Unreal 5

They have new compression features worked on already, Nanite should have like textures a big compression for geometry data.
( And next gen consoles also have hardware build in compression/decompression to manage that even better).

Like Brian said, in the demo they pushed the detail using Zbrush model to the extreme only to stress test.
While it’s exagerated and you really don’t need to keep such amount of micro details on some statue for a game (lot of details could have been stored on a normal map).



So you can always lower the models storage size, and Nanite allows you on game packaging to adjust the game size; you’ll have to choose between cinematic asset high detail and game size.

Games are always pushing graphics details, you can’t avoid it.
With features like virtual texture and virtual geometry becoming the new standard allowing even more details on screen big studios will push detail even further.

Current games like Watch Dogs 3 reach 100 or 130 Go, most big games are around 80g, that’s standard.
On current gen 4K and 8K textures is what makes games so big, on next gen geometry will also become as heavy as textures.

For physical games disc on console, game size is not an issue.
There has been disc, 80 Go disc, Blue Ray, perhaps there will SSD game cartridges or new storage innovations.

Game size will only be an issue for internet game download on PC or console that does not have disc.
But it’s too early to worry about that, UE5 is not available, next gen games will only come lot more later.

Perhaps the solution is simply pre download the game one or two days before , many games already do that allowing to preload the game one month before.

Worst, big studios will no more ask you to worry about game size or installing games on console or PC, this will be their work and they will onloy ask you to play only remote games as service

No more game size or installation worry, just pay subscription and pay games prices they choose :joy:

Be able to use any amount of static geometry polycount directly in the game level, without needing to group objects, without having to worry about having thousand of non duplicated unique objects.

This will be a big workflow change.

https://www.artstation.com/artwork/nYEDvE

Statue with 33 million triangles, no textures and no normal map micro details
(click zoom to see the amount of geomety detail)

https://www.artstation.com/artwork/kDl2aA

More detail screenshot

Just replying after a general/random reading in regards to the game size.

I read somewhere that modern games due to the limitations of SSD or HDD access speed, they need to “place” models and textures within the game in a very space non-optimized way. Basically everything is extremely fragmented or even repeated in order to compensate for the lack of memory access speed.

What the new storage technologies will allow as well, is to avoid this fragmentation, de-facto actually potentially reducing the size of games. True, with the increase in detail of models and size of texture, the memory required to store these will increase. But I believe that, with the new optimization and the price of memory going down, disk space won’t be a big deal for new games.

1 Like

Do you have some source link ?
I guess this is perhaps one of the many new key features necessary for maximum detail.

While on current SSD i guess Nanite will still give incredible graphics details compared to current games without going as high as the Unreal 5 demo.

I can’t remember the exact article, but after a quick google I found plenty of links that state that games on ps5 should be much more optimized regarding disk space, like this one:

I think there are 2 factors to keep in mind: the old consoles were using 5400rpm HDD. What I think this means is that no games what so ever had taken any advantage of the speed of an SSD. Sure you had faster load times, but the underling concepts were tailored on ancient HDD technology. This means that even current high-end SSD could probably give huge boosts in many areas if developers would create games around them. But this couldn’t happen until now because of consoles technology.

Second aspect is that on top of using an SSD, the new PS5 also has a dedicated controller. It is rumored that to achieve the same performance that such controller can give, you need something like 12 zen 2 cores dealing only with the storage IO. I believe however, that we will see an evolution of such technologies on pc as well really soon

Final consideration, is that all this hardware marvels will finally be really taken into real consideration when coding, and this is in my opinion what will allow the bigger step. The new PS5 controller may be revolutionary, but I am sure there’s plenty of speed in current SSDs that is completely useless at the moment, because games or programs don’t take advantage of it.

With all that said, I think this is going to be really exciting, as we’ll see fast storage technologies finally taken advantage of. Looking forward to see the effects of this!

1 Like

Sony will allow developers to control the SSD usage, same for Xbox ( and next gen PC when navailable) with the new DirectStorage Api.

PC will benefit PS5 innovations and will get similar hardware change about motherboard design, transfert rate, controllers, hardware decompression units and other things.

Still on current PC hardware, without pushing polygons to billions, you will be able to display thousand million polygons and thousand unique objects with Nanite.

But Nanite is not the whole game solution, it’s the static (and physic) geometry solution only.
Optimizations systems will reamain like

Automatic grid based world streaming , automatic HLOD and impostor generation

My feeling on all this is, not every part of a game needs this much detail. Sure it looks cool, but if a player is going through a level, how much time are they going to spend looking at a cliff face or a statue unless it’s important to the mission/story/whatever? Unless the hardware and software can do these things easily and cheaply, there’s not really a need to put that many polygons on the screen when normal maps can do it almost as good-looking.
Real-time movies or special FX are a bit different, since 4K is becoming a standard resolution, 8K is around the corner, and photo real CGI is used in a large percentage of shows these days. So maybe that’s the niche this tech would be good for.

No normal map is surface only , its’ not model shapes details.
When you watch at the PS5 demo , you see the big difference at first look, no need to discuss such level of details compared to current gen games.

Even with LOD current games can’t render as many million polygons assets in a full dynamic game scene.

The huge advantage you missed and that have been talked is to also reduce a lot production costs.
No need to retopo , no lods to make , no uv, just drop your asset in your game level.

It’s working tech like the PS5 demo, and it’s games evolution.
It’s not niche, it’s latest innovation, similar when normal maps become popular on games.
Most companies using Unreal on realistic games are very interested to use that feature.

And you did not read a lot about Nanite , as even current PC will get an insane boost on static geometry detail with virtualized geometry; because it loads on the fly from SSD and cache only the geometry that is on screen.

Even Blender does not have virtualized geometry or virtualized textures that would give it a huge boost for real time rendering.

Unreal is used for movies making also like Mandalorian for example.
Movie makers will also gain so much from beeing able to drop insane amount details assets and just render frames in real time.

Lol, this is big. Where did you read that? Maybe you meant “no normal maps”…

I understand that pretty well, thanks. While you seem to not even read what you say yourself.

How do you expect to pack (15+) layers over layers of content for a game? That doesn’t build itself magically. :laughing: Level building time will increase to unbearable heights; it’s all man made, remember? This is going to project a game creation time up to years and years on, into the field of vaporware.
Right now, AAA games can easily range 3-5-7 years of development. Think the last of us 2. So, unless UE is going to automagically autocreate an all around autoenvironment, game creation will be slow, oh so slow.
It doesn’t matter the speed of the machine, if you have to fetch 100’s of pilots into a little cockpit… :laughing:

1 Like

No it’s mixed with big automation like Houdini.

Auto unwrap angle based exist from very long time when you need UVs.
Unreal 5 will propose workflow tools, for example :

  • send Zbrush model to Maya
  • call maya decimate and unwrap on model
  • call Substance Painter calculate curvature map and bake normal map
  • call Painter apply dirt or scratches material on baked material and normal map

All automated when you don’t need to do specific work on models, read previous posts.
Big companies will be able to define their own workflow rules per objects categories , for example apply this Substance procedural for that catgeory of assets.


Houdini or custom procedural tools big companies use.

Also new AI tools able to create assets from photos or videos, perhaps there will be AI tool able to create a full city and level only using some video and producing high quality level.

Nanite is only a tool to support rendering billion triangles scenes in real time, and supporting million polygons assets.

You don’t even search yourself how it works, and you are telling Nanite is not a big innovation.

You are just making fun or Unreal 5 innovation features, and don’t seem interested, why participating in the thread ?

Ah, the mith of automatization… procedural, Houdini, and, yes… curvature maps…
it’s a lot of time they go “procedural”, it always existed. Since Rogue. It means, instead of 5 years doing manual level building, you’ll do 3 years of programming and then 2 years of manual level building, because, you know, don’t try for perfect, it’s never enough…
Then the lead programmer will be struck by lightning (or covid, who knows) and you won’t be able to find a new one before 3 other years have passed, they’re so fu%$£ing rare… and that’s the reason studios avoid them, and use UE alone with dozens of art guys…
I’m not making fun of UE… I’m just pointing out some harsh realities behind the gleaming surface…

The solutions are procedural.
Perhaps later AI image based generation from photos directly able to create geometry and re create textures materials, or photogrammetry.

Everyone knows AAA games costs, and increase with more details , this is not the concern of the thread, this is more about rendering innovation like Nanite or Lumens global illumination.

That is why engines exist, the level designer is simply using tools created by dev. teams that already put in the hard work (ie. software that the studio or the individual did not create himself). Adding Houdini to your workflow gives you a software whose procedural tools are the product of decades of intensive R&D, and not a single line of code has to be written (even though there’s quite a learning curve).

Even in Blender, you will notice that making use of procedural textures for detail, displacement, and other things does not require the artist to write a single line of code. Everything Nodes will bring such automated detailing to other areas.

Unreal 5 Nanite is just about beeing able to render extreme polycount static geometry, and simplifying workflow.
But it will not replace your hard work making a top indie game content and polishing.

Procedural tools will be a solution about storage issues.

Use of procedural generators to generate models based on very simple model shape input, for level design design using guidelines and rules; and have infinite variations, with very lightweight storage (kilobyte files) as it would be mainly math like Substance Designer.


I think this will be one solution to storage when hardware will be fast enough to generate those models and textures graphs fast enough at game loading.

Also new compression techniques innovations.
I guess such system can apply to anything like music or mesh storage.

Convincing people that optimization will be a thing of the past, by bringing up the undisputed, heavyweight champion of false promises and exaggerated claims, friggin No Man’s Sky?

Hehe. Nothing does chaos like a human brain.

It’s not about optimization but procedeural :

  • Substance Designer is the leading procedural texturing tool used by most triple A game companies.

  • Houdini is the leading procedural world generation tool also.

You are very wrong about No Man Sky, that is procedural, it got lot of hard work updates, it improved so much, it become a great game.
It’s the best and the unique on it’s huge space galaxy exploration category.

Another rendering innovation, a point cloud based rendering game, better than many Upbge games, similar to micro polygons, even if you dislike it :

Unreal 5 Nanite renders billion triangles on screen, no other engine does the same, it’s new system others will copy.

Just for you :smiley:

https://twitter.com/BrianKaris/status/1283070552618369024

I’m sure there are bits that are covered many times over such that no part of that mesh are seen at all. Normally this would be shameful sloppy env art. They should optimize it right? Well only if it costs meaningful frame time. With Nanite it’s negligible.

All presentation slides or video are just above, you really did not read anything about it.
I’m not convincing anyone :rofl:

You are not interested in Unreal 5 or it’s features while some others are interested, and it’s all right, just as much as i don’t care about some other things, so just move on and stay on Upbge discussions instead.

1 Like

Triggered much?

I’m interested in every 3D graphics engine, meant for games or otherwise, and even the ones that’d make my computer cry. What I’m not interested in is unrealistic expectations. Saying there’s no more point to optimizing assets at all sounds pretty pwretty unrealistic to me. If you don’t want to discuss the possibility that we’ll all probably still be doing retopo for years to come, be my guest.

Retopo will stay, perhaps it will become even more more automatic and better AI based.

Another way, Guerilla Games 3D artists work only only on the high rez assets and characters.
They use external specialized retopo studio to make the LOD and normal maps for their models.

You mean working realistic expectations ?

For static geometry, Nanite has been inspired by Dreams running on PS4 current gen hardware.
They share similarities and they are both able to render billion triangles in some milliseconds.
That’s why they don’t need retopo or optimization and they don’t care about number of objects.

This is new, and this will become as common as when normal maps became common.

Deal with it.

When Dreams will be available on PC , if they would bring characters rigging deformations i would be happy to make games using it; no UV, no retopo and no normal map bake, no LOD , no care about objects grouping , material drawcalls or other optimization tasks.

While PS4 Dreams or Unreal 5 billion triangles performance rendering is cool, what this brings and that matters lot more to me, is the huge simplification and improvement in workflow.

Just focus making your level design and models, forget polycount and tedious optimization work, let the engine care itself about optimization and level of detail displayed.

https://twitter.com/MartinNebelong/status/1282902161073086464

https://twitter.com/MartinNebelong/status/1270475325047177217

https://twitter.com/MartinNebelong/status/1272291738942681089

Not care about polycount on Unreal 5 or Dreams PS4,
Just make your game :
(billion point clouds rendered real time on simple PS4 old hardware, )

1 Like

It’s not like you can’t do that already (to some extent), but it’s extremely inefficient. Better tech shouldn’t translate to lazier developers, that way you’re bound to hit a bottleneck sooner or later.