Support NeoGAF

[ErRor88]

In preparation for tomorrow's conference.



Ready for what’s next?

Join us on April 5 for our State of Unreal livestream, where we’ll explore what’s in store for the future of games—and announce some exciting news. Want a reminder when the livestream’s due to start?

State of Unreal

The State of Unreal livestream has wrapped up—watch recordings of the keynote and tech talks here.

www.unrealengine.com

Time and Date	Speakers	Details
11:00 AM EDT | April 5​	Tim Sweeney, Kim Libreri, Dana Cowley, Nick Penwarden, Zak Parrish, Teddy Bergsman Lind, Sjoerd de Jong​	State of Unreal Keynote​ Ready for what’s next? Don’t miss our keynote for some very special Unreal Engine news.
12:30 PM EDT | April 5​	Jerome Platteaux, Votch Levi, Scott Clifford​	The Matrix Awakens: Creating a World​ Discover how Epic's Special Projects team leveraged UE5 to create the environment in The Matrix Awakens. We’ll cover our modular procedural approach to building creation; Open World workflows (One File Per Actor, Data Layers, HLODs, level instancing); lookdev breakdowns of materials and texturing; and updates on improvements to Lumen, our global illumination and reflections system.
1:30 PM EDT | April 5​	Quentin Marmier, Robert Osbourne, Julien Marchand​	The Matrix Awakens: Generating a World​ Find out how we procedurally generated the city for The Matrix Awakens demo using a combination of Unreal Engine-native procedural tools and a game-changing new Houdini workflow. We’ll dive into the different construction steps and technical challenges that we faced along the way.
2:30 PM EDT | April 5​	Andreas Suika​	Creating your first game in Unreal Engine​ Tune in for a fast-paced and inspirational tech talk on creating a game in Unreal Engine.

 

[staticshock]

We are less than 24 hours until the official launch of UE5.0 and we being able to download and run all 3 UE5 demos on the PC using only ~7-10 total ram without any superfast SSD, god tier I/O (direct-storage), or decompressor chip (RTXIO). Contrary to popular beliefs.

Let that sink in!

 

[Punished Miku]

This conference should have been 2 years ago prior to the launch of new consoles. Better late than never I guess.

 

[ReBurn]

Now all games will be Unreal. Talk about a boring reality.

 

[Notabueno]

Unreal 5 is a failure because their big marketing around tech like Lumen or Nanite is like Nvidiaworks vaporware: not actually viable for a full-scale product, and doesn't solve the incredibly rising inefficiency of game production tools compared to the accelerated demand.

 

[ChiefDada]

We are less than 24 hours until the official launch of UE5.0 and we being able to download and run all 3 UE5 demos on the PC using only ~7-10 total ram without any superfast SSD, god tier I/O (direct-storage), or decompressor chip (RTXIO). Contrary to popular beliefs.

Let that sink!

You and I have gone back and forth about this and I was suspicious about whether your disagreements were in good faith or if you were trolling. After seeing your takes on other topics, I do believe you come in good faith therefore I will try again:

Those who disagree with you on the UE5 topic do NOT believe a super fast SSD and i/o is necessary for UE5. Just that UE5 streaming scales very well with SSD and I/O performance, i.e. more performant i/o will yield better performance when a game's design is dependent on efficient streaming and management of large assets. I imagine there is similar scalability between UE5 Lumen Tech and GPU performance.

 

[vpance]

Unreal 5 is a failure because their big marketing around tech like Lumen or Nanite is like Nvidiaworks vaporware: not actually viable for a full-scale product, and doesn't solve the incredibly rising inefficiency of game production tools compared to the accelerated demand.

They were just LTTP this gen, sadly. UE5 is supposed to simplify a big chunk of the artist pipeline in modeling and map design with Nanite. Trouble is we'll still be waiting on these titles by year 4 or 5.

 

[mckmas8808]

Is this their post-GDC information?

 

[mckmas8808]

Unreal 5 is a failure because their big marketing around tech like Lumen or Nanite is like Nvidiaworks vaporware: not actually viable for a full-scale product, and doesn't solve the incredibly rising inefficiency of game production tools compared to the accelerated demand.

You don't feel a tad bit silly for saying something like this?

 

[Neilg]

doesn't solve the incredibly rising inefficiency of game production tools

LOL
So there's no time saved by the 3d artists who had to spend well over half their working time creating LOD's instead of getting on with design/detailing still need to do that?
The lighting artists who after placing lights, had to spend days and days dropping reflection probes and building lightmaps, often having to send QA requests up to the 3d team because the UV channel used for the lightmap was fucked in some way?
The level designers who had to completley rethink layouts because there were too many assets in one area and they need to force a background load?

And apart from the massive lift in manpower needed to custom make high detail assets of large open world games (that level of detail being the norm isnt a problem the engine can solve, but also, that's why epic own quixel...), houdini is become better integrated into UE (4&5) all the time for procedural generation.

 

[Xdrive05]

Curious to find out how well UE5 scales down to the RTX 3050, running as executable and not in the bloated editor.

Does the release version actually use RTX silicon to accelerate their GI (lumen, was it?), or is that still done in comp? Also wonder if UE5 leverages RDNA2’s acceleration hardware as well.

One critique of the engine so far is that it seems geared toward high fidelity 30fps on current gen console hardware and may not be so great for high fps by nature of it’s heavy streaming tech. I’d like to see this challenged, too.

 

[elliot5]

Curious to find out how well UE5 scales down to the RTX 3050, running as executable and not in the bloated editor.

Does the release version actually use RTX silicon to accelerate their GI (lumen, was it?), or is that still done in comp? Also wonder if UE5 leverages RDNA2’s acceleration hardware as well.

One critique of the engine so far is that it seems geared toward high fidelity 30fps on current gen console hardware and may not be so great for high fps by nature of it’s heavy streaming tech. I’d like to see this challenged, too.

Lumen Technical Details in Unreal Engine | Unreal Engine 5.0 Documentation

docs.unrealengine.com

Yes it can leverage hardware raytracing vs software

 

[Hobbygaming]

I'm looking forward to seeing more UE5 games that push the hardware

I'm especially excited to see what Bend studio can do with UE5 after how good Days Gone looked on UE4

 

[staticshock]

You and I have gone back and forth about this and I was suspicious about whether your disagreements were in good faith or if you were trolling. After seeing your takes on other topics, I do believe you come in good faith therefore I will try again:

Those who disagree with you on the UE5 topic do NOT believe a super fast SSD and i/o is necessary for UE5. Just that UE5 streaming scales very well with SSD and I/O performance, i.e. more performant i/o will yield better performance when a game's design is dependent on efficient streaming and management of large assets. I imagine there is similar scalability between UE5 Lumen Tech and GPU performance.

That's actually not true. Now when you say UE5 streaming you are referring to nanite and not the typical data loading that happens on a game loading (which scales with SSD & IO).

But for nanite, all the information we have gotten is that at its highest levels nanite has modest SSD and I/O requirement.
Your GPU will crash before any NVME is even 25% fully utilized. That's is because nanite ACTUALLY scales with GPU power and resolution, not SSD or I/O.

This is why the valley of the ancient ran natively at 1080p on consoles.
This is why the Matrix Awaken runs just over native 1080p on consoles.

Remember the streaming requirement of Matrix Awaken was 300 mb/s, just above the Valley of the ancient demo streaming requirement on PC at its peak?

That's not even above or maximizing SATA SSD territory!

If you were to put so much asset in one place with it all intersecting (ala valley of the ancient). Your GPU will melt before nanite requirement even tops SATA SSD speeds. This is why nanite's performance on the valley was 2x more expensive in the Valley demo than the PS5 2020 demo. This is why they had to go from 1440p down to 1080p. Remember 1440p is 2x 1080p.

You can swap all kinds of SSD from the latest NVME 4.0 7gb/s to the lowest SATA SSD drive (500mb/s) and your performance will stay locked. No change. Unless you go back down to HDD (~150 mb/s).

The second you swap GPU or resolution, your performance instantly changes. Proving nanite has a modest SSD requirement and scales highly with GPU power.

This is why you don't need any of the equivalent I/O and decompressing solutions from the consoles (direct-storage/rtxio).

I'm not gonna rehash all the direct quotes from Unreal engine engineers, as you said, you can get them from my post history.
Nevertheless everything i have said is directly from Epic Games and the creators of UE5 and nanite itself.

 

[Notabueno]

LOL
So there's no time saved by the 3d artists who had to spend well over half their working time creating LOD's instead of getting on with design/detailing still need to do that?
The lighting artists who after placing lights, had to spend days and days dropping reflection probes and building lightmaps, often having to send QA requests up to the 3d team because the UV channel used for the lightmap was fucked in some way?
The level designers who had to completley rethink layouts because there were too many assets in one area and they need to force a background load?

And apart from the massive lift in manpower needed to custom make high detail assets of large open world games (that level of detail being the norm isnt a problem the engine can solve, but also, that's why epic own quixel...), houdini is become better integrated into UE (4&5) all the time for procedural generation.

Nobody designs in UE, not what it's made for, which is why LOD is better worked out with other tools like Simplygon until nanite becomes usable, but guess what...it's not because it's badly implemented unless working with Quixel integrated assets, implementing it for your own asset is miserable.

For the light it's not bad however, except again: nobody uses Lumen because it's not viable for a shippable product.

I don't know how UE5 makes Houdini integration better, since it was already integrated.

I think you missed my point: UE5 is for now a "fake" next-gen product like Nvidia vaporworks that nobody ever actually used for it's advertised functionalities and innovation, but that's not even my point either. My point is tools like UE5 do no fix video games and 3D production problems which are becoming massive in that they don't evolve and answer at the pace demand. It's the internet vs social network problem all over again.

 

[Notabueno]

You don't feel a tad bit silly for saying something like this?

If I didn't get paid Ks to work with people you'll never get to, maybe I'd question it, but also I like to make points in radical way.

So let me rephrase: UE5 is a failure in terms of responding to the growing need in accessibility, complexity, expectancy and optimisation of assets and varied experience production. No game or experience actually uses any of UE5 functionalities (maybe except for 2D particles which is old), because it's not viable for shippable products.

The same happened to the internet when social network became popular: fuck paying 50-200k for websites that have barely evolved in 20 years when the demand is so high, let's just use non-proprietary, ugly FB pages or twitter. Well the same is happening in 3D production: tools have been evolving too slowly compared to the evolution in demand, management and production, which is -one- of the reasons why AAA are more and more bland, clanky, limited...and why all indie games look the same without having evolved in scope.

UE5 fails at fixing any of these problems, unlike UE4 which was a major game changer for years. By the time Lumen, Nanite, fast Memory buffering etc...get viable, they'll just be a drop in an ocean of problem that doesn't address the accelerated demand and growing complexity.

 

[Notabueno]

They were just LTTP this gen, sadly. UE5 is supposed to simplify a big chunk of the artist pipeline in modeling and map design with Nanite. Trouble is we'll still be waiting on these titles by year 4 or 5.

They're weren't so late, they're actually the first to implement primitive shaders (which is a quite recent paper) in-engine, and you're right, it'll probably only be viable in a 4-5 years. The problem is larger in that as a platform and engine, it WON'T address the major lack of...let's say heuristics, streamlining and optimisations for 3D production in face of the huge demand growth and therefor expectations in production and management.

3D tools have become huge steam machines that added a LOT of complexity in the past 10 years and barely solved or streamlined the production process.

 

[vpance]

The problem is larger in that as a platform and engine, it WON'T address the major lack of...let's say heuristics, streamlining and optimisations for 3D production in face of the huge demand growth and therefor expectations in production and management.

I think I get what you're getting at. Even if artist pipeline issues are improved there's still other multiple bottlenecks in production.

UE5 won't be the savior of it all of course. I think we're all just hungry to find out what this gen can bring so we want it to succeed. If publishers are happy to put out 4K60 last gen+ games for the next 5 years because of rising costs or expectations then unfortunately that's a hurdle that won't be solved anytime soon by any engine.

 

[kyliethicc]

Gimme Fortnite 3

 

[SLB1904]

someone hug this man lol

 

[jonnyp]

Sweeney is so socially awkward.

 

[niilokin]

i hope twinmotion gets UE5 treatment soon

 

[ckaneo]

Unreal Engine is good, but I just wish they would focus more on performance. The fact that the matrix demo didnt run flawlessly tells us all we need to know about this era. Because the engine is so easy to create for, optimization becomes secondary. Developers add as much shit as possible from teams around the world and just knda hope it works. They gotta put more effort into lowering the performance hits

 

[Notabueno]

I think I get what you're getting at. Even if artist pipeline issues are improved there's still other multiple bottlenecks in production.

UE5 won't be the savior of it all of course. I think we're all just hungry to find out what this gen can bring so we want it to succeed. If publishers are happy to put out 4K60 last gen+ games for the next 5 years because of rising costs or expectations then unfortunately that's a hurdle that won't be solved anytime soon by any engine.

Unfortunately, this could last for longer than 5 years, as the barrier of entry (because of the recession) is getting higher and higher, which puts more pressure on the means, the core of it being tools. The same happened to the internet as I mentioned, and only barely now "codeless" design tools (figma, XD, webflow etc...) are emerging.

But of course the technical aspect of game is not the only thing that matter, game design, art direction, conception etc...does not all rely on tools.

 

[niilokin]

oh shit the repository with material code is awesome

 

[mansoor1980]

that was a glimpse of the next witcher game on UE5

 

[SLB1904]

this thing so far

'' the possiblities... cant wait to show bla bla''

 

[ParaSeoul]

 

[mansoor1980]

new tomb raider..............DAMN

 

[Snake29]

Isn't that the same demo from last time, but just extended?

 

[Xyphie]

Nice to hear Crystal Dynamics isn't stuck in Marvel hell.

 

[Black_Stride]

that was a glimpse of the next witcher game on UE5

Nah that was the medieval level demo for Unreal Engine.
It just so happen to kinda look like a The Witcher game.

 

[mansoor1980]

Nah that was the medieval level demo for Unreal Engine.
It just so happen to kinda look like a The Witcher game.

looked great

 

[mckmas8808]

If I didn't get paid Ks to work with people you'll never get to, maybe I'd question it, but also I like to make points in radical way.

So let me rephrase: UE5 is a failure in terms of responding to the growing need in accessibility, complexity, expectancy and optimisation of assets and varied experience production. No game or experience actually uses any of UE5 functionalities (maybe except for 2D particles which is old), because it's not viable for shippable products.

The same happened to the internet when social network became popular: fuck paying 50-200k for websites that have barely evolved in 20 years when the demand is so high, let's just use non-proprietary, ugly FB pages or twitter. Well the same is happening in 3D production: tools have been evolving too slowly compared to the evolution in demand, management and production, which is -one- of the reasons why AAA are more and more bland, clanky, limited...and why all indie games look the same without having evolved in scope.

UE5 fails at fixing any of these problems, unlike UE4 which was a major game changer for years. By the time Lumen, Nanite, fast Memory buffering etc...get viable, they'll just be a drop in an ocean of problem that doesn't address the accelerated demand and growing complexity.

Thanks for completing that thought. Interesting take.

 

[ErRor88]

 

[Draugoth]

 

[Black_Stride]

looked great

Whats scarier is the original demo was Unreal Engine 4 and still pretty performant and looked amazing.

 

[reksveks]

Unreal 5 is a failure because their big marketing around tech like Lumen or Nanite is like Nvidiaworks vaporware: not actually viable for a full-scale product, and doesn't solve the incredibly rising inefficiency of game production tools compared to the accelerated demand.

Failure for who?

 

[onesvenus]

UE5 is a failure in terms of responding to the growing need in accessibility, complexity, expectancy and optimisation of assets and varied experience production.

Doesn't Nanite solve precisely the complexity and optimisation of assets? I mean, not having to generate LODs manually is a huge time saver in itself.

 

[Bernd Lauert]

Failure for who?

Failure for us 60 fps connoisseurs.

*cries in OLED*

 

[Black_Stride]

looked great

This is the specific screenshot of the competed level the CDPR dev was referencing.
It looks like the notice boards from The Witcher.

Redo this demo in UE5 and as long as CDPR can implement their tools properly then The Witcher (4) will be an absolutely stunning game.

 

[Black_Stride]

If I didn't get paid Ks to work with people you'll never get to, maybe I'd question it, but also I like to make points in radical way.

So let me rephrase: UE5 is a failure in terms of responding to the growing need in accessibility, complexity, expectancy and optimisation of assets and varied experience production. No game or experience actually uses any of UE5 functionalities (maybe except for 2D particles which is old), because it's not viable for shippable products.

The same happened to the internet when social network became popular: fuck paying 50-200k for websites that have barely evolved in 20 years when the demand is so high, let's just use non-proprietary, ugly FB pages or twitter. Well the same is happening in 3D production: tools have been evolving too slowly compared to the evolution in demand, management and production, which is -one- of the reasons why AAA are more and more bland, clanky, limited...and why all indie games look the same without having evolved in scope.

UE5 fails at fixing any of these problems, unlike UE4 which was a major game changer for years. By the time Lumen, Nanite, fast Memory buffering etc...get viable, they'll just be a drop in an ocean of problem that doesn't address the accelerated demand and growing complexity.

Nanite static meshes are a complete game changer because most games dont actually have destructable environments.
So basically not needing to budget for your static meshes changes quite alot.

 

[SLB1904]

This is the specific screenshot of the competed level the CDPR dev was referencing.
It looks like the notice boards from The Witcher.

Redo this demo in UE5 and as long as CDPR can implement their tools properly then The Witcher (4) will be an absolutely stunning game.

i just hope I'm alive when it releases.

 

[Shubh_C63]

Sweeney is so socially awkward.

Don't let that man reduce to being a barber.

 

[thebigmanjosh]

All the studios working on games with UE5:

Where's Blue Box??

 

[mckmas8808]

Awesome!

 

[IntentionalPun]

If I didn't get paid Ks to work with people you'll never get to, maybe I'd question it

 

[digoutyoursoul]

I can’t wait to see what the next Mass Effect game will look like on UE5.

 

[vpance]

The 12:30 presentation



You can download the Matrix project from the Epic Launcher UE5 marketplace

 

[CamHostage]

All the studios working on games with UE5:

PS VR2 being on the list (that's not a developer, so that's a special call-out) is interesting. VR hasn't been emphasized in UE5, but it seems like Sony is making an effort to change that with its anticipated next piece of hardware. (I don't see Meta or Oculus Studios on the list, but hopefully there'll be spill-over and further adoption.)

Currently UE5 is only semi-compatible and rudimentary with VR. You cannot use Lumen (and may not ultimately ever be able to in VR, as Epic has said, "Not supported in VR - extremely high resolution and framerate requirements make dynamic GI infeasible") and as of May 2021 Nanite was not yet supported either... which makes it not a very "UE5'ish" VR experience if it can't use the hot new Nanite or Lumen features, but at least Nanite is on the roadmap. There are still good and vital things you can do with UE5 that you can't do (or do easily) in UE4 even without Lumen or Nanite, so it makes sense to start moving VR projects into the next version of the engine, but I hope the early involvement of the PS VR2 team is feeding back into the VR aspect of UE5 to help that be as extraordinary as the rest of the package (and in turn help make PS VR2 the huge step up in VR that PSVR fans are hoping for.)