Current Generation Games Analysis Technical Discussion [2024] [XBSX|S, PS5, PC]

[Pjotr]

I would prefer mesh shaders with continuous LOD and higher performance over Nanite.

We also have games like A Plauges Tale Requiem that has a lot of foliage on screen and very little noticeable pop-in, and that game doesn't use Nanite or mesh shaders.

Nanite I feel, has just because away for developers to be lazy (and I hate that term) with certain things like LOD transitions, and not only is it not delivering the results on screen, it's also costing a lot of performance.

A Plague Tale Requiem is also not open world which means making LODs is less time consuming. Calling developers lazy means you still don't get the point. They are not lazy, they have a certain budget and time frame to make a game (and in some cases it's a new and less experienced team) and without nanite that same game would have looked worse or not be made at all.
Also Nanite does use mesh shaders for large triangles and I think you are overestimating the performance overhead of nanite. Just look at Satisfactory where they transitioned to nanite for some of their geometry and that made their game actually look better and have less LOD pop-in without tanking frame rate.

 

[DavidGraham]

Nanite doesn't completely eliminate pop in though (at least in it's current state), you still get shadows pop in, characters/dynamic objects pop in, light sources, light areas pop in and HLOD pop in. For old UE5 games, there is also foliage and terrain pop in.

In the beginning of this generation we intensely debated the importance of extreme high poly count vs the importance of accurate lighting (especially in regards to ray traced reflections and illumination). I think UE5 embodies the idea of that comparison perfectly, Lumen has been far more important and effective for image quality improvements and the next gen feel than Nanite.

However, we've yet to see a densely populated city (like in The Matrix demo) replicated in an actual UE5 game, it's quite possible that without Nanite, this kind of environment would be a pop in fest, restoring the importance of Nanite in this kind of environments, and putting it on equal footings with Lumen.

 

[Shifty Geezer]

I think davis.anthony appreciates that - he calls out his own use of the word 'lazy' - but there isn't a straightforward term to replace it. Maybe we should use something like 'compromised'?

"Nanite I feel has become a workaround for developers who are compromised."

 

[trinibwoy]

Resource constrained? After all the primary benefit of a game engine is to trade money that you have for time / knowledge / skill that you don’t.

 

[Scott_Arm]

I think nanite is just an easy scapegoat for problems. I get annoyed by the kind of anti-tech mentality that's seeped into gaming. I imagine it's largely driven by the price of the hardware. The truth is there are things in the world that can't be simplified and rendered with displacement. I can look around and see tons of objects on my desk That would really require a lot of geometry to get looking realistic in-game. Or even just something simple like a smooth curve. It's the most basic thing you'll see in the real world, and something a lot of games fail at spectacularly. People want to draw a line in the sand at "2015" or some other year people want to pretend the games look amazing (they already look dated), but that's really boring.

Nanite is one solution to rendering per-pixel geometry. It's technology in evolution. Many of the games coming out started as UE4, or are locked into a very old version of UE5. Silent Hill is UE 5.1. It's definitely the case that there can be games coming out that don't need nanite because they they don't have assets built to the quality that nanite was purposed for.

Comparing nanite to mesh shaders is disingenous. Mesh shaders is just an API. Nanite is an actual production technology. Nanite is a solution. A better comparison would be Nanite vs the geometry "engine" in Alan Wake 2. Not all mesh shader implementations will be created equal. Alan Wake 2 is a beautiful game. Huge fan. I'm not going to criticize it in any way. It has a geometry solution that is tailored perfectly to that game. It's also a game that is heavily heavily criticized for performance, so I'm not sure where the comparison goes here.

As for pop-in and lod transitions. We need better solutions. It's one of the most annoying visual problems in gaming. Nanite, in my experience so far, greatly improves in that area. Both geometry and shadow pop-in are plagues. If people come up with great solutions that use the hw rasterizer and mesh shaders, then great. What I don't want is the forward looking technology to be "just stick to discrete LODs. 2015 was great."

 

[iroboto]

I think nanite is just an easy scapegoat for problems.

I think what we're seeing is just pure frustration and anger about the general performance of UE5 releases. They want to see something done about it, and its not so much around the technology, I think they recognize it's quality, but the tradeoffs no longer appear to be worth it for some folks, especially when you look at the mid-market region. The high end market seem to be OK with UE5, but down low and in console land, the tradeoffs did not translate as well. Series S especially was impacted heavily.

But things do seem to be improving, albeit slowly. But first party productions on UE5 seem to be OK.

 

[Scott_Arm]

Also, I remember that scammer kid complaining about quad efficiency. Had to refresh myself a bit, but if there are drawbacks to nanite that shouldn't be one of them. Visibility buffer should have 1 material shader launch per material type for all pixels that share the same material. It decouples material from visibility. It's not like deferred rendering where if you had four triangles covering each pixel in a quad that you'd end up with each 1 active lane and 3 helpers per pixel. I'm not going to try to dig through nanite code because it's outside my skill level, but in theory it should not have the same quad inefficiency as small triangles with deferred rendering. If anything it's any non-nanite meshes in the scene, like foliage that might be drawn in a standard deferred rendered way. They do still have a gbuffer, which I think is for doing anything that still requires the non-nanite approach.

 

[QPlayer]

Resolution matters. You can't use too low a rendering resolution because the image quality will be scandalous.

In my experience, native 1080p should be the minimum for a SeriesX or PS5, even if we're talking about 60 FPS mode. 1080p with well-implemented FSR or TSR gives acceptably good quality. The first round of UE5 games often perform well below the 1080p render resolution, which should not be allowed. The newer UE5 games (OfftheGrid, Stalker2) already run in a good enough resolution. So this can be done better. UE5 must be programmed and optimized, not just switched on!

 

[Remij]

I think what we're seeing is just pure frustration and anger about the general performance of UE5 releases. They want to see something done about it, and its not so much around the technology, I think they recognize it's quality, but the tradeoffs no longer appear to be worth it for some folks, especially when you look at the mid-market region. The high end market seem to be OK with UE5, but down low and in console land, the tradeoffs did not translate as well. Series S especially was impacted heavily.

But things do seem to be improving, albeit slowly. But first party productions on UE5 seem to be OK.

Like was already said, many of the games that were in development are still on older iterations of the engine without some of the major performance improvements Epic have made to the engine since.

It will be worth it in the long run.. once we get to a certain point, it will be hard to go back. Even looking at amazing looking games like God of War Ragnarok... you can clearly see it is a last generation game in many ways compared to many UE5 games. It will be interesting when Sony's own first party engines evolve to that point. Their games are heavily carried by the amazing artists they have... but even they will start to look dated if they don't catch up soon.

 

[Below2D]

A bit long at 14 minutes, but I’d like to know what beyond3d thinks about this guy’s take? Seems a bit provocative to me and I disagree with a few of his points, but I’m no dev nor do I have much experience with graphics engines.

 

[Dictator]

The concept from the title alone is absurd - I am pretty sure every graphics programmer would argue that real-time lighting now is better than it was 9 years ago.
Nine years ago we had cascade shadow maps at best + pre computed radiance transfer at best in the rare games that actually used dynamic lighting. Then completly static cubemaps + SSR for reflections. That tech pales in comparison to ray traced shadows, virtual shadow maps, RT reflections or RTGI. Even really cheap DDGI is better and is 100% better than PRT in every way.

There were some rarer techniques back then which were interesting and closer to things we find it titles today, like SVOGI or Hybrid Frustum Traced Shadows, but they were few and far between and almost shipped in no titles.

 

[Shifty Geezer]

Skipping through the vid, seems to be covering a lot of existing and emerging techs like probes and radiance cascades, and saying these should be used instead of RT/Lumen. I didn't pay huge attention though as the presentation is entirely antagonistic, presenting it as championing a mission to fight game developers as if devs have some evil agenda to trash gamer's experience.

 

[Scott_Arm]

The scammer again. Cool.

 

[Charlietus]

A bit long at 14 minutes, but I’d like to know what beyond3d thinks about this guy’s take? Seems a bit provocative to me and I disagree with a few of his points, but I’m no dev nor do I have much experience with graphics engines.

This is a version of the outrage gaming YouTuber but for tech. Just ignoring parts of the discussion and omitting information to rile up the public and generate hate. Just like other outrage YouTubers, he knows that what he is saying is disingenuous, and probably doesn't even believe in it.

 

[Pjotr]

Yeah that person has been on a crusade against TAA, Nanite and Lumen for some time now to the point of attacking UE developers in his posts on reddit. The thing is, he does it al wrong. He just points to other methods and wants Epic to implement them instead of Nanite/Lumen. There are a lot of pros and cons of different techniques and UE is a general purpose engine with a lot of systems that need to be maintained. When swapping to a different method you need to make sure it works with all the different use cases and both on pc, console (and maybe even looking at the future of mobile.)

 

[Scott_Arm]

This is a version of the outrage gaming YouTuber but for tech. Just ignoring parts of the discussion and omitting information to rile up the public and generate hate. Just like other outrage YouTubers, he knows that what he is saying is disingenuous, and probably doesn't even believe in it.

Actually what he's doing is making outrage bait to funnel donations into his "development studio" which is supposed to save gaming by making a custom UE branch that incorporates ai, or something like that.

Donate

This is our crowdfunding page to raise our non-profit funds for fixing Unreal Engine 5. Our goal is to raise 900K to pay a team of graphic programmers to modify UE5’s source code. Including m…

threatinteractive.wordpress.com

This is our crowdfunding page to raise our non-profit funds for fixing Unreal Engine 5.

...

Please note we are not actively seeking donations until we unveil our game prototype and release more information about other effects we want implemented into UE5.

We are only showing this feature because some of our supporters asked for a way to donate early.

Thank you so much for your support!

Fucking lol. My studio of people that have never made anything are going to save gaming. Please send us money through youtube for our non-profit (have papers for that?) so we can have your money and not be legally obligated to do anything.

The ai shit was tacked on to one of his videos where he talked about this amazing scheme.

Edit: Also, 900K for a team of graphics programmers is nothing. He might get some new grads or co-op students with no experience and be able to pay them for a year. Surely that's enough to "fix" Unreal Engine. Honestly sounds like an easy way to pocket $900K at 20 years old, or whatever he is, and just never have to work again. Or, you know, any amount of money he can get because there's no accountability for that 900K. It's not like this is using one of the actual crowdfunding sites that tracks donations etc, and even those have accountability problems.

 

[Shifty Geezer]

Probably best that we withhold discussion about this source for a different thread. I think it's fair to say "Threat Interactive" is not considered a reliable contributor to this discussion and future references should be watched and vetted for a meaningful, informed contribution by those who feel it worth posting.

I think at best his coverage consolidates a bunch of tech from across the industry, without explaining how or why it should be present in all and every game, so that people are made aware of alternatives where otherwise they wouldn't be.

 

[QPlayer]

The best water effect ever made into a video game:




And insanely detailed airplane models:



More can be found in the game topic.

This runs on Xbox SX in 4K image quality. What you think?

 

[Karamazov]

The best water effect ever made into a video game:


View attachment 12500

I don't think you are convincing anyone here with that picture.

 

[RobertR1]

Which UE5 game is most like CP?

LoD transitions in CP when driving fast are most obvious. I’d like to see the UE5 equivalent to see how that’s handled.