Unreal Engine 5 = Game Changer

I guess they may add it in 2020.2 that's why they want to ban everything currently.

https://unity3d.com/unity/roadmap

 

(NSFW)

 

Unity just have the bad habit of overpromising

 

If they can not get rid of lightmaps here are some things they could fix

autoscale: if low variation of direct light across a mesh use lower resolution, if high variation of direct light across a mesh use higher resolution.

auto baketag: Spatial grouping of meshes and tag them for better grouping of lightmaps to reduce draw calls.

remove faces not seen: This one is a bit hard but a combination of navmesh and occlusion data could be used to remove faces that are not scene by camera, this saves both lightmap space and fill rate on GPU.

 

If we are making requests: a LOD system that isn't a chore to use, integrated directly to the MeshRenderer or a derived component class, with LOD assignments made at the asset level instead of littering the scene with extra GameObjects and transforms just to hold the LOD meshes.

Oh, and an in-editor LOD generation system that actually works, not the hilariously bad mesh decimator that AutoLOD uses.

 

Well, the new pricing for unreal is certainly interesting. More competition is always good I just wish that this move by epic would make unity revise the pricing model for the plus license. Unity charges a quiet hefty sum for what is basically unlocking a checkbox that disables Unitys logo. Not only did it increase in price the last time I renewed my yearly subscription, due to the exchange rates going up I need to pay almost 50% more than last year. At least Unity could allow me to attach my license to more than two systems.

 

Saw an interesting youtube video on AMD and the PS5 that mentioned that the speed of it's >800GB SSD was equivalent to DDR2 RAM so it's like having a very large but slightly older RAM drive.

It should make for amazing load times and combined with the GDDR6 RAM that is 10x faster than DDR4 RAM makes for an amazing performance system.

I suspect that PC hardware might struggle to catch up without some major changes e.g. RAM bandwidth (quad or octo channel DRAM anyone) and way faster SSDs.

 

Gen4 nvme are as fast as mid tier DDR2

edit: A 3080 TI will have 16 gig of GDDR6 so you can probably load alot of the level directly to VRAM. I dont believe for a second peasants will have better performance than the master race when PS5 hits the store

 

I mean, this is like Unity themselves confirming that adding the Unity logo to things, devalues products.

 

I downloaded Unreal first time 3 days ago. I cannot believe how BluePrint connects everywhere - I am completely impressed and asked myself why I did not try this before. I confess I was mocking with UI before after 1 hour I realized how well it is actually. Even those colored buttons..
I have checked many tutorial channels but this guy's channel is simply amazing. He is like 3DBuzz on steroids... Super fun... Check him out..
https://www.youtube.com/channel/UC7HRB-YIUrG29zufq-vURsw/playlists

I will keep Unity for my future mobile game developments. (lots of assets i purchased- gotta use them somewhere)

 

Here is why Unity is bad, because even basketballs are hanging the game !

 

It doesn't but "Every other platform that doesn’t have these capabilities will go through a more traditional rendering pipeline, in which we’ll take these assets you’ve built and scale them down to more traditional LODs, rendering them so you can — there will be a version of this demo you could run on Android devices from three years ago."

 

That sounds a lot like it explicitly won't run on mobile.

 

If it allow to build on mobile without any additional hassle by leveraging the same tech for baking it down, I say it's a win. Better than switching from HDRP to URP.

 

About C++/Blueprints vs C#. I've seen few posts that epic is aware of C++ flaws and I guess something may be added in UE5 or later. In Jan 2019 they bought Agog Labs https://www.unrealengine.com/en-US/blog/agog-labs-joins-epic-games that previously worked on scripting asset described as "the sweet spot between C++ and BPs" https://www.unrealengine.com/marketplace/en-US/product/skookumscript.

 

Yeah. "Downscale gracefully" is miles better than "LMAO pick a pipeline and own it kiddo"

 

PS5 has 16gigs of ram too ...
But you are right on one point, I bet they won't compete on "SSD", my bet is on GPU starting to have bigger memory pool integrated directly with specific optimization.

 

NVIDIA announced their latest graphics architecture a few days ago and while they only talked about the server hardware they did mention the amount of memory that it would be coming with and with that we can try to guess the video memory that will be available for desktops.

Volta, the server-side architecture that pairs with the RTX 20 series, had 32GB VRAM while the RTX 2080 and 2070 had 8GB VRAM. Ampere, the server-side architecture that will pair with the RTX 30 series, will have 40GB of VRAM meaning the 3080 and 3070 will likely have 10GB VRAM.

It's a very minor increase but we're at the beginning of the life of the consoles and the current desktop memory standard is still new enough that it's very expensive.

 

Year ago there was Reddit thread about this: https://www.reddit.com/r/unrealengi...eople_at_epic_are_considering_adding/edxha25/ From what I remember guys from epic talked that they are aware of this issue and are considering c#, python, lua or something custom but are not sure yet. It would be nice to know what they choose in the end but it would certainly be a big blow to unity if by any chance they decided to use c#.

 

Looks really cool, I'm glad they're looking to address it (although it'll be hard to beat C# in terms of just general language aesthetic, I'm a huge fan of the language, not even including the context of Unity using it.) Seems like a lot of agents are looking to address C++ as the de facto "game programming language." Unity creating some highly performant C# subset, Beeflang, Jai, Rust for games. It's pretty cool to be in this industry.

Unreal not being able to do 2D well is kind of the biggest turn-off for me though. Most of my game dev at the moment is just game jam stuff, and without a dedicated 3D artist working with me, doing 3D is basically a no-go. To be honest, I've rarely seen Unreal Engine games in game jams anyways. I don't know if that's lack of familiarity or if that is a reasonable indicator that it just takes longer to get something up and running with UE4 than it is with Unity, Godot, GMS, what have you. Say what you will, Unity is amazing in the prototyping space.

 

It's very clear that "downscale gracefully" doesn't mean anything more than "we will do this automatically." This says nothing about the degree of control you get over that downscale, which is also a critical factor to consider.

Again, there's a lot of speculation here about an engine that we won't even see accessible until into 2021.

 

And remember that with the PS5 it is not just the game load times that are improved. The storage IO solution on the PS5 has enough performance to let the game engine stream assets directly from storage instead of loading each asset before using it. That means massive open world games won't need to worry about loading.

So far, I don't think Microsoft achieved the same level of storage IO performance with the Xbox Series X as Sony did with the PS5, so it is possible that the Sony PS5 might have a real winning feature in this next generation. Similarly, most PCs can't match the storage IO performance of the PS5, so PC games probably won't take advantage of direct asset streaming for a while either.

 

From everything I have heard regarding the PS5, it actually does sound like the storage IO will have enough performance to stream assets directly instead of copying them into memory first. It is a combination of NMVe drive, the new AMD APU, and custom storage related design tweaks delivered for Sony to the AMD APU. Nothing is for sure until the retail PS5 is available, though. If the PS5 can direct stream asset instead of loading each of them into memory first, this will open the PS5 platform up to massive world games with amazing detail levels.

Also, I am a PC master race enthusiast myself. I currently have a RTX2080ti in an AMD 3900X rig with three NVMe drives, and a RTX2080ti in an Intel i9-9900K rig with two NVMe drives. The custom water cooling solution in my Intel rig is worth more than an Xbox One and a PS4 combined. So understand that I am a big fan of powerful high end PCs, and I am still excited to see what Sony can deliver with the PS5. If Sony can actually deliver direct asset streaming, then the PS5 is going to be a major force in the industry despite costing a lot less than master race level gaming PCs.

 

We've been provided raw performance numbers for the storage but what interests me the most is what performance will be like under typical workload conditions. Consumers SSDs almost always operate somewhere between Q1T1 and Q4T4 with the former being the majority of accesses. Q1T1 is typically around 50MB/sec and Q4T4 is typically around 200MB/sec even on high performance NVMe SSDs.

Getting within range of the raw performance numbers typically requires you to either be working with random data at the Q16T16 level (almost always limited to server workloads) or working with sequential data greater than a gigabyte.

While it's almost guaranteed to be sequential workloads showing the raw performance of the PS5, if their custom storage approach means that they're getting within range with everyday workloads it's entirely possible PCs will never catch up in the lifetime of the console since the closest drive capable of this performance would be Intel's Optanes that cost $1,000.

 

Yeah, off course I want the consoles to evolve. Better looking console games benefit PC-players too.

 

I completely agree. Unity is going to be the Apple of Game Development. Only difference is iPhones are actually a good value for the money (compared to competition) and you get vastly superior performance than the competition. Unity Licenses are a bad value in comparison to both Godot and Unreal, you don't get access to the source code, and it performs arguably worst than both engines (Godot is more lightweight, takes up less ram typically). Unity is a luxury brand without the luxury product.

 

Do note that GDDR has much higher latency than DDR, which is the reason it is not generally used as system memory. This means that optimising for proper cache usage and allowing the system to prefetch properly will be very important for peak performance. This is precisely the sort of scenario where DOD is necessary.

 

Much as I hate to admit it, mostly because this is specifically why I've come to hate Apple (I was a lifetime user up until last year), I can't find much fault with this at all.

 

Unity still investing in AAA though I want to see how good are those tools, and how low they can scale, My machine takes many second to render a single frame of the 720p version of the demo.

Of course so far, as i'm watching it, it's another demo you can't make without external tools.

Will had edit if that change.

edit: it's useless in the practical sense, it's a marketing video

 

Actually there might be little to no difference in DRAM and GDDR latency:

Source - http://www.redgamingtech.com/ps4-vs-xbox-one-gddr5-vs-ddr3-latency/

You may have noticed that modern DDR4 RAM tends to have higher CAS latency timings but higher mhz speeds.

 

Jumping from 7 to 15 is a major increase in latency and keep in mind it's just talking about the CAS latency and not all the other latencies that will likewise increase with it. Latency here is the number of clock cycles before data is available. If all latencies are increasing relative to the increase to CAS you're more than doubling the time to request data from memory.

 

If unity like to buy small company they should acquire eisko and to get POLYWINK.

 

They say in the first 5 minutes that the job of their team is to do just that. Make demos that unity by itself cant make yet so they can report everything to the RnD team for future development.

 

To me that's whats bothersome about these cinematics though. I mean, with an elite team of the top artists & programmers and direct access to the engine team, you too can create an amazing cinematic using unity. Its not even a game that goes through a product lifecycle and needs to be supported and maintained, its just another one off project with basically zero pressure on the editor/engine because they are working in a way where the changes may or may not(lets be real) happen later, not that they are mission critical to the projects success.
Now dont get me wrong, I am knee deep in terms of working with dots and not planning to switch to unreal anytime at all really(unless they were to announce c# support which is looking like never), but I always wince when I see statements like that because the past in terms of demos featured and what actually filters back into the engine.

 

Keep in mind all the tech demos from Unity 5 and Unity 2017 don't work anymore with the modern Unity render pipelines (URP and HDRP). They have a host of issues I've read about with even vanilla versions of Unity 2018-2019.

On the otherhand, all the tech demos for UE4, from the kite demo to all the paragon assets, still work to this day on the most modern version of UE4.

All the tech demos Unity has released (you can't even use the assets in commercial games) are nothing but smoke and mirrors, while all the assets from UE4's tech demos, you can use in any commercial product.

 

You have an odd choice of wording there. Smoke and mirrors almost always refers to the tricks necessary to make effects look believable without requiring massive performance penalties. If anything effects being successful smoke and mirrors should make them more valuable for commercial use.

 

It's smoke and mirrors because the demo often times doesn't reflect the quality you can actually achieve with Unity out of the box.

Even if that's a bad analogy, I'd still give Epic Games credit for allowing you to use most their tech demo assets in your commercial games.

What's the point of releasing a tech demo like the Blacksmith, if it'll be out of date in only 3-4 years? Why not make a tech demo or game that'll be relevant for game developers for a decade.

 

Two reasons. One, you can't truly make a tech demo or game that'll be relevant for a decade. Crysis was about the closest we've seen and while it looked good at times during the mid 2010s it was nowhere near cutting edge because while Crytek predicted some of the cutting edge features that would come out there were still more they didn't.

Two, you don't want to make a tech demo or game that'll be relevant for a decade. The "Can it run Crysis?" meme continues to exist to this day not just because Crysis was extremely difficult to run at the time it came out but because it continues to be difficult thanks to the architecture moving in a direction they didn't predict.

Hardware is constantly evolving. We don't know where it will be in a decade. We thought we knew for years with Intel constantly releasing quad-core processors but then AMD shows up out of the blue with eight cores that rapidly became sixteen and will be thirty-two soon. All while increasing single-threaded performance.

If Nanite lives up to the hyper its receiving it will be an outlier. Lumen is only valid right now because the hardware we have is too weak for proper raytracing. Once that's no longer the case traditional methods like Lumen will fade into obscurity as raytracing doesn't require any baking while being much more realistic.

 

Honestly, most of them start to break down a single in-cycle point release. Stuff that targeted 2017.1 often doesn't even work in 2017.2. I seem to recall this was an issue with I want to say Adam and its light shape stuff.

 

Lumen is still raytracing, just not HARDWARE accelerated raytracing and not BHV based raytracing. In fact BHV accelerated raytracing could probably benefit from lumen's acceleration structure too, and I bet hybrid renderer will outperform vanilla raytracing anyway.

 

Alright, the tech inside the tech demo doesn't need to be impressive for a decade. I understand that the Kite Demo and Blacksmith are both antiquated compared to next generation tech demos.

What I'm saying is The Kite tech demo can still run perfectly fine in UE4.25 right out of the box, while the Blacksmith will crash and burn if you try to run it on Unity 2019.3.

A tech demo that's relevant for years to come doesn't have to impress gamers for years to come. It just needs to be relevant for game developers and easily usable for at least half a decade. There's absolutely no reason to put precious man hours behind a project that will be unusable in Unity one year from now.

The 2018 HDRP tech demo (with the forest) already has trouble running on modern versions of Unity. All those man hours could have been put towards free assets for all game developers, or better documentation, better tools.

 

Sure but why do that when instead they can charge $250 or whatever for the "icon collective" packs, a bunch of assets that were so specifically tailored to their end results that using them was practically impossible, to say nothing of the absolutely bonkers marketing that never really gave a clear impression of what each pack contained?

You know, these!

https://blogs.unity3d.com/2018/11/1...d-art-content-from-the-unity-icon-collective/

Fun fact: these ridiculously expensive packs do not actually work with HDRP properly anymore.

 

"For this project we collaborated closely with Quixel Megascans, the world’s largest library of scan content, who provided us access to beautiful photogrammetry scanned environment assets from their Lava Field biome to integrate within the scene."

And now Epic Games allows you to use Quixel assets in UE4 for free lol! This has to be a joke. Why spend $250 on assets that won't be supported for more than a year?

 

I got the Buried Memories 2 HD pack, the Construction HD pack and the Asian Urban HD pack from the humble bundle a few months back. I wanted to use them in a game I was making, but apparently they don't work anymore without needing to relink a lot of GUIDs. The official solution is to use 2018 LTS.

 

I'm also quite frustrated by this, like how Occlusion Probes never made it into HDRP, but I have a feeling this is comparing oranges to apples here.

Unity's Book of the Dead assets were released to the public in 2018 when HDRP wouldn't be finished until late 2019. Meanwhile, UE wouldn't possibly release the UE5 demo assets until UE5 comes out in 2021. If they were to release the assets now, while Nanite/Lumen is still in preview, surely they would need custom code as well to make the experimental tech work, and surely it also wouldn't be compatible with future versions of UE5 due to the custom code?

Unity's Book of the Dead was showcasing future tech in development, and we all got angry about that, but we don't get angry at UE's demos also showcasing future tech in development?

Basically, I feel like this problem comes from Unity simply releasing their demo assets too early, which is both a good thing and a bad thing. If UE also released their assets early, they would run into the same problem too.

Please correct me if I'm wrong here! Thanks

 

Quixel already released the environment assets used in that demo on 13th of may...

 

I mean with Nanite/Lumen enabled.

 

Im developing a game for my children in my free time, and maybe some day if I hit some quality I try to sell the game, but obviously i don`t have the experience of many people here.

After read many post here, ignore some trolling and read some fun things, some people (i think) here need to see the POW of other developers, some are developers working in companies, others are freelancers, others like me are doing projects in the free time for money or not, so "what i want" change a lot in all cases.

More or less what I see is:

- People here want the improvements of UE in terms of workflow, QUALITY, AND LICENSING, because are tired of Unity that fell like you have no tools or broken/never finished tools and I know this is frustrating ("Released final stable" HDRP and no grass.. for months?, features available in the editor but no in compilation like change mecanim mask on fly?, multiplayer hello?)
- People that don`t want changes and then come URP, HDRP, DOTS, etc, and the fear of lose months/years of work and get "deprecated" with a message in some forum that no one read, and some day "hello you code is not compatible"
- People that love this changes and embrace it, and think all other can change the workflow/knowledge/code easily like all is doing only free mobile games for Android because are more a developer than a designer.


So if we put all in the desk, and try to see all together, what I see personally.. is a big issue in how Unity is managed in terms of "things to focus".

First of think, Unity is not Unreal, you cant get the same work of a "small company" vs a company with Tencent back (f*** Net income: 36.39 billion USD (2017) by Google), if they try to focus in the Services and the main target of Unreal ("Do AAA games" in terms of quality AND WORK) are gonna die.

They need to focus, focus on finish currently tools and give real stability to allow the people that work with Unity trust on it, you can`t think seriously that Unity is for big projects in something that in the LTS version have bugs that never going to be fixed.

Another thing i think was a big mistake is split the Render Pipelines, i know, i know why they do it, exactly for the reason they cant focus on many things, but right now that many people claim, now we face the issues of you need to stick and never change the RP, you can change it obviously, but with a risk of finish with broken project full of pink assets and red lines from the scripts, and this is unacceptable for big projects or people without time. And in other hand there is the Asset Store, "hey look that nice water asset in offer, i want it!" then "not compatible with HDRP.. f***", half of assets are broken, assets developers need to make the work x3, and in many cases simply ignore the other RP, this is filling the Asset Store of frustration.

Obviously they need to do new things (improve or die), but if this deprecate the base of the actual code, and break the MAIN POINT for many many developers to stick with Unity, this is a shot in the feet, because, "if i need to learn again, why not learn Unreal?" they need to do carefully or can be one reason to die. Think in that, what happen if tomorrow Unreal announce that add C# compatibility in UE? What you think the half of Unity community not working in 2D or Mobile project gonna do?

I think the DOTS idea is not bad one (use main thread for all is a crap), IF allow to use together with the actual code without issues (and don`t deprecating silently the things..) if some improvement can be added and is asked for many devs, ADD IT when is finished if this don`t impact the main version.


And please, ask your users, do a poll don´t kill you, and ask your public "what they want" maybe, maybe... allow to focus and work on things that your public want. Or maybe like others say before, Unity is focusing in other kind of public, and small companies and devs are not the main target now, I don´t know, in that case...


In my case, I like Unity for the simplicity, I´m a web developer that works with C# and other languages to make Windows/Linux applications in free time, so I have C# experience and know other languages better, but i don`t have to much experience in Blender or Design, so I feel like Unity is done "for developers to developers", and this have good things (control ALL you want with code!) and bad ones (like Unreal is years ahead in workflow to NON developers profiles, thats why is best for AAA games that sell by the graphics).

For the people that say "you can do the same result with Unreal and Unity" no please... don`t lie, this depends, here is a game done by ONE guy: Bright Memory Infinite (look on Youtube), tell me, you can do the same WITH THE SAME RESULTS AND IN THE SAME TIME with Unity? let me change the question... you think this guys can do THE SAME in THE SAME TIME with Unity? like I say, Unity is not S***, but the Unreal workflow for people that are f*** artists is better than Unity and the lack of tools to improve time and quality. (And Unity is working hard, HDRP, Shader Graph, VFX, etc)

So, like many people here, i love it, but I hate it at the same time, not to be less than Unreal, to don´t focus on make happy the people that want to use it. (and I´m not talking of me)

Sorry for the block of text.

 

This article is a little lacking in factual analysis and has dubious benchmarking methodology, you do not test the effect of latency on a system by running an application specifically optimised for cache optimisation.

But you are actually right, here's a paper with real tests:

https://arxiv.org/pdf/1902.07609.pdf

GDDR5 specifically tried to balance latency and bandwidth and has comparable latency to DDR3.

 

Epic Games has 1000+ employees by 2019 estimates while Unity Technologies has 2000+ by 2018 estimates. Epic also develops Fortnite and the majority of their developers' time is spent on that since that is their main money maker right now.

Just wanted to give people a comparison point since everyone seems to think that Epic's Fortnite money is magically transforming into lines of code.