Whats wrong with EPIC/Unreal Engine?
Guys, I am following all the developments of Unreal Engine, CRYENGINE and Unity for a couple of months now. Let me say something first - CRYENGINE already had REAL-TIME DYNAMIC GLOBAL ILLUMINATION since a few years back! Frostbite already has the WYSIWYG editor since frostbite 2, and now Unity 5 is coming with the Dynamic Global Illumination that builds on Geomerics' Enlighten technology. Now, we all know who is lagging behind is terms of Dynamic Global Illumination.
I mean what really is wrong with the guys at EPIC? Whats wrong with them? Every guy in the neighborhood is implementing Dynamic Global Illumination while Unreal Engine is still lagging. I know they did implement the LPV system but that's too in the BETA phase and often produces not-upto-the-mark results. I mean Unity 5 got it now. Epic is such a respectable company and I have my trust in them; they have been in the neighborhood since a pretty long time, but this is getting too far. They come out with SVOGI, make everyone super-excited and remove it saying it's extremely resource consuming. I mean, What the heck!?!
Every other major engine (I guess) is starting to implement this feature and I think Epic is the only one in which we have to pre-bake. Man this is time consuming! I love all the blueprints and all that but those aren't an excuse for not bringing in the 'Dynamic Global Illumination' feature. I mean, this sucks so much! While working on a huge level this baking thing totally eats up time.
Guys, I am totally getting sick of this. I am still waiting to hear something on this feature from EPIC. All I have to say is it's really been a long time and they better implement the feature! I mean look at what the devs can gain - no pre-baking shadows and lighting. This will be EPIC! EPIC!
I mean what really is wrong with the guys at EPIC? Whats wrong with them? Every guy in the neighborhood is implementing Dynamic Global Illumination while Unreal Engine is still lagging. I know they did implement the LPV system but that's too in the BETA phase and often produces not-upto-the-mark results. I mean Unity 5 got it now. Epic is such a respectable company and I have my trust in them; they have been in the neighborhood since a pretty long time, but this is getting too far. They come out with SVOGI, make everyone super-excited and remove it saying it's extremely resource consuming. I mean, What the heck!?!
Every other major engine (I guess) is starting to implement this feature and I think Epic is the only one in which we have to pre-bake. Man this is time consuming! I love all the blueprints and all that but those aren't an excuse for not bringing in the 'Dynamic Global Illumination' feature. I mean, this sucks so much! While working on a huge level this baking thing totally eats up time.
Guys, I am totally getting sick of this. I am still waiting to hear something on this feature from EPIC. All I have to say is it's really been a long time and they better implement the feature! I mean look at what the devs can gain - no pre-baking shadows and lighting. This will be EPIC! EPIC!
Replies
VXGI was also announced and Epic said they might look more into it.
SVOGI is extremely slow, its really cool, but not quite there yet, I feel like soon it (or a similar technique) will be more common, but current gen consoles are a little too slow for current implementations, which is a shame.
Cryengine does a screen-space technique that while fast, doesn't look nearly as good as proper GI IMO.
Unity 5 ships with Geomeric's Enlighten, which is available for both UE3 and UE4 too as well. http://www.geomerics.com/enlighten/ Enlighten isn't exactly realtime, it requires a lot of preprocessing and then you can make various changes to the lighting dynamically.
Realtime radiosity is awesome in concept, but its extremely performance heavy, and worse yet, many games don't actually show much benefit at runtime to have radiosity running realtime rather than pre-baked, exceptions would be games with a day-night cycle. For most games with fixed lighting setups, pre-baked radiosity is much, much, much more efficient. Until GPUS that can handle complex raytracing are common place (even the top of the line GPUS from nvidia and ati aren't really there yet), it will be a while before fully realtime raytraced radiosity/gi is feasible in most games.
So, what is wrong with Epic? Nothing really, they're simply playing it safe at the moment. They have a very viable light baking system currently that has been used in countless games, and if they rush out some realtime radiosity solution that is half baked and not really fast enough to run along side an actual game (its important to note tech demos aren't representative of what you can do in games), people would bitch just as much about poor performance as they are that SVOGI wasn't included. Damned if they do, damned if they don't. With UE4's base platform being a subscription based model, they can introduce features when they feel they are actually ready, rather than rushing to market.
What's the big deal?
Everyone always rants about realtime radiosity but if your prebaked solution looks great who cares?
Even LPVs don't necessarily update every frame because its super expensive.
https://forums.unrealengine.com/showthread.php?44278-Realtime-Dynamic-GI-Reflections-AO-Emissive-plugin-AHR
Err, how shall I put this: unless you've tried an engine in production, it's very hard to make a judgement. You seem to think Geomerics' Enlighten is some magical solution, but it has its issues as well for certain scenarios.
No reason to make such a big fuss. There's more to an engine and its lighting system than the big words their marketing departments throw around.
But yeah like Xoliul said, Enlighten and similar tech isn't some magic wand, there's a downside to everything. We used a very similar solution to Enlighten for the dynamic GI in Alien: Isolation and there was quite a bunch of problems with it let me tell you, we just had to work around them. Same as you would with Enlighten. Same as with Lightmass and other prebaked GI.
Screenspace GI is complete garbage for numerous reasons.
Let me relieve you mate.
You ARE NOT a drama queen and you ARE NOT overreacting.
You are asking the absolutely obvious question that few people think over. The same question I asked myself all over the summer.
I tried switching over from CE to UE4 because their services felt much more " friendly" On the other hand since UE4 was still "young" many many features were missing including real time GI. There was the promise of " they will be there soon" .
So I WASTED 2 months for deciding over one for another. Watched the twitch streams. I noticed they always "dodged" the question of real time GI. They still do.
I have a conspiracy theory. Would you like to listen ?
They don't implement a "great" real time GI solution... Because they simply can't.
Either it is something in the fundamental engine layout that doesn't allow great working real time GI or simply they do not have enough expertise on the subject to do a good one.
I am no code whiz. So ... yes a " conspiracy" theory
If you want to use a great real time GI for your work the only one open to access is CE. It is really designed for high level users though. So it will take some time to learn it.
Also all the new engines are leaning towards the real time GI for obvious reasons. Like Snowdrop ( my favorite) , Fox Engine etc... So it completely LOGICAL to use a real time GI solution.
Sounds like you really wasted 2 months then.
Btw, those of you waiting for a dynamic GI solution in UE4 check this out: https://forums.unrealengine.com/showthread.php?44278-Realtime-Dynamic-GI-Reflections-AO-Emissive-plugin-AHR
If you realy want GI that bad, you can just use enlighten.
if you don't have the budget for enlighten then you probably shouldn't need GI.
honestly GI realy is not important at all IMO.
in most games that actually utilitize GI the avarage user will never notice it or its effects.
a lot of games that use enlighten end up not using the realtime GI at all because its way too expensive and its effects too subtle.
it is, and they even discuss how they faked it for that Ryse on console in one of their tech paper (i'll try to dig it up). essentially they took their SSAO pass, tinted the shadows based on the colour pass, downsampled heavily to simulate a kind of blurry-illuminated haze, and overlayed it.
You pointed out something nice here! I did notice someone say this exact thing in the Unreal Engine forums! I dont know what causes it to crash though!
I really hope they do release it soon! I have great respect for EPIC but also for Crytek! Now somebody here will say maybe I'm wrong on this one lol.
@PogoP Am I being that dramatic? lol. Try to understand here! I am just trying to say that if cryengine can do something why not unreal engine! Now, I know cryengine uses screen-space GI; so why not give a minimum screen-space GI thing that cryengine uses. At least it can suffice for now until the guys at EPIC come up with something usable! Tell me I'm wrong!
Thanks for the help out there! About the conspiracy theory - lol, maybe true! haha, but I really want them to implement that. We are humans lol. We can sure do it! Sure!
@RexM Are you outta your mind? Are you kidding me right now? lol
@Goeddy Maybe you are right! I am a student and I dont have the budget! Btw, cryengine has it, so why the heck cant UE have it? huh? Silly! ain't it?
Thanks for that! uahahaha! Brilliant one!
Dynamic GI in most cases is not better than baking lights, and even as the tech behind it goes forward and gets faster so will the tech behind lightmapping which will always be more accurate.
What people choose to use for their game generally has more to do with what their games bottleneck is. Since one hits your gpu cycles hard, and one hits your memory.
Also epic is working on a solution, so what do you want them to rush and give it out as buggy crap? Or take the time to get it right?
UE4 already has the exact same GI solution as CryEngine, light propagation volumes, implemented as a beta feature that is pretty easy to turn on. It's completely usable.
I'd be interested in seeing that documentation though, does sound interesting.
Game is for games.
Dev is Development
de·vel·op·ment
dəˈveləpmənt/
noun
1.
the process of developing or being developed.
synonyms: evolution, growth, maturation, expansion, enlargement, spread, progress; More
Haven't actually seen this before. Made my day
Thanks mate
Considering there is probably going to be an iteration of maxwell soon since mass 20nm production is way overdue...
mebbe the tech is fine for an enthusiast market?
I thought epic buried the GI into an optional beta feature fer now because they were surprised the new generations of consoles were so "underpowered" compared to what they were predicting?
Till then I think they were full steam ahead on the tech as it exists? ( isn't it mostly a hardware issue at this point? )
90%?
I have not read that report?
The Jon Peddie Research group printed an interesting report last year on the value of the enthusiast gaming market.
yes enthusiasts will pay more money for value...
( that same report finds: although "enthusiast" gaming hardware market makes up a small fraction of the overall PC hardware peripheral market compared to mid to entry level products... the high cost that "enthusiast" market is willing to pay by some accounts adds up to 50%? of total earned revenue overall )
Enthusiast
PC
Gaming Market
table of contents ( purchase required however all the finding have been published )
http://www.jonpeddie.com/download/PC_gaming_hardware_market_report/The-Enthusiast-PC-Gaming-Market-TOC-Exec-sum.pdf
There has been one follow up to the initial research which seems to vindicate the health of the performance market.
PC gaming hardware sales unaffected by overall PC decline
http://hexus.net/business/news/components/58669-pc-gaming-hardware-sales-unaffected-overall-pc-decline/
http://www.vgchartz.com/gamedb/?name=battlefield+4
http://www.vgchartz.com/gamedb/?name=battlefield+3
http://www.vgchartz.com/gamedb/?name=call+of+duty&publisher=&platform=&genre=&minSales=0&results=200
you can search for any modern title, but those are shooters, games which are supposed to played better with mouse and keyboard for the lack of precision on a controller.
And we are talking about times where a PCgame costs 50 euro while the console version costs 70.
go check the sales of the crysis titles, crysis 1 being developed for this huge market of pc enthusiasts, on crysis 2they had 4 times the sales on consoles on crysis 3 even tho a lot broke away they still had about 4 times the sales on consoles.
check tombraider, now the sims could be a good example of pc sales, but this is not a game for the hardware enthusiasts. its a game for masses, the moms, the kids, the couples.
dota, lol, blizzard games, sims, tons of them play browsergames (not in a specific order)
this is not the hardware enthusiastic market, and this market - as with consoles - isnt one market you can grab in one strike. shooter players might not care for lol, mobaplayers might not care about the witcher.
I am sure a forecasted 6.5 billion dollar enthusiast market would not mind running a lil GI on their $3000 dollar boxes and feeling "special" cuz consoles couldn't do the same
The report isn't just covering hardware either but pointing out the undervalued percentage of overall sales ( and growth ) the enthusiast market represents. A market which also traditionally claims that their experience is hamstrung by console development concentration.
1. Every rendering engineer would gladly throw away all other work and start researching dynamic GI. Simply because it's hot and very interesting topic. It's holy grail of real time rendering.
2. But.. There are few problems. There are only 5 rendering programmers at Epic.
3. One of them is dedicated to working on Niagara.
4. Other guys are working for more than half year on multithreaded rendering. They also have other tasks like DirectX 12 (which will happen to public sooner than later), DFAO etc.
5. Epic is not Crytek with dedicated R&D department dedicated to rendering.
6. Which is shame, but hey, priorities.
7. Hiring new people to work on rendering, wouldn't help with current tasks, but could probably allow to start new ones.
8. It's not helping that one of the original authors of SVOGI no longer works at Epic.
Screenspace GI is good complementary solution. It can provided quality on tight spaces, where no other real time solution can.
It's the same with screenspace AO. It doesn't really make good AO, but combined with medium and large scale AO solution it works great.
Anyway. We prolly are going to get VXGI before end of the year. So sit tight and wait.
The problem is, when it doesn't.
Dynamic ToD is not the only thing where static lighting doesn't work.
Big open world maps also are no option for lightmaps. Baking then is taking ages. But worse, they take ridiculous amount of space and memory. My simple map, I have been working as showcase for marketplace is taking 500MB with lightmaps. And I'm not even 1/4 done with it, nor I'm using big landscape. Only 1k.
At this point I'm just goint to provide fully dynamic lighting map, with lightmapped map. Just for comparsion.
Why not both ? ;p
Ohhh... So THAT WAS WHY !
Also the MOST INFORMING and MOST HELPING post in this thread EVER ! Hands down ! No question !
I love you mate. I really do. And I wanted you to know that
What is this Niagara?