I also like the kite demo, no question about that but the performance is horrible “sadly”… And yeah GTA5 looks amazing… Ghost Reacon looks even better than the kite demo not just because of the buildings, hopefully UE4 will reach soon the same kind of quality and performnace like the Magic engines with they’e lighting solutions.
So we have to wait at least until the release of Dead Island to hear a bit more about the Yager solution of lighting ? I’m not into engineering engines or renderers but i would be totaly satisfied with the same kind of solution that Cryengine is using, it has also his weak points but i could absolutly live with the that
You are right! What I meant with that is, its different when you look at Naughty Dog who works only on Playstation and Uncharted (yeah they did more games, but TLOU uses a lot of tech from uncharted) compared to Epic who work on a multiplatform Engine that has to do soooo many things that all take time to develop. So yeah…we will get fancy and optimized stuff in the end, but it will take longer because the scope is larger with a broader horizon.
And yes, running fast has nothing to do with the game itself, it depends a lot on having good people that know exactly what to do for a given situation^^
@Adlik, read the paper about lighting in Crysis3, multiply the effort of lighting that level by maybe something between 50 and 100 for an open world and…nope^^ Its extremely timeconsuming to set up
But Crysis 3 was using the old LPV system, the new Environment probe based system works a lot better on huge worlds, at least that is what i read about the CE3 lighting…
By the way, any news about Enlighten ?
I’m not too convinced from trailers coming from a huge PR show, even a bit sceptical. While it would be nice if they find new technologies, I have my doubts. There have been too many lies in the past, so until I play the game for myself and see that their engine does what they promised, I won’t applaud anything. (Unless they talk about it, like at GDC)
As far as I understand it, many of these engine are made for specific cases and are not generic usable like UE4. Sometimes they even come with limitations in some form or another. What I do have to admit is that the effort put behind them is astonishing, many implementations are based on clever tricks since they had the need to find some way.
It feels like many technologies powering engines running on consoles have more effort put into them (since console sales are a huge), while on PC we end up with more or less brute-force attempts.
Anyways, it’s nice to read these posts, while I have to believe they are true, many informations mentioned are interesting and I do hope that someday I can talk alongside you properly ^^
Enlighten falls under “pseudo-dynamic”, and has several drawbacks. First off, it’s hard to make major changes to a scene after the initial preprocessing steps. You can have different “states” of the scene for different cases (like a building being destroyed like you’d see in Battlefield, or a wall being destroyed), but generally it’s difficult to have fully dynamic worlds with Enlighten. There’s also problems such as it doesn’t handle rapidly updating light states very well - there’s usually some perceptually noticeable “lag” with the GI updating with respect to fast moving light sources, and similar goes for quickly changing light color.
It’s neat if your lighting changes slowly and you’re not going to have a lot of geometry changes going on in the scene in general, but otherwise it’s quite limited in that regard.
Carmack went too far imho though, he wanted to use voxels for everything, stop using triangles. Rasterization works great for solid surfaces and direct lighting, voxels should be used for GI and similar stuff.
But the oclusion quality of cryengine doesnt look half bad , while you mentioned that your low quality mode is going to have a lot lower quality oclusion , or do you think that it is going to look as good as that voxel based global ilumination tecnique , personally i would prefer to have high quality indirect shadows cheaply , while not sacrficing the gi shading quality ( you said that you were calculating it at 1/4 the resolution , so i dont want to go any lower than that) but i wouldnt not having reflections with ahr , especially since i am looking at real time reflection tecniques that use the harware pipeline instead of the shaders .
I do the trace at half res actually, 1/4 is pushing it too much. When I said that the oclussion would have less quality, I meant that you wont have the fine sub-Voxel oclussion that raytracing gives. I imagine that the results will be similar to the ones if crytek, specially the AO on the house. Also, reflections are the cheapest part of the system, they add about 1 ms at 1080p on my weak gtx 750 ti, so no point in disabling them.
I really need to implement I to see the results though.
Having problems building from the Clone Zip. Could use some help, I’m a bit lost. Followed the usual plan. Downloaded clone zip, unzipped to D:\Unreal Engine\4.8_AHR, ran setup.bat and downloaded the dependencies, ran generate project files.bat, opened the SLN in vs2013 CE, ran build. I get error.
That’s strange, maybe a problem with administration rights?
BTW, there isn’t a 4.8 version, i put 4.8 update on hold, working on multiple bounces, interpolation and the cheap preset.
That’s really strange then, sorry I can’t be of more help.
Thanks! 4.8 will take a while, but the other stuff should be coming up earlier.
I have 5 exams though, on the 30 of June, 2 , 4 , 8 and 10 of July, so busy week ;D
The results look really nice, but I’m confused about the method. For diffuse GI do you just sample several rays? With a fixed step size, too? That sounds inefficient compared to voxel cone tracing. Tracing one cone is only about as costly as tracing one ray, but you get much smoother results with fewer traces, and the step size increases with each step, reducing cost further.
