• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

[DF Article] Is Unreal Engine 5 really the best choice for the next Witcher game?

3liteDragon

Member
Digital Foundry on what CDPR gains - and loses - in the tech transition.
CD Projekt RED has officially confirmed the existence of a new title based on The Witcher. In an announcement that reveals no concrete information on the game itself, the focus is very much on technology. The firm's in-house REDengine is being mothballed in favour of Epic Games' Unreal Engine 5. CDPR talks of a 'multi-year strategic partnership... [covering] not only licensing, but technical development of Unreal Engine 5, as well as potential future versions of Unreal Engine where relevant'. In essence, CDPR has chosen to throw in its lot with Epic Games, even though core Unreal Engine 5 technologies are still at an early state, with the firm helping to shape the tech 'with the primary goal being to help tailor the engine for open-world experiences'. We're at the embryonic stages here as no development time frame or release dates are being shared at this time.

It's a seismic shift for the Polish developer, which has shipped all of its titles from The Witcher 2 onwards on its own engine - but to be clear, while the REDengine's days are numbered, it is still the core foundation of Cyberpunk 2077, so the upcoming expansion for that title will still be based on the same technology. Beyond that, however, it seems that Unreal Engine 5 is the future. From the perspective of the staff members at Digital Foundry, this is not the best news and while it's clear that there are reasons why CDPR has moved on, the end of development on a hugely impressive independent engine is a blow. After all, REDengine in its various guises has produced a visually unique presentation that has undoubtedly helped to shape some brilliant games. Going all the way back to The Witcher 2, CDPR delivered a title quite unlike any other that genuinely pushed back barriers - a game that still looks brilliant today. Environment detail, lighting, character rendering and post-processing were well ahead of their time. The Witcher 2 launched during the PS3/Xbox 360 era yet from a technological perspective, it was a class apart, a generation beyond.

While it's clear that Cyberpunk 2077 has suffered from a range of technical problems, from our perspective, the PC version has always been performant and scalable, and always delivered the absolute state of the art in game visuals from day one. Bugs were undoubtedly an issue across the board, but the vast bulk of the technical problems centred on the last-gen console renditions of the game. Graphical aspects are inherently scalable - whether you're talking about features, the quality of those features or the native rendering resolution.

However, the last-gen consoles simply did not have the CPU and storage capabilities necessary in delivering a strong experience. Meanwhile, on PC, a modern CPU combined with an RTX GPU could deliver an impressive visual experience at 30fps minimum with RT features beyond what appeared in the game's actual next-gen patch for PS5 and Xbox Series X. Higher-end PC kit ups the resolution and the range of RT features you could run at a playable frame-rate and there's the sense that Cyberpunk 2077 will continue to scale going into the future - just like The Witcher 2 did back in the day. The need to service the games machines of the past will not be a factor for the next Witcher title - likely several years away from release - so there are two key questions to consider in CDPR's decision to move away from independent engine development. First of all, why not continue to evolve the REDengine and secondly, what advantages does a shift to Unreal Engine 5 offer?

First of all, the perception that the REDengine in Cyberpunk 2077 is an evolution of the same REDengine found in The Witcher 3 should be challenged. On Twitter, ex-CDPR developer Bart Wronski lays it all out. 'Every game they dropped the whole engine, rewrote it from scratch hoping this time it will be better and work, but then due to crunch hacked the hell out of it with it not being maintainable or usable at all.' Wronksi says that aside from some rendering systems, REDengine was effectively re-written from scratch between titles and goes on to suggest that the process of engine development would need to reset again moving from Cyberpunk 2077 to the new Witcher title. With that in mind, he considers moving to Unreal Engine 5 to be a good choice - and there are many good reasons why development should shift. It's built for flexibility, for scalability across platforms and it also grants CDPR access to an enviable 'brain trust' of technical talent. It's no secret that even the most successful developers have had recruitment issues in a world where many of the most talented engineers are increasingly gravitating to the likes of Epic, Unity and Nvidia.

In terms of the advantages of the shift to UE5, that should be pretty clear beyond the obvious workflow advantages, based on the showcase reveals of key technologies: Lumen, Nanite and Metahumans. Lumen is UE5's stunning real-time global illumination system, accurately simulating the properties of light bounce within any given scene. Nanite is the new geometry system developed by Epic that aims to effortlessly stream extreme quality assets at unrivalled levels of density with no draw distance 'pop-in' whatsoever. Meanwhile, the Metahuman tech is Epic's stab at delivering photo-realistic character rendering. The best and most recent showcase for all this technology is The Matrix Awakens demo, which presents almost like an interactive Blu-ray, in its scripted chase sequence at least.

The Matrix demo also serves to highlight another key challenge facing UE5 developers that may affect the new Witcher title. The technology jump delivered by Lumen and Nanite is immense, but it comes at a cost: performance. We must remember that the Matrix demo is exactly that - a demo - but in all likelihood, titles that use both of these key technologies may target 30fps rather than the 60fps that has come to dominate the current-gen gaming landscape. While the upgraded rendition of Cyberpunk 2077 for Xbox Series X and PlayStation 5 didn't quite deliver the bulk of the PC experience as we had hoped, at least its 60fps mode looked good and played well. Hitting that same level of performance in UE5 is likely to require a very high-end PC.

CDPR's arrangement with Epic Games seems to be more than just a mere licensing deal, however. CDPR's statement talks about a collaboration with Epic Games 'with the primary goal being to help tailor the engine for open-world experiences', which highlights a key point about UE5 in general and its potential application in a Witcher title specifically: the fundamental systems at the heart of UE5's innovations are still in development. The Valley of the Ancient demo, for example, highlighted that expansive worlds were possible, while The Matrix Awakens' city focused on the strengths in procedural generation in making those worlds more easy to create. However, nothing we've seen on Unreal Engine 5 yet resembles the kind of environments seen in the Witcher series. Castles, cities, quarries and deserts wouldn't be an issue, but as far as we're aware, Epic Games has yet to solve the issue of integrating non-opaque objects into Nanite, nor skinned geometry. To translate that into plain English, elements such as trees, foliage and characters would need to be rendered in the traditional way in a separate pass to Nanite, subject to the kind of draw distance pop-in and other limitations we see in games today.

That's just the tip of the iceberg in terms of the kind of systems required for an open world game, and it'll be interesting to discover whether the collaboration between Epic and CD Projekt RED sees the kind of skill and expertise that built Cyberpunk 2077 and The Witcher feed back into technologies that will benefit all Unreal Engine 5 developers. That may be what's implied by the concept of a partnership and would distinguish this kind of arrangement from, say, a project like Housemarque's Returnal - where UE4 was the foundation, but the final game was heavily customised with the Finnish studio's in-house technology. The stage is set for the next Witcher game, but we can't help but feel sad about the end of REDengine (or perhaps, the REDengines in the plural). It's a shame because fundamentally, diversity in core technologies results in visually distinct games that can propel visuals into new and exciting territory: Cyberpunk 2077 on PC topped our list of the best game graphics of 2020 - and rightfully so. Visually, it's streets ahead of any Unreal Engine title released to date, though obviously UE5 is a different kettle of fish.

We're moving into uncharted territory with this new tech, but what's clear is that this is one more step towards a homogenisation of game technology - advantageous in so many ways, but also worrying in an industry that thrives on new and diverse ideas and techniques. The best case scenario? The new Witcher game retains the identity of a CDPR title but fully taps into the potential offered by Nanite, Lumen and Metahuman technology. These are all exciting features, but still in flux - and maybe their deployment in a game of this stature will pay off for both Epic and CD Projekt RED, while potentially raising the bar for all UE5-based developers.
I'm just wondering how a game like the next Witcher's gonna be made on UE5 when Nanite doesn't even support foliage rendering yet.
 
It's smart of them, but for unfortunate reasons. They had such a massive staff turnover due to crunch. and lost a lot of people to Rebel Wolves, that the number of people familiar with their Red engine may be dwindling. I used to work at a company which had areas of code that nobody knew how to work because the authors had moved on years before. More people in the industry will have experience with Unreal Engine, which means less training needed - and less risk when people inevitably move on after just a few months due to having their soul crushed.
 
Last edited:

CamHostage

Member


I'm just wondering how a game like the next Witcher's gonna be made on UE5 when Nanite doesn't even support foliage rendering yet.

Nanite doesn't support foliage (it might never work with those types of objects and the tricks that developers use to portray foliage), but that's fine, they can use other methods for those aspects of the game world. In cases where Nanite is used to provide a view miles into the horizon, asset density into the background might prove a challenge where only so much vegetation can be flocked to that area, but again, further tricks and other methods can be used to make the most of the new technology mixed with the proven technology.
 
Last edited:

SlimeGooGoo

Party Gooper
Well, as long as they can customize the engine it shouldn't be much of an issue.

It's better for outsourcing, but it risks feeling to "samey" to other Unreal Engine games.
 

Max_Po

Gold Member
Hasn't made a single fucking game, counts FPS of other people works and writes a fucking article without seeing work or footage ....."OH-PEE-N-ONION"...

 

Buggy Loop

Member
Probably a fortune in R&D would go into a geometry engine that competes with nanite. Not a lot of devs have Fortnite « Fuck You money », UE5 legit has something to offer to most dev houses this gen.
 

Sega Orphan

Banned
Redtech is a massive CPU resource hog. It's really poorly optimised for CPU usage.
Some engines suck at that, same as Cryengine does. Other engines are just really good at it such as IW8, Forzatech and UE.
Only good things can come from this move.
 

CosmicComet

Member
Sure. CD Projekt Red aren't exactly tech wizards. They are pretty much Bethesda level.

I dont trust them with their own engine.
 

Dr Bass

Member
It’s an opinion piece lamenting the death of an independent engine that has traditional produced great and industry leading visuals while acknowledging the downsides to it, even including an ex CDPR engineers take. It’s not meant to be an investigative report.
You don't see the issue with DF writing an article about the technical merits of UE5 vs CDPRs in house tech when not only do they have absolutely zero clue about programming video games, the assumption is that they might in fact know better how to do CDPRs work than ... CDPR? How do you think they would respond to anyone giving them a suggestion on just how to edit or present a video, something you can do with freakin' iMovie?
 

Anchovie123

Member
I dont even understand how UE5 doesnt do foliage. So it can render a chainlink fence with pixel perfect accuracy but it cant do a fucking tree or plant?
 

elliot5

Member
You don't see the issue with DF writing an article about the technical merits of UE5 vs CDPRs in house tech when not only do they have absolutely zero clue about programming video games, the assumption is that they might in fact know better how to do CDPRs work than ... CDPR? How do you think they would respond to anyone giving them a suggestion on just how to edit or present a video, something you can do with freakin' iMovie?
Its a headline to grab attention, the content isn’t like its mansplaining directly to CDPR. So much butthurt over something so trivial.
 

Fake

Member
No. As much I like UE, seeing every company using make their lack personality. Kingdom Hearts 3 is a good example of that. As much as the pixel movie chars got good using UE, the Kingdom Hearts folks really looks bad.
 

LordOfChaos

Member
It's not a big deal, it gives them a breather to work on their engine rather than jumping in with another game, could be good all around.


With the Witcher 3, I always envisioned a next gen version being about the SSDs enabling Ciri's planet/dimension hopping to be more of a core game mechanic, and UE5 is built to leverage that.
 

Hugare

Member
They have such a narrow vision about how good Cyberpunk looks, because it's all made of smoke and mirrors.

It looks good, yes. But the engine is unoptimized as all hell. It's clearly made with spaghetti code.

Rendering crowds are very CPU expensive. And crowd AI is just some mindless dolls walking around with no level of interactivity at all.

After more than 1 year they were able to make the crowd not disapear when you turn your back to them and run away after you fire your gun.

There are no people riding bikes besides the player because they couldnt figure out how to implement it. No car chases. No working police system.

They were suffering with this engine, and it shows. Crazy how stupid DF looks with this article.

Not saying that UE 5 will solve all of their problems, but it will be easier to recruit new employees and to make the game faster

The "problems" mentioned like not being able to render trees yet and etc. will obviously be resolved in no time.

There are no games released yet using the engine, ffs
 
Last edited:

Arioco

Member
I dont even understand how UE5 doesnt do foliage. So it can render a chainlink fence with pixel perfect accuracy but it cant do a fucking tree or plant?


It's not that UE5 can't do foliage, it can do it just fine, but non-opaque geometry is not supported by Nanite at this moment, and foliage and trees are rendered using lots of transparent geometry. You can render all the foliage you want in UE5, but I'll have all the limitation we currently see in other engines (like visible LOD changes). I mean, it doesn't support characters either and we've already seen many moving humans in the demos (thousands of them in The Matrix one), there're just rendered in a more traditional way, without the benefits Nanite brings to the table. Same for the cars, they're integrated in Nanite, but as soon as a car crashes (and deforms as a result) it stops being supported by Nanite and it has to be rendered like any other engine would do.


But let's not forget UE5 isn't even finished yet, by the time it launches it may support all those things. It already shows huge potential, but there are just a few limitations that Epic will have to deal with.
 

IntentionalPun

Ask me about my wife's perfect butthole
I dont even understand how UE5 doesnt do foliage. So it can render a chainlink fence with pixel perfect accuracy but it cant do a fucking tree or plant?

They've focused on UE5's new features and what those new features enhance.. those enhancements right now won't work well with foliage but it doesn't take away from existing good foliage in UE.
 
I'm just wondering how a game like the next Witcher's gonna be made on UE5 when Nanite doesn't even support foliage rendering yet.

Errr... you just... you know... don't render foliage using nanite.

Devs aren't compelled to render everything in a scene using nanite.

It just means that they'll still have to author content the traditional way for those dynamic objects that they can't use Nanite for. But then that's no different to how they currently do it for those types of objects in their engine.
 

iorek21

Member
Games produced on UE usually take a lot less dev time, no?
I mean, it’s very hard to hear about a game on UE being delayed or being very messy performance wise.

CDPR’s experience with Cyberpunk might have changed their minds about some things.
 

ZehDon

Member
Hasn't made a single fucking game, counts FPS of other people works and writes a fucking article without seeing work or footage ....."OH-PEE-N-ONION"...

Well, I'm glad you already posted a clown picture for yourself...

On topic, I'm guessing after the issues they ran into with Cyberpunk, maintaining their own Engine at a AAA level has proven cost or labour prohibitive. They commented that they had to cut things due to engine limitations. I'm guessing after reviewing Unreal, they believed it was easier to simply upgrade Unreal with their open world stuff than it was to upgrade their RED Engine to compete.
 
Last edited:


I'm just wondering how a game like the next Witcher's gonna be made on UE5 when Nanite doesn't even support foliage rendering yet.
And emissive or vertex displacement or color maks or decals or tesellation or parallax…

I mean you can use and mix nanite and normal mesh and all, but epic said more you mix worse is the performance.
 

SlimySnake

Gold Member
As others have said, you dont need nanite to render foliage. You can look at dozens of UE4 forest demos to see just how well they look even in 4-6 tflops GPUs.

The real problem is going to be what Bioware faced with Frostbite. Introducing RPG systems in a new engine from scratch. The engine they have right now needed to be scrapped and rebuilt but i can promise they didnt have to rebuild everything. They will now need to officially code every single system in the game from scratch.

Expect this game to be out in 2027 at the latest.

All that said, it is the best way forward. It's time for custom engines to go. UE5 is a league ahead of anything ive seen and their nanite tech will one day support foliage. Everyone needs to stop rewriting their own engines and just start to license these things, and focus on game development instead of engine development.
 
I think the REDengine looks great from Witcher 2 & 3 to Cyberpunk but if CDPR want to use UE5 instead of their in house engine i feel something is going on there for them to make the switch
 
huh ? that's weird, STALKER 2 is UE5 and that's doing a lot of foliage.

Where did this UE5 can't do foliage thing come from ?




not UE5 in general but Nanite -> UE5's new system for geometry virtualization which allows you to drop in "infinite" poly assets and gets rid of the need for LOD management
 


I'm just wondering how a game like the next Witcher's gonna be made on UE5 when Nanite doesn't even support foliage rendering yet.
Nanite may not support foliage rendering per se, but UE5 come bundled since last december, I think, with Quixel Megatrees, that, while still in beta, altready allows regular people to easily create stuff like this :
 

Barakov

Member
If their in-house tech wasn't causing them issues, there wouldn't be any reason for them to want to switch.
Fair point. After Cyberpunk, they're probably looking for a big hit right out of the gate. Unreal Engine 5 is the path of least resistance.
 
Hasn't made a single fucking game, counts FPS of other people works and writes a fucking article without seeing work or footage ....."OH-PEE-N-ONION"...
This article is from Richard Leadbetter.
I like Richard. He strikes me as a genuinely curious person and I don't really get the perception of him being biased (unlike someone else on DF).
 

Lysandros

Member
This article is from Richard Leadbetter.
I like Richard. He strikes me as a genuinely curious person and I don't really get the perception of him being biased (unlike someone else on DF).
You don’t remember his damage control articles on XboxOne? Or his sceptical interview with Cerny on PS5 more recently? He is the main reason behind closer ties with Microsoft.
 
You don’t remember his damage control articles on XboxOne? Or his sceptical interview with Cerny on PS5 more recently? He is the main reason behind closer ties with Microsoft.
To be honest, I didn't really follow Eurogamer back in XBOne days. Yes, DF has always been tilted towards Xbox, though it might be because they were rooting for the underdog.
They all thought the Series X was going to run circles around the PS5, probably out of sheer ignorance about the bottlenecks in a GPU, and/or out of the fact that Microsoft shared with them a lot of exclusive details about their console whereas Sony told them as much as they told everyone on public videos.
But ever since the consoles came out, I've been finding Richard to be a rather balanced voice.

I mean at least Richard hasn't been going off on social media telling the world the PS5 is "the less advanced console", nor does he make ridiculous claims about Nvidia GPUs like stating the RTX2060 is better than all current-gen consoles.
 
I don't think they decided on this lightly. They probably talked at length with Epic about what is to come for the engine in the coming years and such. So, yea, I wouldn't worry about it.
 

Thirty7ven

Sony make cringe trainers.
What a sad article by DF, shows how much they know about actual game development.

U5 is a powerhouse that will allow the studio to focus on delivering a game instead of struggling to make the tech work. It’s also an engine used by thousands of developers, tech that is crossing over to Hollywood, has a ton of support and is constantly being improved.

It’s going to be easier for CdProjekt to attract devs also, as experience with Unreal engine is much more widespread, and will let team additions hit the ground running.

Cyberpunk was barely holding together because of the tech, yet all these guys care about is “mah Ray tracing” “mah content”.

Kindly fuck off DF.
 
Games need to be developed faster. i dunno about you guys but waiting 5 +years for a new game from the big guys is too much. Always been too much. So anything that speeds up developing of you know, the actual content of the game will always take priority for me. Probably why Fromsoft never changes their jank ass engine.
 
Top Bottom