PS: No actual opinion on UE5 early access yet, I'm downloading it now to check it out, but it may be a while.

 

I'm curious if the terrain collision is going to be as detailed as what is rendered. Complex colliders are expensive.

 

I'm not dismissing anything. It looks awesome. But I'm keeping an open mind. As I said we will find out more at Siggraph.

There is no best way of doing anything. Nearly every problem there exists multiple solutions. It's up to the developer to work out which solution is best in their case.

 

Is a high detailed ground that looks like an actual real world ground but has imperfect foot IK worse than a plasticky lumpy ground that also has slightly imperfect foot IK?

The way these terrains are usually made is using a combination of displacement/tesselation shaders which work based on distance from the camera (so its optimized), plus lots of meshes which are scattered strategically (also making use of LOD's). This displacement is usually only doing small things like pebbles and twigs that would only cover like half of a human characters foot for example, and larger meshes just collide like any other mesh.

This isn't anything new, the only thing new is that now there is more triangles rendered.

The idea that "if you do one thing well you have to do evverything well" is wrong. This isn't how humans think. If a person gets big fake tiddies we dont all expect them to suddenly have a great personality too. We just oogle at the tiddies.

Graphics work the same way. Player sees something shiny, they are less likely to notice the turd right next to it. If all there is is turds, how can they not notice the turds?

Consistency is important, but not as much as developers seem to think. People have no taste, so we can all relax a bit.

 

https://i.gyazo.com/d98c641461f0aa4e0be8c7d8a34eb739.gif

Still downloading the Valley of the Ancients demo. Checking out blank project in the meantime.

Am I the only one having these strange effects on the ground? I wonder if it's connected to lumen.

 

Do you have lumen enabled? That looks more like their Screen space solution.

 

It's a blank new project. I wasn't aware you have to enable it manually?

When checking project settings for lumen, this is what's displayed:
https://i.gyazo.com/6a97766c3de4b7f2d2ab3c74f812bda1.png

 

Well, you can look at lumen source code and shaders now. I think they mixed the different techniques together. I saw SDF, mesh card, screen probe, ray tracing, path tracing, voxel lighting, radiance cache etc. Not sure what it means but that was in their source file. i'm not breaking eula, am i?

 

It's hard to come away from that demo not feeling impressed. Particularly at the ease of everything - with Quixel built into UE5 now, you could probably 'speedrun' a level build and smash out some amazing looking scene in 30 minutes. In Unity, it would probably take 30 minutes just to open the editor, download all the miscellaneous Asset Store packages & import them.

Nothing is really 'new' in terms of Epic's direction here - they're still the guys who build stuff that devs are actually gonna use, the FBBIK is a great example of that (and one area where I think up until now, Unity has actually been 'ahead' with the help of Final IK).

I am a bit skeptical of their new AA solution they showed there though - you could see the smearing when they placed any of those Quixel assets into the scene, and the camera never really moved too fast, and had lots of bokeh (a great way to hide those artifacts). Not that Unity's TAA has historically been great, mind.

 

I was getting 5 FPS on the crashed/moon vibe scene (forgot it's name), so the second scene you transition to.
Disabling lumen sent it to a pretty stable 35 FPS/30!

My hardware is pretty old. GTX 1070 (mobile) i7-6700HQ CPU @ 2.60GHz (mobile), 32 RAM.
It also didn't take ages for shaders like in UE4.
Nice.

 

I would like to temper people's expectations. I ran the UE5 demo on basically the best gaming computer that exists 5950x, rtx 3090, nvme pcie gen 4 ssd, etc. The performance and the visuals were not that impressive with all the artifacts. Pan the camera too fast and you see tons of lighting artifacts. Move through the world too fast and you can see the geometry not loading in fast enough.

We don't have the windows direct storage api yet and this is very early days, but overall this tech is only designed for people targeting the bleeding edge (next gen consoles or high end pcs). I don't know about you guys, but I want to reach a much wider audience than that particular subset of users. Overall this is very cool tech, but it's something that as an indie doesn't really matter atm.

 

I was mostly looking forward to the new UX and scripting changes. The UX changes are similar to the Unity font update in 2019.3, where all the icons have shrunk and favor a flat, simplistic look. The addition of drawers that can be minimized is awesome, and unity should seriously be taking notes.

Blueprints looks almost identical to UE4, which is very disappointing given we've seen better UX workflows with tools like Bolt 2 (before it was scrapped). I'd prefer a choice between vertical and horizontal scripting, with proxy connections preventing the common issue of spaghetti blueprint wires.

Not to mention we still haven't heard any news about the rumored Skookum successor scripting language (Verse), which is supposed to be compatible with UE4 and UE5. I'd prefer a scripting language more similar in syntax to c# and be built into the engine with a built in IDE.

The Graphics Tech is impressive, but I think the tech is going to be mostly used for making game development easier than ever, rather than creating extremely detailed CG graphics on next gen consoles. The problem is next gen consoles are difficult to get, and mobile gaming is growing rapidly. So more developers are going to want to target weaker devices, while the engine tech is growing more demanding. Many developers won't bother using Nanite for CG graphics, but instead will opt to use the technology to skip painful billboarding and LOD techniques all together.

I think the data layers for terrain will be fantastic for multiplayer games, which are always trying to add new content and update maps periodically.

The biggest reason why I'm excited for UE5 is because it'll push the competition to offer at least "a little more" than what they are right now. When UE4 was first made available for free in 2015, Unity had to make the free version much more enticing by essentially providing all the pro engine features for free. Hopefully with UE5 being available in preview and the new $1 million revenue model will push Unity to give us developers more.

 

I mean lumen is not beyond our imagination but unity should take a serious look at it. Some tech already existed in HDRP. Try to mix and match to make dynamic real-time global illumination solution for unity.

 

It's early days but Lumen looks less exciting to me when you can just do hardware ray tracing and skip all the crazy artifacts caused by Lumen. One of the best looking game out now is Metro Exodus Enhanced Edition which is doing all it's lighting with ray tracing, which runs and looks better than Lumen imo.

 

Yeah, I'm seriously curious about this too. I have the fear though that it won't work only on super-powered desktops. As soon as you want to target even just the average PCs, especially Macs (where usually everything is severely underpowered comparing to the screen resolution) you will need to develop a second set of features too. Which would mean the simplicity and developer-friendliness worth exactly nothing. Cool tech though and all and I'm seriously crossing my fingers to not to be right on this.