A first look at Unreal Engine 5 • Hacker News Board

Re: A first look at Unreal Engine 5

#671

Post by floatboth » Thu, May 14, 2020, 6:13 PM UTC

simias wrote: post #161

ksec wrote: post #8

In the past, I have always looked at Demos from Games and thought this is very good and getting close to movie quality. But that "close to" remained "close to" for quite some time. Even though It is improving every year but you can still tell it is gaming graphics. Even if some of the shots are not real time and pre rendered, they are still gaming like. That Unreal 5 Demo was the first time ever I thought this is Hol…

At the risk of sounding a bit negative I personally find that graphics have plateaued since about the PS3. Sure, there are more polys, sure, there are higher res textures, sure, there are more complex and dynamic lights. But you don't really have the kind of gap we used to have between, say, the PS1 and PS2 for instance. Diminishing returns and all that. The problem is that, in my experience, this eye candy only matt…

> there are more polys, sure, there are higher res textures, sure, there are more complex and dynamic lights

What I'm noticing is there's too much focus on fancy lights, over-the-top postprocessing effects with annoying crap like ambient occlusion, screen space reflections that hog a lot of performance for just reflecting your character in water puddles placed everywhere just to show the effect off, and so on.

And not enough focus on good old textures and polygons. Approaching spherical objects closely still makes the polygons very very obvious. Staring directly at walls still shows how low res textures are.

> You could destroy everything, dig holes, build things

You might be looking for Minecraft! :D

Re: A first look at Unreal Engine 5

#672

Post by kbenson » Thu, May 14, 2020, 6:19 PM UTC

kevstev wrote: post #523

simias wrote: post #402

Earlier quoted context omitted.

I must have expressed myself poorly, my point was not that I don't think it looks good, my point is that I feel like all this effort is focused on a single metric, making stuff look good in trailers. Remember when HL2 was announced you had this super fancy physics engine? How objects would bounce and realistic react with each other, how you could stack things and come up with puzzles that would just use the physics e…

I will even go farther and say that I feel like I have now been playing the same game for 25 years. The FPS genre gets some new tricks, but it's still the same basic game with a few new mechanics and tricks- I can pick up a game I have never played before and generally be reasonably competent at it within a few minutes- a stark contrast to when I picked up tribes for the first time and was completely useless in onlin…

I think part of that is that the UI and control system has standardized on a mostly similar set of features with somewhat standardized mechanics. They've iterated over the last twenty years towards what is likely the peak (or at least a local maxima) of what you can do with a mouse/keyboard or controller (and controllers are generally normalizing as well, except for small differentiators). I.e. there's only so much control over your physical space you can have with a mouse and keyboard while still allowing good character movement, so there are limitations on what can easily be done.

On the other hand, VR based games are wildly experimenting and iterating because the control interface is so different (while at the same time fairly intuitive in that it maps to our reality better). Superhot in VR (which I have on the Quest) definitely doesn't feel like the same game from the last 25 years. Half-Life Alyx is supposed to be an amazing experience (I haven't had a chance to try it yet, and I won't for a long time likely), but that's not supposed to be because it looks so much better, but because it offers a wildly different and new experience.

So, if you're tired of feeling like you're playing the same old games, try VR. The Quest is probably the cheapest way into this space, but it's not actually cheap at $400, and it will gate you from a lot of the premier experiences available if you have a powerful PC with a good graphics card and VR headset that can interact with it (which the Quest can also do with a cable, so you aren't locked to Quest games, although maybe not at quite the visual quality of some other headsets).

Re: A first look at Unreal Engine 5

#673

Post by Arelius » Thu, May 14, 2020, 6:28 PM UTC

rsp1984 wrote: post #613

trynewideas wrote: post #575

Earlier quoted context omitted.

That video's amazing because the part of the UE5 demo I can't wrap my head around is where Nanite magically turns 1 billion triangles of source art to 20 million triangles of rendered art, per frame. And then I get a minute into the UE3 video from 2004 and what do I hear? "The scene you're seeing here is about 1 million polygons of in-game assets, which came from about 200 million polygons of source art." Then, deja…

That video's amazing because the part of the UE5 demo I can't wrap my head around is where Nanite magically turns 1 billion triangles of source art to 20 million triangles of rendered art, per frame. I was wondering about that too. My best guess is that they put the heavy geometry items in an acceleration data structure and then ray trace / cone trace these in a first pass to get the points where the camera rays inte…

This doesn't really seem volumetric in nature to me.

Firstly, UE4 has broad-phase that I understand they are quite happy with, I doubt they would add in a ray-trace, since not only would that not really improve anything, it would also not integrate well with any non-nanite techniques of rendering geometry, and my understanding is at least for now, nanite can only handle (mostrly)static geometry, and I doubt they would want to give up the ability to work with other rendering techniques.

If I had to take a wild guess, I would guess they have a somewhat standard process to generate a low detail mesh, which is what get's dispatched to the GPU, then, atop of that mesh they may build triangles acceleration structures parameterized across the surface perhaps similar to this paper [1] From there, you could do similar to what you suggest, and just generate additional triangles such as is in that paper.

However, given how poorly GPUs handle pixel-sized polygons, that may not be the best approach. so I wouldn't be surprised if tessellation is used to around a quad level and then from there, the rest is either done using compute-style rasterization in the pixel shader, or some outputs are written and additional rasterization is actually deferred for a compute job.

I'm sure there are all sorts of exceptions and edge-cases, but I wouldn't be surprised if it looks something like that general workflow. On the other hand, I'd be very surprised if there was anything that looked like ray-tracing on the scene level, perhaps traversing the surface-space data structures looks a bit like ray-tracing.

[1] https://graphics.stanford.edu/~niessner/papers/2016/4subdiv/...

Re: A first look at Unreal Engine 5

#674

Post by naikrovek » Thu, May 14, 2020, 6:30 PM UTC

fluffything wrote: post #645

ahaferburg wrote: post #644

Earlier quoted context omitted.

It's quite possible that they make heavy use of instancing, which might reduce the memory footprint by a lot. Not every pebble or rock needs an individual geometry.

From the video it sounded to me that the 100s billions of triangles were from single instance geometry. The models they mention in the video are in the ball park of >1 billion triangles each (that's 12 Gb per model). So if they have 100 models of that resolution, that's 12 Tbs, and whether they instance each model 1 or 100 times doesn't matter much because the problem of having to fit those 12 Tb in triangles somewhe…

The statue instances do not require additional storage or RAM usage per instance. That's the whole point of instancing.

A moment earlier in the video, the poly count of a single statue mesh was given: "more than 33 million triangles."

Re: A first look at Unreal Engine 5

#675

Post by vertex-four » Thu, May 14, 2020, 6:32 PM UTC

syockit wrote: post #621

vertex-four wrote: post #595

Earlier quoted context omitted.

https://en.wikipedia.org/wiki/Source-available_software#Dist...

http://hyperlogos.org/article/Who-Invented-Term-Open-Source

O... k? But what does open source mean today? What has it meant for approximately the last twenty years? Did Epic claim their software is open source? Did Microsoft claim their source-available software was open source? I wonder why not, it would've been a good marketing term, what with all the developers who want open source.

The answer is that pretty much everyone agreed that open source means something roughly equivalent to the OSI definition, and even big companies agreed with that... until a year ago or so when somebody decided that they wanted all the benefits of claiming to contribute to open source without actually doing so, and a bunch of people who never actually cared about the FOSS ecosystem aided them in doing so.

Re: A first look at Unreal Engine 5

#676

Post by naikrovek » Thu, May 14, 2020, 6:32 PM UTC

notaplumber wrote: post #386

0xCMP wrote: post #346

Earlier quoted context omitted.

Good point, but OP's point is that they can't hide behind demos. Anyone who requests the source will most likely get it and be able to see what's going on.

Indeed, that part is accurate. It would be technically interesting if the tech demo was a part of UE5 source release. I'm not familiar if they've done that in the past.

Demos are sometimes released in the Unreal Marketplace. Unity often does the same in their own store.

Re: A first look at Unreal Engine 5

#677

Post by kevstev » Thu, May 14, 2020, 7:19 PM UTC

kbenson wrote: post #672

kevstev wrote: post #523

Earlier quoted context omitted.

I will even go farther and say that I feel like I have now been playing the same game for 25 years. The FPS genre gets some new tricks, but it's still the same basic game with a few new mechanics and tricks- I can pick up a game I have never played before and generally be reasonably competent at it within a few minutes- a stark contrast to when I picked up tribes for the first time and was completely useless in onlin…

I think part of that is that the UI and control system has standardized on a mostly similar set of features with somewhat standardized mechanics. They've iterated over the last twenty years towards what is likely the peak (or at least a local maxima) of what you can do with a mouse/keyboard or controller (and controllers are generally normalizing as well, except for small differentiators). I.e. there's only so much c…

I guess I just feel there is a real lack of creativity in gaming these days- There used to be a rich variety of games- puzzle/adventure games, RTSes, Simulation games, Flight/Space games of all kinds, and now its just kind of condensed down to FPSes, repeated sports franchises, and a few outliers- KSP, Minecraft, to name a few.

Interestingly, I won an Occulus Go at a meetup about 9 months ago. I almost gave it away, I didn't really know what I had won, I thought it was going to be an upgraded google cardboard.

But man I was surprised at how cool it was. And yeah- everything feels really raw and unpolished and experimental, kind of like the early days of the internet. My biggest gripe is that the games are all kind of vapid and on rails, there are no games where you can kind of go and explore, but still its all quite interesting and has me interested in buying a "proper" VR setup.

Re: A first look at Unreal Engine 5

#678

Post by fluffything » Thu, May 14, 2020, 7:52 PM UTC

naikrovek wrote: post #674

fluffything wrote: post #645

Earlier quoted context omitted.

From the video it sounded to me that the 100s billions of triangles were from single instance geometry. The models they mention in the video are in the ball park of >1 billion triangles each (that's 12 Gb per model). So if they have 100 models of that resolution, that's 12 Tbs, and whether they instance each model 1 or 100 times doesn't matter much because the problem of having to fit those 12 Tb in triangles somewhe…

The statue instances do not require additional storage or RAM usage per instance. That's the whole point of instancing. A moment earlier in the video, the poly count of a single statue mesh was given: "more than 33 million triangles."

If one mesh is 1 billion triangles, and you have 100 different meshes, you need to store 100 billion tringles, unless you do something radically differently.

If you have 1 triangle mesh, and instance it 100 billion times, you don't need to do that.

> "more than 33 million triangles."

The assets the company they mention makes have 100s of millions of triangles, and billions as well. So I wonder if they are not using those original assets after all.

IIUC they are, but they are only rendering at lower triangle counts and the LOD happens automatically.

Re: A first look at Unreal Engine 5

#679

Post by Miraste » Thu, May 14, 2020, 7:56 PM UTC

alkonaut wrote: post #666

Miraste wrote: post #657

Earlier quoted context omitted.

> I don’t think the SSD but rather the GPU would be doing the tesselation... Ha, I phrased that badly. I meant that, if the high-poly models can't all be stored in VRAM at once, is the SSD fast enough to load them back onto the GPU every frame?

If the tesselation is performed on the GPU (surfaces uploaded to GPU in some non-triangle representation e.g image geometry/SDFs/patches) then I don’t think it’s ever viable to have the tesselated triangles loaded back from GPU memory even to RAM (never mind disk). This isn’t a lot of data. 2 triangles/1 vertex per pixel is just 8K triangles and 4K vertices every frame, that can be overwritten each frame. This is tin…

Ah, that makes sense. I was thinking the problem lay in loading the original meshes, but I didn't consider they could be using a smaller non-triangle format. I really hope they share more about how this works.

Re: A first look at Unreal Engine 5

#680

Post by jayd16 » Thu, May 14, 2020, 8:03 PM UTC

random32840 wrote: post #637

jayd16 wrote: post #522

Earlier quoted context omitted.

>And why is it so damn hard to get clothing to actually fall on the body right, instead of always looking like a stretchy foam skinsuit? Mostly because it is a stretchy foam suit. A model is deformed and translated based on animated bones in the model. Essentially each vertex is tied to a small number of transforms matrices(bones) with multiplier/weight. Animations are updating the transform matrices over time. You o…

I can see deep learning shaking up dynamic cloth animation dramatically in the not too distant future. It's already happened for fluids.

How so? We can already render really nice cloth offline. The issue is the runtime cost of sending that animation data to the GPU.

Are you suggesting building some kind of GPU side physics protocol that takes material properties, and existing bone data and fills in some of the blanks? I suppose you could start to push bone motion vectors and interpolate some things.