The scalability and evolution of game engines *spawn*

[iroboto]

You might want to tell this to the folks clamouring for Bethesda to rewrite CreationKit. Like most engineering endeavours, it's often simpler to modify (or fix) something that already works (or mostly works) than start from scratch and never more than when addressing something very complex. Modern engines are fairly modular and we don't need to throw the baby out with the bathwater. It's the same reason that genuinely new design airframes for aircraft are a rarity, they're mostly variations of something already proven.

We saw a lot of genuinely ground-up new engines last generation because Xbox 360 and PlayStation 3 were radically different architecturally but this generation we're seen more iterative building on existing engines. I think people get too caught up in he name of the engine and assume that because it has the same name, it's not really different but that's necessarily the case. The commercial engines don't change their name because they have brand recognition and for companies using internal engines, I don't think they change the name because it really doesn't matter.

Curious to see creation kit going. I think they’ve put so much focus on gameplay side of things and let graphics fall behind. Which is fine, visuals aren’t actually needed for long term play, gameplay is.

It’s one of few games I know where you can hot drop a whole bunch of stuff into the game at run time and that matters. Run time virtual texturing hits hardware limits faster than streaming virtual textures for instance.
Not to mention just being able to overload the world with stuff.

for reference the players can add and delete world items at run time. Shadows will be added etc Lighting, Time of Day.

It may not be a looker, but it’s pretty dynamic. Probably amongst the most dynamic AAA titles I can think of. Some people compare it to 360 lighting, and yea you’d be right; because everyone else has been baking their GI, and you can’t here. You need real time GI and lighting and shadows. Compute intensive stuff.

Imo ray tracing will solve so much for them, it should allow them to compete In next gen.

 

[Deleted member 11852]

It’s one of few games I know where you can hot drop a whole bunch of stuff into the game at run time and that matters. Run time virtual texturing hits hardware limits faster than streaming virtual textures for instance.
Not to mention just being able to overload the world with stuff.

for reference the players can add and delete world items at run time. Shadows will be added etc Lighting, Time of Day.

I don't think many people appreciate how complicated this is. I've seen people argue, or ask, does it need to do that and the answer is no, you could probably fudge this any one of number of ways but Bethesda don't want to fudge it and their engine is crazily ambitious in ways the average person does not understand and that has implications for the whole engine stack.

They've prioritised immersive world over graphics for quite a few years so it boggles my mind that folks don't seem to realise that. Graphics is not their priority.

 

[iroboto]

I don't think many people appreciate how complicated this is. I've seen people argue, or ask, does it need to do that and the answer is no, you could probably fudge this any one of number of ways but Bethesda don't want to fudge it and their engine is crazily ambitious in ways the average person does not understand and that has implications for the whole engine stack.

They've prioritised immersive world over graphics for quite a few years so it boggles my mind that folks don't seem to realise that. Graphics is not their priority.

Yup. Well ray tracing will solve that however. Power will be heavy but I think we’re going to see RT come in a big way for them for next gen. That will make all of this sing. They can toss resolution to do it.

which is great. This is what I love about RT. I called it the equalizer between studios, because studios with insanely high production budgets can’t be caught. But you can’t get better than RT, so... people can now have the best at the expense of asking for power.

 

[eloic]

Yup. Well ray tracing will solve that however. Power will be heavy but I think we’re going to see RT come in a big way for them for next gen. That will make all of this sing. They can toss resolution to do it.

which is great. This is what I love about RT. I called it the equalizer between studios, because studios with insanely high production budgets can’t be caught. But you can’t get better than RT, so... people can now have the best at the expense of asking for power.

I guess you may be right IF they use RT just for lighting and shadows. I'm not sure next-gen machines are able to use RTRT for both lighting and reflections, on top of a very complex scene full of NPCs in an open world... I wish I am wrong, though.

But honestly, yeah, even though I think RT reflections are cool, if we have to choose instead of having it all, I choose proper lighting. Fake reflections can be quite decent. Bad/not dynamic lighting... not so much.

 

[Shifty Geezer]

They've prioritised immersive world over graphics for quite a few years so it boggles my mind that folks don't seem to realise that. Graphics is not their priority.

I don't know about not seeing it. I think the primary issue is that design is seemingly carrying heavy baggage that could be much improved. We have Fortnite StW and Conan running on Unreal Engine that both support similar world detail and user manipulation, I think (I don't know how persistent they are and when stuff resets).

The main problem with Bethesda titles are the bugs and performance. The more complex your engine, the more numerous the interactions and bugs. If you patch a fix for a bug that's found, you may miss a more serious underlying design issue. Then you patch a few more bugs. Then you find your hull is all patches and no original wood. The issues around Fallout 76 smell to me like legacy system pushed past their usefulness, and I am confident a ground-up engine designed for modern systems with fast IO would be worlds apart in performance and stability. I also question whether the design choices for an object-oriented world model will adapt well to RTRT and future lighting. And then designing for good network play, you are better off designing that in from the beginning rather than adding it on top of an engine.

Basically, ideas have completely changed since Elder Scrolls began. Entity based development instead of object oriented design is 10x - 100x faster, but you have to design for it. The future of rendering isn't so much meshes and objects but spaces and optimising for spatial representations. An optimal open-world game is going to need the most efficient database backend and a completely different renderer to what we're used to, and the chances are these are so intertwined in Creation Engine they can't be retro-fitted in.

Fallout 76 launched as a buggy mess missing key features of the engine like NPCs, and it's taken significant time just to get it to where it should have been at launch. That is, the engine support Radiant AI and has for years, yet it's not present in FO '76. Why? Because the adaptations presumably broke it. I think FO '76 points to the devs wrestling with the engine, and if the same period had been spent building something new, the game may have taken longer to release but it'd work better and be on a far stronger footing for new games over the next 10+ years. In essence, I think it's typical business decisions to keep the old thing working than update to new and improved.

 

[Shifty Geezer]

Yup. Well ray tracing will solve that however. Power will be heavy but I think we’re going to see RT come in a big way for them for next gen.

RTRT isn't going to be fast enough for that, and to get the most from it, you'll want spatial representations that can be heavily optimised. Games built from the ground up for RTRT will likely look a half-gen above engines where RTRT is added on top, by my guess.

 

[iroboto]

RTRT isn't going to be fast enough for that, and to get the most from it, you'll want spatial representations that can be heavily optimised. Games built from the ground up for RTRT will likely look a half-gen above engines where RTRT is added on top, by my guess.

I think it depends on how much rebuilding you can do. If you opt to support traditional lighting models as well as RT models I would agree.

I think if you opt to only support RT then you could be better optimized.
But I agree there probably isn’t enough power to do everything, but there should be enough to get them to a much better place than they are now.

 

[Globalisateur]

Indeed, there's a base improvement of 2.7 - 2.8 times the memory capacity and 150x faster storage. Even multiplatform games should look amazing, given time for developers and artists to rebase their toolsets. Though there's always the matter of base PC systems dragging down the consoles.

Well, only if they don't release their game on Lochkart. From 7.5 GB to 14GB is a lot. Or they are not going to use all of that 14GB on PS5. I predict some (japanese) multiplatform devs won't release their game on Xbox anymore. I'll stop being annoying about this now but I am pretty sure we'll talk about this later in the gen. Games like FF16 or MHW2 could not be released on Xbox (or only later) because of this.

 

[BRiT]

Well, only if they don't release their game on Lochkart. From 7.5 GB to 14GB is a lot. Or they are not going to use all of that 14GB on PS5. I predict some (japanese) multiplatform devs won't release their game on Xbox anymore. I'll stop being annoying about this now but I am pretty sure we'll talk about this later in the gen. Games like FF16 or MHW2 could not be released on Xbox (or only later) because of this.

Then I guess they won't release on PC at all.

And yes, memory quantity is still a concern of mine.

 

[dobwal]

Well, only if they don't release their game on Lochkart. From 7.5 GB to 14GB is a lot. Or they are not going to use all of that 14GB on PS5. I predict some (japanese) multiplatform devs won't release their game on Xbox anymore. I'll stop being annoying about this now but I am pretty sure we'll talk about this later in the gen. Games like FF16 or MHW2 could not be released on Xbox (or only later) because of this.

Porting is often dictated by market size and user demand. Nobody is going to forego Lockhart if it represents a significant share of the market unless they are underfunded or have non-financial reasons not to support the system.

 

[Unknown Soldier]

Porting is often dictated by market size and user demand. Nobody is going to forego Lockhart if it represents a significant share of the market unless they are underfunded or have non-financial reasons not to support the system.

I could have understood your comment if you said XBSX, but Lockhart? HAHA.

 

[AlNom]

Market size led to certain Switch ports too.

---------

this sorta stuff should be in a different thread.

 

[eastmen]

Well, only if they don't release their game on Lochkart. From 7.5 GB to 14GB is a lot. Or they are not going to use all of that 14GB on PS5. I predict some (japanese) multiplatform devs won't release their game on Xbox anymore. I'll stop being annoying about this now but I am pretty sure we'll talk about this later in the gen. Games like FF16 or MHW2 could not be released on Xbox (or only later) because of this.

your still going from a traditional platter mechanical drive to a 2.5GB/s ssd and thats before compression that ms is using and any ML done on the textures or other tricks. So even Lockhart will be a big step up vs the ps4 and xbox one.

If all the fud about japan wanting small consoles that we've heard of since the original xbox relased is true then they may not show up for the ps5.

 

[Lalaland]

I'd somehow missed the earlier leaks about Lockhart's RAM, that does seem like a pretty major gap whether it results in games on Lockhart being seriously compromised likely does come down to how successful it is. If Lockhart customers are a small proportion of your overall user base I can't see devs doing much more than the bare minimum, for devs to make significant efforts at Lockhart optimisations MS will have to have a majority of the market and Lockhart will have to be a major fraction of that.

 

[BRiT]

I'd somehow missed the earlier leaks about Lockhart's RAM, that does seem like a pretty major gap whether it results in games on Lockhart being seriously compromised likely does come down to how successful it is.

I don't think it'll be that bad, because resources will be smaller, possibly 1/4th the size if targeting 1/4th the resolution, and with the same speed SSD as Series X but loading less will have Series S having faster loading/streaming performance than even PS5. So if the PS5 can stream things dynamically on demand then so too can this theoretical Lockhart.

 

[dobwal]

Lockharts memory will probably be a lot easier to deal with than Durango’s.

What’s 7.5 gddr6 vs 13.5gddr6 when devs are coming from 68 GBs of DDR3 + 32 MB of esram vs. 176 GBs of gddr5. LH vs. XSX vs. PS5 is no different than PC hardware tiers.

Now consider the previous 2 console gens where devs had to deal with esram, edram, split memory pool vs unified, and XDR vs gddr3. I’m sure those two gen represented more memory management nightmares than just a straight difference of less memory.

Devs aren’t even under any pressure to make LH’s game look as good as others. Not in the same way they were when it was the one vs. PS4 or 360 vs. PS3.

Im guessing Epic can more readily scale UE5 across consoles, pc and mobile because the consoles (NS is more of a mobile design) and PC are basically the same. Where as two gens ago MS, Sony, Nintendo, PC and mobile were all supported by varying archs mostly unique to each platform.

 

[Lalaland]

I think we're debating what is "significant" when it comes to differences between XBSS/LH vs XBSX, I don't think it will be difficult to create a XBSS version of a game as tool chains are significantly more mature at dealing with multiple targets (especially with the .5 gen consoles). We can already see what happens when the performance profile for a title skirts the limits of the boundaries between the two consoles with multiple titles lately coming with more inconsistent frame rates under load such as with Control on PS4 vs Pro but the idea of launching with that gap is very strange to me.

I mean at launch MS is asking devs to create 4 performance profiles for Xbox platform titles (XB1, XB1X, XBSS, XBSX) which I will imagine most will handle with multiple texture resolutions, varying target resolutions, game engine settings and tolerance for performance variability. The first two of those factors are fairly easy to adjust and offer the most gains for least effort, it's the third factor that takes the most time and energy as you discover whether using 1/4 res on alpha buffers is as effective as fiddling with this or that LoD horizon. All of those factors are ultimately ruled by the fourth, how willing are devs or publishers to see releases for non-target platforms ship with significant framerate issues. For me the signs on that are overwhelmingly negative, even in the scant 4 years since the introduction of the PS4 Pro/XB1X the "upgrade" platforms have become the base platform for performance tuning with most companies more than willing to leave owners of the base consoles dealing with a degraded experience. Now MS is proposing to sell a half step console on day one, will XBSS customers ever get their day in the sun when their platform is treated as the base for performance tuning? How bad will the performance be for them in 2 years? Or in 4 After the next .5 gen console launch?

 

[BRiT]

Microsoft isn't asking 3rd party devs to do anything about supporting current-gen consoles. That's their own decision, even with first party titles that was a decision for the devs.

Why do you say its a half step console when it has a faster cpu than PS5, may have better resource streaming than PS5, and supports all cpu and graphical features as series x?

 

[Globalisateur]

Microsoft isn't asking 3rd party devs to do anything about supporting current-gen consoles. That's their own decision, even with first party titles that was a decision for the devs.

Why do you say its a half step console when it has a faster cpu than PS5, may have better resource streaming than PS5, and supports all cpu and graphical features as series x?

You can't be serious.

 

[BRiT]

You can't be serious.

Its because they won't be targetting 4K or 8K resources but resources 1/4 the size, essentially giving them a 4x multiplier in streaming speed. Or using resources 1/2 the sizes gives a 2x multiplier in streaming speed.