What are the "Limits" of UE4?

[darthviper107]

1b) By USB do you mean flash drives?
Sorry if it is my misunderstanding…I just cant see how the USB port is becoming obsolete?
Every hardware appliance nowadays has USB jacks on one end to allow connection to power, computer, Audio equipment, etc.

-USB has undergone changes over time, there was USB 2.0 which was faster, then there’s USB3.0 and USB3.1 which again are faster. Now there’s USB C, which is faster and provides more power, and it is a smaller port which can be plugged in either direction, it’s an all-in-one port that could provide power and video signal to a monitor and can be used for pretty much anything, you can even charge your using USB C. But of course, it’s no longer compatible with any other USB ports since it’s a new shape.

2a) I see so the main trade off is…

      **UE4**

 **Pros**
 -Real time lighting
 -Can see rendered out models(with life like interactions with wind,etc) 
 as you place assets and world build 

 **Con**
 -Must do all the lighting work before hand to have realtime lighting
 -**If you don’t want to do the work…**
 -Would use dynamic lighting which
 = Doesn’t look as good

• In order for dynamic lighting to look better you would use Dynamic GI which
  as of right now, doesn’t have a high enough level of performance for use.

VS.

      **3DS/Maya**
 **Pro**
 -Building lighting labor does not need to be done beforehand

 **Con**
 -No real time lighting 
 -No rendered out models(with life like interactions with wind,etc) 
 as you place assets and world build 

-UE4 is real-time lighting once you bake it first, the process of baking takes a long time since it’s doing everything, but it allows things to be much faster after it’s done. With something like 3ds Max/Maya you have to do many of the same processing every frame so while it’s only rendering what you see, it’s rendering a lot of the same things over and over again.

**2b) ** So essentially, labor has to be done somewhere right?
(Either you…

 -Do the labor beforehand to see what your world looks like real time 

Or
-You don’t and you have to press the render button every time you want to see what your world really looks like)

-Correct

** 2c)** The only change to this process would be in the future if a dynamic GI was released that was able to produce the same quality as building up the light by hand…

Which would allow real time lighting without any of the drawbacks of labor in building the lighting, correct?

-dynamic GI would eliminate the need to bake lighting beforehand, you would be able to see your lighting instantly, plus you wouldn’t have to prepare your meshes for lightmaps which is also a pain.

3a) Are there any specific information hubs you go to keep up to date the releases and what they are capable of?

-With each release/update they have notes on what has changed—HOWEVER–keep in mind that it’s very common for a project to break when you upgrade to a new release. That’s mostly for the new version upgrades like going from 4.10 to 4.11, while there’s smaller hotfixes like going from 4.11.0 to 4.11.1 which should not break your project.
What is recommended is to only upgrade to a new version when it has a feature you really want, like say if dynamic GI was added to the engine in the future. If you choose to upgrade, keep a backup of your project in the old version just in case.

3b) Do you think the next new line of CPUs would incorporate the best of both the Xeons and the i7’s?

-No, they will keep things the way they’ve been doing them. Some of the higher end i7’s are pretty much Xeons that have some features disabled (this allows them to save on production costs) and usually Xeons are chips that have been tested for reliability. But, since they are a smaller market, the costs for Xeons are higher.

3c) I’ve read that the Quadros are industry standard more so because they have ECC’s.
Even so…

 **Quadros**
 **Pro**
 -Can improve performance in Maya/3ds Viewport
 -ECC in case of data loss

VS.

 **Con**
 -Ridiculously high price in proportion to the GeForce with similar performance**

**
Thought)** As you see, I just can’t understand how the industry standard (Quadro) is worth?
Is a slight improvement in 3DS/ Maya viewport and ECC really worth that much more money? (Since extra resources spent is multiplied by the amount of computers they have in the studio)

-I think some programmers develop simulation programs that are designed for Quadros, but Quadros also usually have more GPU memory, which can be important for things like GPU renderers where everything must be loaded into the GPU and if it doesn’t fit then it can’t use the GPU. It’s becoming less of an issue with the GPU’s with 8GB or 12GB now.
Quadros are also more reliable similar to the Xeon processors, they can stay running for a while with having less likelyhood of issues.

4) When using UE4 for…

**Only Rendering ** Making the level for - cinematic shorts ( w/High level of realism) ]

                                  **VS**

Real-time performance Making the level for
- playability as a game
And/or
-cinematic short( w/lower level realism) ]

Thought) Does one have to still be concerned with the technical problem of “making things interactive or getting workarounds for lighting and other effects?”

-Yes, that’s one of the things with a gaming engine, the tools are designed to work in real-time, so even if you don’t care about making it real-time and just want to use it for rendering in most cases you can’t just turn up the quality or realism since they don’t offer that feature–like with particle effects.

[QUOTE]
5a) Amazing thought though right?
In Unity 3’s latest cinematic “Adam,” apparently all that highly realistic rendering was real time…
If so, its possible something close to what we are imagining might actually be around the doorstep

In case anyone hasn’t seen it, here is the link. https://youtube.com/watch?v=44M7JsKqwow[/QUOTE]

-There’s a lot you can do if you’re targeting a cinematic, like with the Infiltrator demo for UE4 it has fire that they simulated in 3ds Max with FumeFX and then they developed a process to bake the fire and use it in the particle system in UE4, it looks good from a limited number of angles though and since it’s pre-simulated it wouldn’t work in a game where everything has to be dynamic, but it looks great in a cinematic.

5b) So hypothetically speaking, if game engines become able to render photorealistically, then render farms wouldn’t be necessary anymore right?

 **Because, essentially the thought process is…**

 **A)** We would like Photoreal Cinematic Renders(in which Game Engines are incapable as of yet) 
 = So we **need** to use **3DS Max and Maya**

 **B)** We want faster render times
 =So we purchase and build a **render farm**

Thought) Afterall, hypothetically speaking…

If A) (the need to use 3DS/Maya) was rendered obsolete due to engines like UE4 and Unity3 becoming photoreal in realtime…

Then B) (Render farms) would also become unnecessary correct?

-Right, but like I said, the shortcuts that games take mean that there’s a lot of things that are nowhere near the level you get rendering in 3ds Max/Maya, there’s a lot more things in those programs that are not possible to do in a game engine, like the amount of polygons, the full global illumination rendering, particle effects with millions of particles, liquids, fluids, reflections, anti-aliasing, cloth, hair, etc.

6) Regarding using UE4 to render(at a higher level than its real time counterpart), instead of real time performance…are there any examples of rendering ability of UE4 in which people have used it to create a cinematic of any kind?

Most of the photo real renderings I’ve seen are of a sofa or a single room architect’s preview for their clients.

-I haven’t seen people doing that type of thing, really if you go that option then you need to download the UE4 engine source from the Nvidia branch and compile it so that you can use VXGI. Then you can turn up the dynamic GI quality and render. I’m sure people are doing things like turning up the reflection quality and doing things like Supersampling (render at a higher resolution than your target, then scale it down which will improve anti-aliasing). But there’s other stuff you can do as well, like increase polygon counts, increase texture/lightmap resolutions, all types of stuff that you wouldn’t do in a game when you have to be able to play at 60fps on mid-range hardware.

By the way, if all you’re planing on doing is cinematics, then you don’t have to pay royalty for using UE4. If you do a game that you are going to sell then you would. But not for non-interactive media.

[anonymous_user_b0cf5855]

With each release/update they have notes on what has changed—HOWEVER–keep in mind that it’s very common for a project to break when you upgrade to a new release. That’s mostly for the new version upgrades like going from 4.10 to 4.11, while there’s smaller hotfixes like going from 4.11.0 to 4.11.1 which should not break your project.
What is recommended is to only upgrade to a new version when it has a feature you really want, like say if dynamic GI was added to the engine in the future. If you choose to upgrade, keep a backup of your project in the old version just in case.

1) So every time someone updates UE4, their projects are going to break?

Do you mean it will break in terms of…
-It wont run
**Or **
-The file itself actually breaks?

Because if its just the file itself and the project will still be able to run on the next version, then every time someone updates they could just move the project to a different HDD then bring it back right?

**geneconcept)**Regarding using UE4 to render(at a higher level than its real time counterpart), instead of real time performance…are there any examples of rendering ability of UE4 in which people have used it to create a cinematic of any kind?

Most of the photo real renderings I’ve seen are of a sofa or a single room architect’s preview for their clients.

**)**I haven’t seen people doing that type of thing, really if you go that option then you need to download the UE4 engine source from the Nvidia branch and compile it so that you can use VXGI. Then you can turn up the dynamic GI quality and render. I’m sure people are doing things like turning up the reflection quality and doing things like Supersampling (render at a higher resolution than your target, then scale it down which will improve anti-aliasing). But there’s other stuff you can do as well, like increase polygon counts, increase texture/lightmap resolutions, all types of stuff that you wouldn’t do in a game when you have to be able to play at 60fps on mid-range hardware.

By the way, if all you’re planing on doing is cinematics, then you don’t have to pay royalty for using UE4. If you do a game that you are going to sell then you would. But not for non-interactive media.

**2a)**Is the only dynamic GI available right now is through the VXGI?

---

**2b)**What exactly is the difference between the VXGI and the unreleased SVOGI by UE4?

There must be a fundamental difference, I’m sure, since one GI (VXGI)was released and the other GI (SVOGI) wasn’t?

For example…maybe VXGI builds light worse than the SVOGI = better performance = why it was released to public?

---

3) Is there a name for the method of using UE4, only for rendering, to boost up the quality for cinematic production?

---

4) When it was mentioned of using UE4, only to render…

Previous message) Right, if you want to use UE4 to just render stuff and don’t care about real-time performance then it becomes much easier, you can render things pretty nicely and there’s quality settings that you can turn up if you don’t care so much about it being real-time.

There’s also stuff like Nvidia VXGI which is available for UE4 if you want to try a dynamic global illumination system that you could use and turn up the quality if you don’t are about framerate.

How does this recent message…

Recent message) …in most cases you can’t just turn up the quality or realism since they don’t offer that feature…

…fit together with the original/previous message?

They seem to be complete opposites of each other, but I believe you have a reason for both messages, and that I am the one misunderstanding the messages somewhere along the way.

[Vi]

If you worry about a huge world to use in a game, the answer is : it depends on how UE4 can stream this world.

I am not sure but I think the UE4 can stream an open world, so this is possible.

[darthviper107]

1. So every time someone updates UE4, their projects are going to break?

Do you mean it will break in terms of…
-It wont run
Or
-The file itself actually breaks?

Because if its just the file itself and the project will still be able to run on the next version, then every time someone updates they could just move the project to a different HDD then bring it back right?

-It might break, no guarantee that the project will work if you download a new version and open your project with it. Once a project is converted to the new version you can’t go back so keep a copy of the project at the old version just in case.

2a)Is the only dynamic GI available right now is through the VXGI?

-The only really usable solution is VXGI, UE4 has another method called LPV but the quality is very poor and the performance isn’t very good either and you can’t turn up the quality as high as VXGI.

2b)What exactly is the difference between the VXGI and the unreleased SVOGI by UE4?

There must be a fundamental difference, I’m sure, since one GI (VXGI)was released and the other GI (SVOGI) wasn’t?

For example…maybe VXGI builds light worse than the SVOGI = better performance = why it was released to public?

-They use similar methods, but from what I gather VXGI is a bit faster. The reason SVOGI wasn’t included was because they were using GTX680 graphics cards for it to work, which most people don’t have something like that. Also, it doesn’t work on game consoles (PS4/Xbox One) which all means that if they wanted to keep the feature in the engine then it would only benefit a small group of people–basically PC gamers with high-end graphics cards.
VXGI is developed by Nvidia, and it’s not fast enough either, but Nvidia made it available anyway. I don’t know of any games that use VXGI.

3. Is there a name for the method of using UE4, only for rendering, to boost up the quality for cinematic production?

-It’s already doing what it can, I mentioned before some things that can be done to improve quality–higher quality assets and the few settings that you can turn up.

4) When it was mentioned of using UE4, only to render…

How does this recent message…

…fit together with the original/previous message?

They seem to be complete opposites of each other, but I believe you have a reason for both messages, and that I am the one misunderstanding the messages somewhere along the way.

-People don’t turn up the graphical features as high as they can, but once you do there are limits to the quality available. For example, you can’t increase the resolution of reflection probes, and anti-aliasing quality only goes so far. I think even GPU particles are limited to 1,000,000

[anonymous_user_b0cf5855]

Hey

I’m really grateful to you for essentially assisting in making this seemingly unreachable dream of mine, into one where I have a solid plan and foundation.

In response, I’ve come up with a hybrid solution based off of the knowledge garnered within the thread of discussions

**Based on everything we’ve discussed this is the ideal solution I’ve arrived at, hows this sound? **

---

                                                            **Hybrid Solution: UE4 and Maya/3ds Max, The Best of Both Worlds **

             **Goal**

A) Lifelike renders for cinematics
B) Explorable levels based off of Sets(& Vice versa / Sets based off of explorable levels)

---

                                                                                            **Hybrid Solution: Idea 1** 

Steps

1) Create explorable level/city with UE4…
Goal B) accomplished
…& utilize dynamic GI (VXGI for now/use stronger dynamic GI in the future)

**=**No need for labor with building light and light maps,etc

2) After creating city level, import (scenes/“parts of the level” or “set”) into 3ds Max/Maya.
(Essentially, selecting parts, of the cities’ levels created in UE4; and bringing them into Maya/3DS as the virtual “set” to film on)

**=**Utilize rendering ability of 3DS/Maya to achieve photoreal renders that UE4 can’t do.
Goal A) accomplished

---

                                                                                    **Hybrid Solution: Idea 1 Inverted** 

Thoughts behind this idea) If the above stated hybrid method is possible…then what about the opposite?

Steps

1. Create virtual “set” (or parts of the city where film is shot) in Maya/3DS to film cinematic (this set will have areas of empty 3D space because only specific camera angles needed for shot, similar to a CGI movie workflow)

**=**Using 3DS/Maya to get photoreal
Goal A) accomplished

-After finishing cinematic…

2)Export all the finished cinematic’s sets from Maya/3DS,
into UE4 and unify the sets (by filling empty 3d space w/architecture, nature, landscape assets) to make an explorable level in UE4.
Goal B) accomplished

---

THUS, in combining both Steps 1) and 2) in both the Idea 1 and Idea 1 Inverted

-We have a back and forth system that allows for high quality renders for cinematics…
whilst also not letting the hard work that goes into making the sets for the film go to waste!

(Afternote) “Waste” used here is based off…
-Since many movies only create the parts of the set that the audience can see,
the rest is empty 3D land and for the most part never used again.)
Which = “waste”

[willmypower]

I think the “limits” is not on UE4,but on the hardware.
current CPU,GPU,RAM and disk’s speed and volume can’t support your idea.
But if regardless of hardware problems,the huge world can be created.

[Nawrot]

I think unreal for single person is limitless, ie. you never produce enough to exceed longint size limitations.
I once estimated sizes and scales for earth (to see maximum scale of planets in accurately represented solar system), where i assumed 1au equal to half of size of grid area.
Longint makes possible to put 20000 x 20000 x 20000 cubes that are size of grid.
So solar system made to the scale of 2au = grid size was still quite small compared to whole longint space.

There is small problem however, that not everything in unreal works outside grid, i think physics and replication start do weird things. Also animation start to have rounding errors.
But you have lifetime you can learn C++ and improve (or fix those places, get them merged into main branch etc). So assuming you do not want to exceed 32bit limitation, you still can do huge world. Also because games like elite dangerous (i think) implemented 64bit, there may be reason for Epic to consider this (maybe not necessary, but good marketing point). So there is big chance that at some point coordinate system of unreal goes 64bits everywhere. That probably would make possible to create world that spans light years in size (when you keep my test scale of 2au=grid size).

Another estimation to visualize scale:
earth surface is 500 mil square km. Assuming you make square of it, it will be ~22400x22400 km (same surface as earth surface on ze ball).
now integer is from −2,147,483,648 to 2,147,483,647.

If you take 1uu = 1cm (like Ue4 now assumes) you get 1.9173 of earth squares side by side in longint of coordiantes.
That makes whole grid 3.83 times of earth surface when you do 1uu = 1cm.
But it is not all, you probably could stack layer of such surface one above another and turn them on/off with script, or LOD).

So there is no chance you can ever reach limit of filling up single unreal map to the max. Realistically your limit here is your PC, sdd size, ram size etc.
But then you can do levelstreaming which is only limited by size and speed of hdd.

I am quite sure that doing it alone or even with 100 of other people, you will fill up that world slower than hardware power grows and slower than epic improves engine.

If i was you I would heavily dig into proceduraly generated maps, its bit harder than making it all manually, but then you can truly do infinite worlds.

[Kinos141]

“Is it possible, for example, to run a map the size of the United States on Unreal Engine?”

Yes. You have to use certain tricks to make that happen.
Remember, most video games are illusions. Normal maps make low poly meshes look high poly. Occlusion culling only shows the player what he’s looking at and hide the rest to save memory(or something like that.)
Most likely, you will use Steaming levels to load certain chunks of game depending on where the player is.

The real question is WHY would you do that? A game that size would be full of empty space, which is boring.
Personally, I’d take a smaller, busier map of Yakuza game over the huge, empty map of GTA5.

Edit: I don’t think the size of the team is much of a factor for UE. There are many modules that make automating processes simple, i.e footsteps. You can take a single footstep sound, modulate it and randomize it via audio cue to create several sounds that would have taken longer in another app.

[anonymous_user_e168914d]

The limits are not in unreal so much as they are on the developer. But, You might have to rip out parts of unreal from time to time, and throw them away in favor of developing something that is scalable. Scalabel is NOT something that unreal is. And architecturally, it is not something it will ever be. If you want to make the next great MMO then you will have a lot of work ahead of you replacing unreal stuff, or perhaps wait untill Unreal 5 comes out. Which should actually be getting discussed at epic if they want to stay on top of the industry.

[Nick_Calabrese]

The importing and exporting capabilities are really up to you. I don’t know of any 3D software that exports levels to engines by default (although they all have object exporting capabilities) and certainly the engines themselves don’t have any exporting capabilities, but every engine has importing capabilities. Technically you could just treat your entire level in your 3D software as one object, but on import this would be a nightmare for several reasons; one is that unreal by default creates collision for you, which is nice, but it does it per object so your collision won’t be even close to usable, and this would be disgusting to deal with for even a small level, let alone a giant level like you’re talking about.

The solution is to write your own importers and exporters. I use blender and unreal together. I wrote a script for blender that export levels for unreal, and a script for unreal that imports and loads levels made in blender. The ideal way to do it is to have your exporter create a custom scene description file (mine just has the obvious transformation stuff (location, rotation, scale) for the specified object) and exports each object that gets instanced at least once in blender. Then the unreal script reads the scene descriptor, imports the necessary objects, and creates each objects instance according to the transformation in the scene description file). This makes later adjustment nicer. There are ways to cheat this that are a lot less optimal but are easier (although this isn’t too much work anyway).

One upside to this is that you do almost all 3D stuff in your software, and so if you do decide to use a different engine, all you need is a new importer that can read your scene description file.

Like other people have said, this project is pretty unrealistic, so I would definitely plan on just making a city (which is already a big task). I think the idea of just doing it gradually over your life is a good mindset. Games take a long time to make, but a little bit every day goes a long way, and when it starts to take shape and look like what you envisioned, it’s a really cool feeling.