Unreal Engine 5 - unbelievable.

[Proto]

Fortnite in DX11 vs DX12 is a good example of frame pacing.

 

[FlyingScotsman]

Defo Proto

 

[Proto]

 

[Calin Banc]

As I said before - it's a demo, and sadly we never seems to get close - atlease with those darn UE!

Immortals Of Aveum looks pretty decent (aside from the childish art style - what I hate about Cyberpunk as well!, especially compared to Deus EX) and is using Lumen and Nanite. It would have been nice to use Chaos amongst the other features, as the game world seems too static. And a bit of that extra sound as well

At launch it was a mess, stutters all the way. Now is pretty good, not perfect, but miiiiiiles better. As per devs:

Our studio’s journey continues as we update Immortals of Aveum to UE 5.2 and begin work on our next project in UE 5.3. Here’s a sneak peek into some of what we’ll be exploring with the engine upgrade:

• Lumen – In UE 5.1, Lumen solved the indoor-to-outdoor lighting transition seamlessly, allowing four lighters to light over 15 levels, and also allowed our modelers to instantly view assets in a variety of lighting scenarios. In 5.2, we want to take that even further by improving lighting detail around characters and visual fidelity of animations.
• Nanite – Nanite gave us unprecedented geometric fidelity, while saving our artists countless hours of setting LODs. In 5.2, we’re looking to further enhance overall game visuals as well as faster geometry calculations that can help further reduce pop in.

These things, like @Nexus18 kept posting about 4A with their Metro, help develop games way faster and better. You can try the engine yourself and the demos are the proof it can look pretty. Now... if with 5.2 can run as good is it does now, I'm happy that it should run better with newer, better versions.

All in all, after playing again some Immortals with some patches... it does feel on the right track.

 

[MikeLoader]

Immortals Of Aveum looks pretty decent (aside from the childish art style - what I hate about Cyberpunk as well!, especially compared to Deus EX) and is using Lumen and Nanite. It would have been nice to use Chaos amongst the other features, as the game world seems too static. And a bit of that extra sound as well

At launch it was a mess, stutters all the way. Now is pretty good, not perfect, but miiiiiiles better. As per devs:


These things, like @Nexus18 kept posting about 4A with their Metro, help develop games way faster and better. You can try the engine yourself and the demos are the proof it can look pretty. Now... if with 5.2 can run as good is it does now, I'm happy that it should run better with newer, better versions.

All in all, after playing again some Immortals with some patches... it does feel on the right track.

Awesome news buddy, just awesome!

And yeah those 4A are so frikking great. All games from now on will Only be RT all the way they said, when releasing the version of Metro that wont run raster old crap

 

[Grim5]

Why Unreal Engine 5's Nanite sucks

 

[mrk]

Epic fail?

 

[humbug]

Why Unreal Engine 5's Nanite sucks

Eh... i'm not saying this is an unnecessary click bait video... but Nanite from Epic Games perspective was never meant to replace LOD's entirely, they never said that and all of the documentation explains you shouldn't use Nanite as flat a replacement for LOD's.

Nanite is used to optimise extremely dense tessellation, both visually and performance, this is does extremely well, lets say your scene is looking over 4KM^2 of dense foliage, a forest, if you wanted to optimise something like that using LOD's about 70% of your assets would be rendered at the lowest LOD, that being a 2D card of the asset, or 2 of them spliced together to form an X so that it looks vaguely like a 3D object from distance, the truth is it doesn't, it looks awful, the drop shadow from something like that, if you can even render shadows at that distance is equally as awful.

What Nanite does is retain the 3D detail of even the most distant asset giving your scene a much better look and it will do it at the same or more likely higher performance than using LOD's.

What you would do is use LOD's where appropriate and Nanite where it is appropriate.

 

[humbug]

Another thing that i would add to that ^^^ is LOD's is simply replacing the asset with a lower polygon count assets on the fly based on distance from the object, what that does visually is create often noticeable pop-in , you may have seen it in most game, where as you approach the asset it jarringly pops slightly different shapes getting more detailed as you get closer.
Nanite doesn't do that.
----------

Anyway...

This is what 750,000 very high polygon trees look like.

A few years ago when Nanite was first introduced i had a quick go to see how far i could push it, each one of these trees is 2 million polygons, about 10X higher than you should have it in a 3D game, on top of that there are 750,000 of them spread across a 4KM map, each tree, even those at a 3KM distance retains its detail and shading, while the performance is very low, as i said i'm deliberately pushing it as far as i can... this is impossible to do with LOD's at anything close to this detail, it just wouldn't even run, not even at 1 FPS.

The visual shimmering you're seeing is Temporal Antialiasing not being able to deal with very fine edges on some of the trees, like pine needles, it needs work and i didn't bother just for this.

 

[Grim5]

What you would do is use LOD's where appropriate and Nanite where it is appropriate.

I believe that's what he's complaining about - in the video he's saying that some developers don't know what they are doing, and are using nanite, lumen, shadow maps and LODS incorrectly, resulting in massive amounts of lost performance in modern games and then needing to embed DLSS into their game to try and make up the lost performance, when they could have achieved better graphics for less performance cost

Another one his videos touching on this at a broader level

I'm not a game programmer so I don't know if what he's saying is right, but what I do know is we have plenty of modern games that look worse than older games and yet run like ass - a very recent example is the new Star Wars game that came out two days ago. So when he says that modern games running poorly and looking bad is a self inflicted problem created by bad developers and poor tools, I kinda want to believe him.


Though I don't know what the guy's end goal is here; he claims to own an indie game studio that was working on a new game in UE5 and then they stopped and instead now he claims they are going hire more developers and start writing new tools for UE5 so games can be made the way he says they should be. It sounds a bit weird, like if they didn't like UE5 then just use another engine, but instead he claims the game was cancelled and now the company will instead focus on building graphics tools

 

[Proto]

Is the glass half empty or half full?

That’s the definition of a classic manufactured question.

All glasses are half empty or half full. It’s your definition that’s the problem.

Human beings are not perfect.

 

[humbug]

I believe that's what he's complaining about - in the video he's saying that some developers don't know what they are doing, and are using nanite, lumen, shadow maps and LODS incorrectly, resulting in massive amounts of lost performance in modern games and then needing to embed DLSS into their game to try and make up the lost performance, when they could have achieved better graphics for less performance cost

Another one his videos touching on this at a broader level

I'm not a game programmer so I don't know if what he's saying is right, but what I do know is we have plenty of modern games that look worse than older games and yet run like ass - a very recent example is the new Star Wars game that came out two days ago. So when he says that modern games running poorly and looking bad is a self inflicted problem created by bad developers and poor tools, I kinda want to believe him.


Though I don't know what the guy's end goal is here; he claims to own an indie game studio that was working on a new game in UE5 and then they stopped and instead now he claims they are going hire more developers and start writing new tools for UE5 so games can be made the way he says they should be. It sounds a bit weird, like if they didn't like UE5 then just use another engine, but instead he claims the game was cancelled and now the company will instead focus on building graphics tools

He's probably right then, tho i find it bizarre if they are reading documentation, i only watched the first 5 to 10 minutes of the video and from that i got the impression he was just bashing Nanite, fair enough then my fault...

The video you posted here is quite interesting, and he's right.

I'll give you an example of how ugly that can get and this problem from a game developers perspective, because that i am not, its just a hobby to me what i know is limited.... but learning more and more all the time.

And the result as they haven't yet got around to fixing it.... pay attention to the edges of the doors in the floor.

 

[Chuk_Chuk]

Why Unreal Engine 5's Nanite sucks

This is an interesting video. I guess there is no such thing as a free lunch.

I like his ending comment though. Stop wasting money on band aids and put that money towards process improvements to optmise scenes. Literally everyone wins, but that doesn't sell new GPUs.

I believe that's what he's complaining about - in the video he's saying that some developers don't know what they are doing, and are using nanite, lumen, shadow maps and LODS incorrectly, resulting in massive amounts of lost performance in modern games and then needing to embed DLSS into their game to try and make up the lost performance, when they could have achieved better graphics for less performance cost

Another one his videos touching on this at a broader level

I'm not a game programmer so I don't know if what he's saying is right, but what I do know is we have plenty of modern games that look worse than older games and yet run like ass - a very recent example is the new Star Wars game that came out two days ago. So when he says that modern games running poorly and looking bad is a self inflicted problem created by bad developers and poor tools, I kinda want to believe him.


Though I don't know what the guy's end goal is here; he claims to own an indie game studio that was working on a new game in UE5 and then they stopped and instead now he claims they are going hire more developers and start writing new tools for UE5 so games can be made the way he says they should be. It sounds a bit weird, like if they didn't like UE5 then just use another engine, but instead he claims the game was cancelled and now the company will instead focus on building graphics tools

I think these videos should be a thread of their own. It would start an interesting discussion to analyse and review his comments and raise awareness.

As you to say in your post, a lot of people have complained that games don't seem to look significantly better yet seem to require bonkers hardware.

 

[Calin Banc]

Why Unreal Engine 5's Nanite sucks

So... when he talks about complex worlds, with lots of detailed objects (07:14), he speaks about how developers had previously a solution: for a pile of junk, make it just "one" object, one draw call - cool, but it means is fixed, no interaction, so not really a solution for something complex, rather "fakery".

Then in an earlier section, he compares a scene with around 6mil triangles vs what, 10, 20, 30million used in a modern engine at 4k? He also talks about temporal AA, but presenting a 1080p resolution.

At the end of the day, a bit similar to RT, the question is: even with the performance overhead, doesn't this mean that:

1) you save dev time by not needing anymore to create a lot of LODs for each asset?
2) you have larger, more detailed world and objects due to nanite, otherwise not possible with the regular approach (high polly vegetation, fences, etc)?

So if this is true (more so the 2nd point), basically, use a more traditional approach for less complex games and scenes, go with nanite for the rest?

 

[Chuk_Chuk]

So... when he talks about complex worlds, with lots of detailed objects (07:14), he speaks about how developers had previously a solution: for a pile of junk, make it just "one" object, one draw call - cool, but it means is fixed, no interaction, so not really a solution for something complex, rather "fakery".

It is only one object when far from the player, as they approach it splits up.

Then in an earlier section, he compares a scene with around 6mil triangles vs what, 10, 20, 30million used in a modern engine at 4k?

From his explaination it seems like heavier scenes might have a bigger margin in favour of LOD.

He also talks about temporal AA, but presenting a 1080p resolution.

Not sure what issue you are highlighting.

1) you save dev time by not needing anymore to create a lot of LODs for each asset?

Some assets, I believe nanite can only be used on static meshes. It is also why he concludes that money should be invested on better process rather than upscaling methods.

2) you have larger, more detailed world and objects due to nanite, otherwise not possible with the regular approach (high polly vegetation, fences, etc)?

Might actually be the opposite if his issues about GPU power being wasted is worst with higher density scenes.

 

[humbug]

 

[humbug]

Good news, we have started the en****ification of Unreal Engine.

As of October we will be moving all your Quixel and Megascanns content from the excellent integrated Quixel Bridge to Fab, why did we do this? well the really great news is Quixel Bridge was intended for free in house high quality assets, so it only needed to be simple, easy to use and ergonomic, the great thing about Fab is in early 2025 we intend to charge for those assets, on top of this fantastic news we can now move all of that content to a place that is bloated, full of crap constantly trying to capture your attention, for monetary reasons of course, that will make it difficult to navigate and annoying.

What you shouldn't do, you shouldn't, under any circumstances purchase every single asset from Quixel, right now, for free while you still can, you must not do this.

Upcoming changes to Quixel related to Fab's launch in October

Last year we announced that Quixel Megascans would eventually be moving to Fab, Epic’s new digital marketplace. Our team is excited about the future, and we’re committed to supporting our community in developing breathtaking projects and experiences. Ahead of Fab’s launch in October, we wanted...

forums.unrealengine.com