Game dev isn't my field, I always find it fascinating though.

Why can't Unreal handle something like CP2077? Could be an "explain like I'm five" - details would be cool but I'm not sure I would even fully understand them, haha.

I'm not an expert on the topic either, but from similar discussions I've followed, the answer tends to be that out of the box and with the base settings, an engine like UE5 will be optimised for certain use cases. Different genres will have different specific needs, and may require your developers to do a lot of heavy lifting to make the engine performant under their specific conditions. A good example might be sim racing titles, which generally have the player moving a lot faster than the default scenario for an engine (which is mostly a slow-moving FPS, for example). At these relatively high speeds, the default antialiasing methods may result in serious graphical artifacts, and at the rate at which physics needs to be calculated (90-300 hertz) the default physics of the engine may be less useful, and you're forced to implement your own from scratch. That's not to say it's impossible, but you're opting out from a lot of the engine features, or in some cases, it's actively making your life hard.

Awesome answer. Thanks for those examples, makes a ton of sense!

I just assumed in 99% of cases, you could just "tweak" Unreal but if its issues with core components of the engine... I could see why an engine from scratch may work better.

"can't handle" here means "can't handle performantly".

Think of it like using the wrong data structure for the problem.

And not just for the game itself but for creating the world in the first place. Often you need radically different data structures for the creation and editing of the data as compared to what you need for running the game on player's machines.

In Cyberpunk 2077: you can go from one end of the open world to the other, without ever seeing a loading screen, you can even install a mod that allows your car to fly and you can see the whole city from the top, I haven't seen a game like that in UE5 at that level therefore there is a possibility that it would struggle with something like that if it's not designed to cope with it.

Video games use a trick called Level Of Detail, which means if something is far away, you only render a very low definition version of that object, UE5 has it as well, however we don't know if it's capable enough for something like Night City, obviously the RED engine was built with a very hardcore LOD system that might be a massive pain to do in UE5.

So the investment into seeing if UE5 would work with something like CP2077 would be greater than the investment into a new engine? Or perhaps the risks are greater? I could see getting 75% of the way there and realizing its starting to buckle would be pretty detrimental.

Looking at the Witcher 2, 3 and CP2077, I'm highly skeptical that the engine was the problem.

I also have my doubts if they would haven been able to do the same thing better on older consoles with Unreal Engine.

Most of the worst issues with CP2077 have nothing to do with the game engine, things like the NPC AI and broken difficulty curve, the rough melee combat and hiring a shady testing company.

The one thing that could be blamed on the engine is the potentially difficult tooling for the content creators, which can easily be solved, just look at the amount of mods out there.

From the outside, this looks more like a business decision rather than a technical one, that's not to say that it could not work out for CDPR if Epic works with them very closely, however I'm very annoyed when games journalists imply that CP2077 would have been amazing at launch if they would have used Unreal.

If Unreal Engine 5 can't handle a world like CP77, why are they making the sequel to CP77 in Unreal Engine? Are they betting on the engine being ready for CP77's sequel when they need it to be?

Interesting! I’m not at all familiar with game engine design or even game dev, but the only game series I can think of that reliably handles large, open-world scale well is GTA.

Here’s a dumb question: in theory, would Rockstar’s internal engine (RAGE) have worked for CP77, or are the technical challenges simply too different?

The easiest dumb answer is that Rockstar has never licensed RAGE to an external developer and presumably doesn't develop the engine with that in mind at all so it is probably under-documented and hard to work with for any team that isn't Rockstar.

Also, I think there are other engines that do as well as RAGE, but none of them are licensed for external use either and are secret sauce to their various developers.

I think that’s fairly self-evident, hence my inclusion of “in theory” :)

Are there other internal engines that have been used for open-world games with detail and scale at the level of GTA or RDR?

Just off the top of my head:

Volition's multiple unnamed/unbranded engines for the Saints Row franchise.

The RenderWare fork used for Crackdown 1. The unknown engine for Crackdown 2 (presumed to be an iteration of the Crackdown 1 engine). (Crackdown 3 used Unreal Engine 4. Useful to mention for those elsewhere in this thread concerned Unreal "can't" do it.)

You can argue that BioWare has tried its raw best to push EA DICE's Frostbite engine as much as possible in that direction. DAI and MEA didn't have truly seamless open-world games, but they tried to get the biggest scale of levels that Frostbite seemed capable of and at their best felt "close". (Obviously at their worst they felt nothing at all like GTA/RDR level of detail/size.)

Thanks for the list!

How is Unreal limited in a way that prevents it from doing something like CP77?

Do they license their engine? Building it just for CP77 would be a shame.