I'm not sure if anybody else has heard, but I saw this on the Unreal Engine Facebook page. What do you guys think? is this for this gen as well? or a set up for next?
It would be interesting to hear what will be different, will it directly replace Nr 3 even for the free SDK and how will it be different anyway. I guess we have to wait.
Epic wants to start signing up licensees for the next gen.
I'd be surprised if it ran on current gen console hardware. I'd also be surprised if it was all that different from UE3, or if it gave you any sense for what the next gen consoles will be like. There will be a lot of assumptions built in, and it will run on experimental hardware. Expect it to be a beefed up UE3, with a cool demo. They'll do behind closed doors demos at GDC, then put the video up online after GDC.
We don't even have proper access to Mips and LOD's for textures in the engine, let alone cubemaps, I really doubt they should even be thinking about this when we can do Mip levels in Max at DX9 level.
Given the amount of research and development going into it, I'm assuming it'll be a fair bit more than a beefed-up, more efficient UE3. They say they will reveal it later this year (with what I'm assuming will be a demo video), but the actual engine probably won't be in use for a good while after that (I'm guessing at least a year or two after the reveal, which puts the performance target in 2014 territory). UE4 is bound to be different enough to warrant being an entirely new engine, likely with entirely new approaches, and not just an update to what UE3 is doing (they actively update UE3 itself and have reached Samaritan quality with it). I sort of hope that we're entering into SVO territory (which is feasible given the growth of interest and what id's discussed), but who knows.
We don't even have proper access to Mips and LOD's for textures in the engine, let alone cubemaps, I really doubt they should even be thinking about this when we can do Mip levels in Max at DX9 level.
I dunno, why wouldnt they want to promote it to the community? That would be wierd, considering they are gicing the latest UDK out to the community. They WANT every artist to know their engine, its good business, its working out pretty well for them I would say : ))
I remember a while ago I read somewhere that they were going to do a full rewrite of the next iteration of the unreal engine and that wouldnt be released till like 2014 or so .Either way I look forward to this .I got my start playing unreal tournament 2003.Great game .
Hmm, I thought all these incremental UDK updates were, technically, Unreal Engine 4? What else could they possibly have tucked away that needs a whole new version number for it? Fancier SSS? (and if following recent trends, add about 15 more S's to that.)
Hmm, I thought all these incremental UDK updates were, technically, Unreal Engine 4? What else could they possibly have tucked away that needs a whole new version number for it? Fancier SSS? (and if following recent trends, add about 15 more S's to that.)
Well if the samaritan demo was any indication, they are focusing on next gen and dx 11 with the new engine. Its supposedly built from the ground up and not an expansion on udk or unreal 3. Tessellation + Reflections + Film Effects +
Subsurface Scattering + Better AA + Deferred rendering + glossy reflections + stuff we haven't seen yet, hopefully. I wonder if they are throwing light baking completely out the window, I kinda hope so, and I'm also wondering if they are doing any megatexture features like in rage for terrain. Maybe we will so some cyrsis like real time GI.
and i don't see a reason to keep the new Engine from the public as a free dev kit. They went very well with it to build the supply for experts on the engine, build the interest on the engine and the amount of sold full licenses and indie licenses. Iam not seeing a reason to make people familiar with it from the start. Well If it's as big as promised Crytek might be having a problem...
i dont see how much more intuitive a program can be... where could they possible improve from ue3
Tightly integrated and friendly cross development with Max/Maya! ( opportunities to iterate at the very first stages of dcc content authoring. ) http://area.autodesk.com/blogs/stevenr/project_skyline_at_siggraph
Also Visa versa!
would be aweome if integration would allow for, custom Maya python solutions Like IK blending for character navigation of unique terrain. ( perhaps even leveraging my 12 years of previous dev libraries and aging talents ) :poly136:
Watching project Skyline examples of tight Maya UDK intergration has me as giddy as a little schoolgirl. ( I really wished art to engine never died! I think during Maya 2.5 - 3 ish? I still have some early promotional cd-Roms if anyone is interested in some ancient Mayan artifacts ).
I will likely fo my body more harm ( inadvertantly spazing, or holding my breath till I pass out ) If they do not follow up with Project Skyline at GDC 2013 )
UE4 isn't an incremental iteration (like the UDK updates are). Like UE3 to UE2, it'll feature a new rendering pipeline geared towards next generation console hardware - I'd also expect that other core libraries may well have changed (such as physics and audio), but at this point in time can only speculate on which.
Some fundamentals in the engine have changed; you'll already notice tiny snippets of the 'Kismet 2' libraries have fallen into UDK as minor references. Although I do not know yet if it is the case, I'm expecting to see a more library driven approach (such as runtime hot-switchable DLLs), which may suggest that UnrealScript may be wholly deprecated.
Other than the above speculation (and this is speculation), I'm contractually obliged not to say what I already do know.
@ambershee why would they need tp prepare for the next console generation, last time i checked, the average computer from today will have way more rendering power behind it than the xbox 720 when it releases.
no new console tech to prepare for just what we currently have with slightly more power behind it, since it seems the new xbox and PS will be based on what is considered to be fairly old tech that is already being used in the average computer.
You don't know anything about the next hardware generation (and those of us that might are contractually obliged to not say anything); in any case it makes perfect sense to concentrate on building a stable development platform that supports the appropriate major hardware. Strong, flexible out-of-the-box solutions with first-class tools means better games.
please, take a cue from crytek's lighting and do away with lightmaps.. oh pretty please.
Crytek's dynamic lighting has issues - it's fiendishly performance hungry for a start, and seems to exponentially deteriorate with increasing number of light sources. Their shadowing on the other hand is pretty sharp. Lightmaps in UDK under Dx9 are essential, especially on console hardware, as the lighting is forward-rendered (like CryEngine) and costly. Under Dx11 it's deferred and signficantly better until you start using shader lights, then it's back to forward rendered again (but how many of you use shaders on lights?)
It seems like fancy shaders and vertex blending is the future, but then again it seems like deferred is the way to go, it'll be interesting to see what wins it out.
We don't even have proper access to Mips and LOD's for textures in the engine, let alone cubemaps, I really doubt they should even be thinking about this when we can do Mip levels in Max at DX9 level.
Id like to think the biggest reason they would be working on a completely new engine is to drastically change the engine design and business model, creating more of a Unity clone than the old "internal use only" engine design they did in the past. Along with a better engine/api design they could sell cheap licenses similar to how Unity manages licensing and completely do away with the need for c++ level source access, while implementing all or most of the advanced graphical features of the current version of unreal, and of course opening up the engine/api on a low level so you can implement your own tech as needed.
Thats just wishful thinking though and I really doubt they will attempt to address the api access and script execution speed issues
Id like to think the biggest reason they would be working on a completely new engine is to drastically change the engine design and business model, creating more of a Unity clone than the old "internal use only" engine design they did in the past. Along with a better engine/api design they could sell cheap licenses similar to how Unity manages licensing and completely do away with the need for c++ level source access, while implementing all or most of the advanced graphical features of the current version of unreal, and of course opening up the engine/api on a low level so you can implement your own tech as needed.
Thats just wishful thinking though and I really doubt they will attempt to address the api access and script execution speed issues
If they were to take some cues from Unity (and they had it available to everybody, like UDK) I would LooOoOOoOOoOove to be able to make browser builds.
It seems like fancy shaders and vertex blending is the future, but then again it seems like deferred is the way to go, it'll be interesting to see what wins it out.
Nice! Looking forward to it.
I really hope they have better single player support (from scratch) and improved the AI.
I also hope they have it in full real time so that we don't have to build lighting or
brushes.
I don't think there are haters, just people with bad experiences. I mean it was made pretty clear from epic, that UE4 won't be running on X360 or PS3 so why bother about their technical limitations? We all know the new generation is coming, we all know roughly what feature set it will have (atleast DX11 support as far as it's known right now), they will be out 2014 round about so it makes sense to get the engine out this year, assuming the dev cycle for a game 2-3 years and you want to be in first line of releases.
I guess you could have both, I was just thinking about how limiting of a material editor the cryengine 3 has (compared to udk), and I've heard its a lot more expensive to have a fancy shader in a differed engine.
I think lightmaps are still going to be around for a few generations. You can't quite get the same amount of detail and lighting complexity with realtime compared to offline baking as well as being cheaper.
Yeah I know - it was a reply to "get rid of lightmapping" posts. I believe realtime lighting still has a long way to go when it comes to proper area lights and light distribution in general. You can have rich ambient and reatime direct lighting/shadows at the same time, so no need to throw away ligthmaps just yet.
i think ue4 will mainly be geared to the next gen of consoles. I dont think it will be THAT much better than the current versions of UDK. Especially considering the DX11 capabilities.
I have to say I do somewhat enjoy having lightmaps. Compression artifacts often give that subtle dirty look, while low res lightmaps sometimes produce happy accidents and create some cool gradients.
I have to say I do somewhat enjoy having lightmaps. Compression artifacts often give that subtle dirty look, while low res lightmaps sometimes produce happy accidents and create some cool gradients.
I've been noticing this in the snow scene I recently started. In areas where it's just snow over a fairly flat surface the lightmaps give it some strange reflections that make it look like certain parts are iced over which is cool.
EDIT: And just as I post that UDK starts lighting my shit super dark for no reason.
I'm talking about the unintuitive way a MaterialInstance and Materials themselves, are setup, not really Kismet or Artists friendly at all if you want external access for certain real-time modifications.
So that if you want to use LOD's and Bias's of a texture for say SSS because you don't have access to DX11 and want to custom tailor your own stuff, ceramic and metal light bleeding, it's next to impossible, not mention if you do a little HLSL which requires certain nodes, you're later on unable to access them through any instance.
Also, considering your able to do ALL this in Viewports for modeling packages like Blender, Max and Maya, at DX9 level, I find it strange the way all this was dropped and introduced only in DX10, but you had to know some HLSL to access certain functions.
But now that DX10 was dropped, and DX11 is the only other renderer they have, it really leaved me peeved out on just how much of the 'basic' stuff which should be readily accessible to people is closed off and so limited, even if one buys the engine, unless they have a programmer who knows C on their hands, chances of you getting back much of the closes stuff it still hard.
My hopes is that UE4's entire shader systems kinks are worked out and made more transparent. Seriously, accessing something as simple as LOD's of a 512px texture shouldn't be a hidden feature, especially for Cubemaps.
I'm talking about the unintuitive way a MaterialInstance and Materials themselves, are setup, not really Kismet or Artists friendly at all if you want external access for certain real-time modifications.
So that if you want to use LOD's and Bias's of a texture for say SSS because you don't have access to DX11 and want to custom tailor your own stuff, ceramic and metal light bleeding, it's next to impossible, not mention if you do a little HLSL which requires certain nodes, you're later on unable to access them through any instance.
Also, considering your able to do ALL this in Viewports for modeling packages like Blender, Max and Maya, at DX9 level, I find it strange the way all this was dropped and introduced only in DX10, but you had to know some HLSL to access certain functions.
But now that DX10 was dropped, and DX11 is the only other renderer they have, it really leaved me peeved out on just how much of the 'basic' stuff which should be readily accessible to people is closed off and so limited, even if one buys the engine, unless they have a programmer who knows C on their hands, chances of you getting back much of the closes stuff it still hard.
My hopes is that UE4's entire shader systems kinks are worked out and made more transparent. Seriously, accessing something as simple as LOD's of a 512px texture shouldn't be a hidden feature, especially for Cubemaps.
You first seem to be narrowing in on an issue which is important to you, but in the whole scheme of things isn't that important.
Secondly I don't see how you can sit there complaining about UDK Vs DCC apps like Maya and Max, its apples to oranges. How about comparing UDK to other "Game Engines" and then tell me how inflexible it's material system is.
I hope in the new UDK they address alot of the issues and other people have, but those features aren't there probably because few people want or need to use them.
^
So I should say everything is fine and move along, that's it? Not ask for improvement in a certain area or be ticked off about it?
Are there other things they could fix? Yes, but isn't everything important to someone, so why should one thing be worth less then the other? Looking back, I don't see where I implied the UDK engine was a less then able engine, or that other things didn't need a spruce, hell, I think my Kismet comment alone is easily translatable to other areas on how unexposed it is sometimes.
Also, I cannot reference other engines because everyone has a hissy fit when I do, you can't even use the name of a certain Anglo-Nordic-Turk engine or one that has a developers name as that of a rolling numbered cube with some frosting on it, on Epic's site for comparison or even ask if such features exist since you topic or post will get a nice lockdown.
The reason I reference Max and Maya instead is because Max and Maya are poor in term of viewport bling, you have to pull every single trick in the book to get them look and nice and have an acceptable performance, as opposed to UDK which is super optimized and refined, yet you can have better definition of your material if you wish in the prior as opposed to the latter, which is awkward to say the least.
Also, considering we're well past the prime of using Specular all alone as the highlight of a model instead of a proper Cubemap which represents the roughness of a material, I would say this isn't a very narrowed down point at all, if anything that should be the next step in proper graphical fidelity which every bloody engines and their stockholders like screaming out of their slabbing man-tits.
The gist of it is simple; If someone is going to tout their engine as 'next-gen', they better bring stuff which have been available for ages now in laggy DCC viewports for a while now, because if I wanted soft bouncing tit physics, I would buy a really fat cat.
I'll agree that not having access to custom mip-levels and such is annoying (and caused me issues before), but that's a fairly niche feature that UE3 doesn't support.
Getting changes and features into UE3 from the UDK forums can be done - and has been done several times before. It just needs to be done objectively.
Replies
I'd be surprised if it ran on current gen console hardware. I'd also be surprised if it was all that different from UE3, or if it gave you any sense for what the next gen consoles will be like. There will be a lot of assumptions built in, and it will run on experimental hardware. Expect it to be a beefed up UE3, with a cool demo. They'll do behind closed doors demos at GDC, then put the video up online after GDC.
We don't even have proper access to Mips and LOD's for textures in the engine, let alone cubemaps, I really doubt they should even be thinking about this when we can do Mip levels in Max at DX9 level.
..what are you even talking about? ^^
I bet GDC it will be shown off behind closed doors or E3.
I guess well have to wait and see what happens.
Well if the samaritan demo was any indication, they are focusing on next gen and dx 11 with the new engine. Its supposedly built from the ground up and not an expansion on udk or unreal 3. Tessellation + Reflections + Film Effects +
Subsurface Scattering + Better AA + Deferred rendering + glossy reflections + stuff we haven't seen yet, hopefully. I wonder if they are throwing light baking completely out the window, I kinda hope so, and I'm also wondering if they are doing any megatexture features like in rage for terrain. Maybe we will so some cyrsis like real time GI.
Tightly integrated and friendly cross development with Max/Maya! ( opportunities to iterate at the very first stages of dcc content authoring. )
http://area.autodesk.com/blogs/stevenr/project_skyline_at_siggraph
Also Visa versa!
would be aweome if integration would allow for, custom Maya python solutions Like IK blending for character navigation of unique terrain. ( perhaps even leveraging my 12 years of previous dev libraries and aging talents ) :poly136:
Watching project Skyline examples of tight Maya UDK intergration has me as giddy as a little schoolgirl. ( I really wished art to engine never died! I think during Maya 2.5 - 3 ish? I still have some early promotional cd-Roms if anyone is interested in some ancient Mayan artifacts ).
I will likely fo my body more harm ( inadvertantly spazing, or holding my breath till I pass out ) If they do not follow up with Project Skyline at GDC 2013 )
Some fundamentals in the engine have changed; you'll already notice tiny snippets of the 'Kismet 2' libraries have fallen into UDK as minor references. Although I do not know yet if it is the case, I'm expecting to see a more library driven approach (such as runtime hot-switchable DLLs), which may suggest that UnrealScript may be wholly deprecated.
Other than the above speculation (and this is speculation), I'm contractually obliged not to say what I already do know.
Your kidding, right?
no new console tech to prepare for just what we currently have with slightly more power behind it, since it seems the new xbox and PS will be based on what is considered to be fairly old tech that is already being used in the average computer.
Crytek's dynamic lighting has issues - it's fiendishly performance hungry for a start, and seems to exponentially deteriorate with increasing number of light sources. Their shadowing on the other hand is pretty sharp. Lightmaps in UDK under Dx9 are essential, especially on console hardware, as the lighting is forward-rendered (like CryEngine) and costly. Under Dx11 it's deferred and signficantly better until you start using shader lights, then it's back to forward rendered again (but how many of you use shaders on lights?)
Seriously... What are you talking about?
Thats just wishful thinking though and I really doubt they will attempt to address the api access and script execution speed issues
If they were to take some cues from Unity (and they had it available to everybody, like UDK) I would LooOoOOoOOoOove to be able to make browser builds.
Im not sure i follow you there. Why not both? ^^
I really hope they have better single player support (from scratch) and improved the AI.
I also hope they have it in full real time so that we don't have to build lighting or
brushes.
And have them already in udk
deferred rendering has already been in full use, and flashy looking vertex blending is easily done on shader model 2.0.
Nothing harsh, people discuss things.
There's no point in immunizing epic against criticism in any way, even if it might be unfounded at some points, it's mainly just speculation going on.
Discusser's gonna discuss.
Cheers
I guess you could have both, I was just thinking about how limiting of a material editor the cryengine 3 has (compared to udk), and I've heard its a lot more expensive to have a fancy shader in a differed engine.
I think lightmaps are still going to be around for a few generations. You can't quite get the same amount of detail and lighting complexity with realtime compared to offline baking as well as being cheaper.
-ah gotcha
I've been noticing this in the snow scene I recently started. In areas where it's just snow over a fairly flat surface the lightmaps give it some strange reflections that make it look like certain parts are iced over which is cool.
EDIT: And just as I post that UDK starts lighting my shit super dark for no reason.
I'm talking about the unintuitive way a MaterialInstance and Materials themselves, are setup, not really Kismet or Artists friendly at all if you want external access for certain real-time modifications.
So that if you want to use LOD's and Bias's of a texture for say SSS because you don't have access to DX11 and want to custom tailor your own stuff, ceramic and metal light bleeding, it's next to impossible, not mention if you do a little HLSL which requires certain nodes, you're later on unable to access them through any instance.
Also, considering your able to do ALL this in Viewports for modeling packages like Blender, Max and Maya, at DX9 level, I find it strange the way all this was dropped and introduced only in DX10, but you had to know some HLSL to access certain functions.
But now that DX10 was dropped, and DX11 is the only other renderer they have, it really leaved me peeved out on just how much of the 'basic' stuff which should be readily accessible to people is closed off and so limited, even if one buys the engine, unless they have a programmer who knows C on their hands, chances of you getting back much of the closes stuff it still hard.
My hopes is that UE4's entire shader systems kinks are worked out and made more transparent. Seriously, accessing something as simple as LOD's of a 512px texture shouldn't be a hidden feature, especially for Cubemaps.
Secondly I don't see how you can sit there complaining about UDK Vs DCC apps like Maya and Max, its apples to oranges. How about comparing UDK to other "Game Engines" and then tell me how inflexible it's material system is.
I hope in the new UDK they address alot of the issues and other people have, but those features aren't there probably because few people want or need to use them.
So I should say everything is fine and move along, that's it? Not ask for improvement in a certain area or be ticked off about it?
Are there other things they could fix? Yes, but isn't everything important to someone, so why should one thing be worth less then the other? Looking back, I don't see where I implied the UDK engine was a less then able engine, or that other things didn't need a spruce, hell, I think my Kismet comment alone is easily translatable to other areas on how unexposed it is sometimes.
Also, I cannot reference other engines because everyone has a hissy fit when I do, you can't even use the name of a certain Anglo-Nordic-Turk engine or one that has a developers name as that of a rolling numbered cube with some frosting on it, on Epic's site for comparison or even ask if such features exist since you topic or post will get a nice lockdown.
The reason I reference Max and Maya instead is because Max and Maya are poor in term of viewport bling, you have to pull every single trick in the book to get them look and nice and have an acceptable performance, as opposed to UDK which is super optimized and refined, yet you can have better definition of your material if you wish in the prior as opposed to the latter, which is awkward to say the least.
Also, considering we're well past the prime of using Specular all alone as the highlight of a model instead of a proper Cubemap which represents the roughness of a material, I would say this isn't a very narrowed down point at all, if anything that should be the next step in proper graphical fidelity which every bloody engines and their stockholders like screaming out of their slabbing man-tits.
The gist of it is simple; If someone is going to tout their engine as 'next-gen', they better bring stuff which have been available for ages now in laggy DCC viewports for a while now, because if I wanted soft bouncing tit physics, I would buy a really fat cat.
Getting changes and features into UE3 from the UDK forums can be done - and has been done several times before. It just needs to be done objectively.