Unreal Engine 5, [UE5 Developer Availability 2022-04-05]


Not unexpected with an engine switch...

Shifting to a new engine is a big undertaking, so we want to be clear that we will not be announcing any new projects or titles for some time.

So, I'm not expecting anything new out of them for at least 2-3 years. Considering how extensively they modified UE4 to have it run as well as it did within their requirements for Gears 5, I'd be curious about whether they'll have to modify UE5 just as extensively or if UE5 has incorporated some of the modifications The Coalition made to their fork of UE4.

Regards
SB
 
It's very risky, as you never know how many issues UE5 will run into before it is ready to release games with, but it is also very important for Epic to have companies like this moving to UE5, who can not only find issues in UE5, but come with fixes and improvements themselves, to get the UE5 Engine to where it needs to be for others.
 
Not unexpected with an engine switch...



So, I'm not expecting anything new out of them for at least 2-3 years. Considering how extensively they modified UE4 to have it run as well as it did within their requirements for Gears 5, I'd be curious about whether they'll have to modify UE5 just as extensively or if UE5 has incorporated some of the modifications The Coalition made to their fork of UE4.

Regards
SB

Wouldn't be surprised if Epic would be happy for them to merge changes and additions to UE5, they certainly did a good job with Gears 5, and thus the seemingly early access for them.

They also don't seem to be working on Gears for their next project. While I can't find it now, just a few days ago I spotted some apparently official concept art for an "insert random project code name here" by an artist from The Coalition. Some very decidedly not Gears super colorful and out of this world trees, guess they OKed releasing it under the project codename as it's just early exploration concept art you can't make much heads or tails of. Interested to see what they'll do next though, Gears 5 was surprisingly fun.
 
Wouldn't be surprised if Epic would be happy for them to merge changes and additions to UE5, they certainly did a good job with Gears 5, and thus the seemingly early access for them.

They also don't seem to be working on Gears for their next project. While I can't find it now, just a few days ago I spotted some apparently official concept art for an "insert random project code name here" by an artist from The Coalition. Some very decidedly not Gears super colorful and out of this world trees, guess they OKed releasing it under the project codename as it's just early exploration concept art you can't make much heads or tails of. Interested to see what they'll do next though, Gears 5 was surprisingly fun.

Apparently, the next project is an experimental side stuff to help them acclimate to UE5 and explore its systems since they are Xbox's UE experts. Makes sense too. This will help them get used to the engine and maybe develop a cool new IP in the process before going full steam into Gears 6. Anyway here is the clip from Jeff Grubb, a journalist from Gamesbeat who seem to know a lot about Xbox in general:
 
Apparently, the next project is an experimental side stuff to help them acclimate to UE5 and explore its systems since they are Xbox's UE experts. Makes sense too. This will help them get used to the engine and maybe develop a cool new IP in the process before going full steam into Gears 6. Anyway here is the clip from Jeff Grubb, a journalist from Gamesbeat who seem to know a lot about Xbox in general:

Really looking forward to what it may be, and i think it may be good for the team as well. They only developed gears for now, and to work on something else and unload ideas that are not constrained to the gears universe can be good and refreshing.
 
Really looking forward to what it may be, and i think it may be good for the team as well. They only developed gears for now, and to work on something else and unload ideas that are not constrained to the gears universe can be good and refreshing.

Microsoft also has a lot of shooters now , Halo , Gears , Quake , Doom , Wolfenstein and rage off the top of my head. So they can let Gears rest for a few years
 
My realistic interpretarion of the dev's comment about "unloading" assets behind the camera is:
Actual streaming engines don't exactly "unload" stuff. By that, I mean, they don't wipe data out of memory. They tag it as replaceable and it gets substituted by new data when it is needed. Most engines probably have multiple priority levels of how needed and uneeded each asset is so it can prioritize better.

So my guess is, in this game, the SSD is reliable enough that they can knock assets out of the view frustrum a couple levels down in their priority level, and that means that in some extreme cases where there is an absurd amount of asset variety in view at a time, some of that stuff ends up getting replaced, but that is probably not that common, and does not happen to ALL data out of view at a time.

But condidering their portals, and how you can see parts of multiple wildly different levels simultaneously in this game, a View-Frustrum aware streaming system makes much more sense. Say, for when a player aproaches a portal, it prioritizes streaming in the assets that are directly visible from that portal's opening, and delay the stuff behind (that would only be visible after the player character GOES THROUGH the portal)

They will not discard everything because they would need 10 TB game size to be able to do this and maybe faster SSD with tons of unique assets. I think in a few years the biggest constraint will be the size of the game on storage.

Another interesting game will be God of War Ragnarok if the game is exclusive to PS5. Some Sony Santa Monica devs describe what they wanted to do for the first fight against if they were not limited by the HDDsaid they would have want each big attack send at high speed Kratos in the air with big destruction and so far he reachs another realm.
 
They will not discard everything because they would need 10 TB game size to be able to do this and maybe faster SSD with tons of unique assets. I think in a few years the biggest constraint will be the size of the game on storage.

I disagree here. I think the biggest constraint will be rendering of those assets. The PS5 has a finite GPU like all hardware. Being able to put more assets into VRAM for the GPU to crunch on will reach a limit of how fast (and how complicated) can it render in a given amount of time. It does no good to throw a trillion transparent triangles, for example, if the GPU can't light and shade them fast enough for the given render target.
 
They will not discard everything because they would need 10 TB game size

How would they create that much game data in the first place.

I think in a few years the biggest constraint will be the size of the game on storage.

I think the biggest constraint will be the size of development team. Can we expect more procedural creation of assets, is that a thing?

Another interesting game will be God of War Ragnarok if the game is exclusive to PS5. Some Sony Santa Monica devs describe what they wanted to do for the first fight against if they were not limited by the HDDsaid they would have want each big attack send at high speed Kratos in the air with big destruction and so far he reachs another realm.

I can't wait for PS5-specific gameplay design because of the fast ssd and i/o.
 
I disagree here. I think the biggest constraint will be rendering of those assets. The PS5 has a finite GPU like all hardware. Being able to put more assets into VRAM for the GPU to crunch on will reach a limit of how fast (and how complicated) can it render in a given amount of time. It does no good to throw a trillion transparent triangles, for example, if the GPU can't light and shade them fast enough for the given render target.

Wait when we will have the detail of the size of the Unreal Engine 5 demo. With rock assets of millions of polygons, 8 K textures, 16K shadow maps. There is cleary enough power to light and shade this demo. The quality of assets is not a limit anymore . If it means the game will be 1440p or 1080p reconstructed at 60 fs with a deep learning reconstruction tech this is ok. This is a developer choice. They can and for a long time improve the rendering but Brian Karis of Epic said the UE 5 quality of assets is not realist for a game because of assets size on the SSD, This is a first demonstration assets size on disc will be a limit.

In 10 years, GPU will be able to render the same assets at a higher level of fidelity but out of new miraculous storage solution, it will probably be impossible to create a game with this quality of assets. Storage size limitation is a reality since this demo. From an asset perspective out of special case like hair rendering or vegetation rendering this demo reach polygon per pixel and a 1 texel per pixel for the scenery you can't go further. Some engine begin to use hair spline technology using compute with analytical AA on Dice side. From an asset point if view, the gap is huge between PS4/XB1 and PS5/Xbox Series.

Currently Nanite does not support transparent assets and animated assets but they think solution exist but hair and vegetation are out of scope and impossible to render with this technology.

There are some big, obvious features we can see in terms of the benefits of the PS5, like the fast load times or the rifts that pull you into a parallel world immediately. But are there any examples of smaller, less obvious things that are cool or that you’re really proud of that wouldn’t have been possible on the PS4?

With the SSD, it’s easy to say there are no load times, and look how fast we can load this other area, but it has all sorts of knock-on effects. We don’t need to be as careful with how we package our data. All of the assets for an area don’t need to be collated on the spinning hard drive to get the right streaming speed out of it. It makes the game smaller on your hard drive; it means we can patch it more easily.

Currently the game are the same size than last gen but I doubt it will stay like this for a long time.

And people need to understand Ratchet and Clank Rift Apart is a first year exclusive PS5 game like every generation we will see better later.

More polygon, better geometry level of detail solution, realtime GI solution and so on but assets size will begin more and more to be THE limit. This is only the beginning of the generation.

EDIT:
The biggest interest of the SSD on next gen console is on the design side it means if you want to create a portal this is possible, if you want a game where the player can travel at high speed with high quality graphics this is possible. The RAM size and streaming speed aren't a limit on what you want to design.
 
Last edited:
Wait when we will have the detail of the size of the Unreal Engine 5 demo. With rock assets of millions of polygons, 8 K textures, 16K shadow maps.

The UE5 demo has only the triangular detail. The lighting/shading is still relatively cheap compared to a full RT pipeline. 8k textures are great if they have a lot of variation in them. 8k textures with very little variation (like that UE5 demo) isn't going to push photorealism. 16k shadow maps can't be streamed in from SSD. They have to be computed on the fly. They are not an asset like textures and geometry.

There is cleary enough power to light and shade this demo. The quality of assets is not a limit anymore .

Of course the quality of assets are a limit. Once they are in VRAM, the GPU has to render them.

In 10 years, GPU will be able to render the same assets at a higher level of fidelity but out of new miraculous storage solution, it will probably be impossible to create a game with this quality of assets. Storage size limitation is a reality since this demo. From an asset perspective out of special case like hair rendering or vegetation rendering this demo reach polygon per pixel and a 1 texel per pixel for the scenery you can't go further.

1 triangle per-pixel isn't the upper end of quality of rendering. We still have several orders of magnitude of sub-pixel rendering during calculations. I would never clamp my rays cast to 1 per pixel. That leaves me very little accuracy to model plausible lighting, tessellation and shading - all of which should be constraint only by the precision of double values. I'm used to the quality of 16-32 rays per pixel to start rendering.

In short, we still have a long way to go.
 
Im not sure whats going on here, seems they are teleporting elsewhere, but some things are the same (or very close) in the new scene in the foreground/background, so perhaps not the best example.
I disagree here. I think the biggest constraint will be rendering of those assets. The PS5 has a finite GPU like all hardware. Being able to put more assets into VRAM for the GPU to crunch on will reach a limit of how fast (and how complicated) can it render in a given amount of time. It does no good to throw a trillion transparent triangles, for example, if the GPU can't light and shade them fast enough for the given render target.
No GPU in 10 years will be able to render a trillion visible triangles in a frame at 60fps.thus LOD is needed, could the above R&C city scene vista contain a trillion triangles? Yes it could (perhaps it does?)
IMO, the best thing that has happened this gen is that devs are not chasing for native 4K. Upscaling has come a long way and in some titles it can really provide crisp final visuals. We are even living in a world where Returnal devs intentionally chose 1080p over 1440p "because it looked better" for the grimey and unsettling artstyle of the game. Lower resolution was the "artistic choice".
I agree forcing yourself to go 4k is a terrible decision, though I dont believe they went for 1080p because of artistic choice but performance, you could argue 1080p @ 60fps looks better 'artistically' than 1440p @ 40fps and I would prolly agree
 
The UE5 demo has only the triangular detail. The lighting/shading is still relatively cheap compared to a full RT pipeline. 8k textures are great if they have a lot of variation in them. 8k textures with very little variation (like that UE5 demo) isn't going to push photorealism. 16k shadow maps can't be streamed in from SSD. They have to be computed on the fly. They are not an asset like textures and geometry.



Of course the quality of assets are a limit. Once they are in VRAM, the GPU has to render them.



1 triangle per-pixel isn't the upper end of quality of rendering. We still have several orders of magnitude of sub-pixel rendering during calculations. I would never clamp my rays cast to 1 per pixel. That leaves me very little accuracy to model plausible lighting, tessellation and shading - all of which should be constraint only by the precision of double values. I'm used to the quality of 16-32 rays per pixel to start rendering.

In short, we still have a long way to go.

Did I say we don't have a long way to go on a rendering perspective?



Again the geometric level of detail inside the demo is too big for a game not because they can't render it but because the size of the data is too big. And Brian Karis told performance is not a problem and read the tweet thread he created, he explains performance is improving because this is a new technology. For example Nanite took around 4,5 ms on the demo since then it is less than this.

And out of a miracle from a storage point of view, this will be the case in 10 years. The rendering will improve but it will be impossible to use assets of this quality in a game.

You can take the same assets for scenery and probably can do in 10 years a much better demo from a visual point of view but at the end the problem will be the same not enough place on the storage.

I think what most people are point but not quite able to put into words in the on rails section shown is that, even though that scene takes place in the same level as you're showing in this screenshot, the fact that it's on rails has some advantages because it is constrained. For example, if you place the camera/gameplay on a rail, with a fairly fixed speed, you know exactly what parts you need to load in and when. In the gameplay area in your screenshot, you can move in any direction at variable speed. I'm not saying they need to constrain the game to an on rails section for technical reasons, but it would certainly simplify the IO to know what assets and at what LOD you need them and when, and to be able to discard what you know you will no longer need. I doubt you could even turn the camera around during that section. Again, for gameplay reasons you wouldn't want to.


Hold my beer. Also, size is not always of quality.

And you will be able to do the same in an open world because the I/O is not the limit anymore. The problem you talk about are an engine problem. The I/O has nothing to do with this. This is not the I/O system taking the decision of the data needed to be loaded for the next quarter second. I take a tweet by Fabian Giesen from RAD Game tool company behind the compression technology oodle kraken and oodle texture to illustrate a problem with the more realistic figure of 10 or 11 GB/s you can go through the complete data of a 200 GB games in 20 seconds.


For the 10/11 GB/s figure it comes from RAD tool games
http://cbloomrants.blogspot.com/2020/09/how-oodle-kraken-and-oodle-texture.html

Why do they need 10GB or 11 GB/s on PS5 because it is for extreme case when you need to change the vast majority of assets in memory like with a portal or the example I gave about god of war. Same lead gameplay developer said in EasyAllies video preview than it is the first time he did not ask to cut from the game some ennemy type because of RAM and streaming constraint in a game. This is exactly like in the Road to PS5 video presentation. It helps visually with much better assets quality but the main benefit is from a game design perspective. Devs can do portal, can go very fast without loss of assets quality in the game, they can do great transition realtime and cutscene game easily. They can do realitme flashback cutscene in at totally different place without problem. It will be easy to do interior/exterior transition without doing like in Spiderman a cutscene for load the data. I expect transition between different biomes to be more progressive and more natural. And probably tons of other things I don't think about.

Easy Allies video preview

When I talk about game size problem this is not define by the number of assets but the size of each individual assets like the statue in the UE 5 demo. You can do a full Open world game not bigger than a game today but the assets quality will be so high it will take TB of data probably more than the SSD inside the PS5 and Xbox Series consoles.

And Nanite reduce the job of artist if you can import the Zbrush model directly in the game engine.
 
Last edited:
I wouldn't say its a certainty that these consoles can handle the level of throughput shown in the UE5 demo in an actual game.

It is a certainty they can't because the SSD in the console are probably not big enough to sustain a full game with this quality of assets without any compromise.
 
I wouldn't say its a certainty that these consoles can handle the level of throughput shown in the UE5 demo in an actual game.
It's a very efficient LoD and rasterizer which doesn't bog down hugely with small polygons.
Actual rendered polygon amount is not that high, few millions. (The source polygon amount is meaningless metric.)

If they would render porous materials or objects with dense highly variable geometry, then the geometry amounts would be quite different. (possibly tens or thousands of polygons in pixel.)
Hopefully they will introduce rasterizer/tracer for those objects as well.
It is a certainty they can't because the SSD in the console are probably not big enough to sustain a full game with this quality of assets without any compromise.
Rendering side certainly is the 'easy' part in this.
Certainly will be interesting to see how large the objects will be in storage.
 
Back
Top