Predict: Next gen console tech (10th generation edition) [2028+]

[DavidGraham]

Is it really when they're releasing a Lego Horizon spinoff game on the Switch ?

Yes it is. One game on the Switch is hardly anything significant.

How many of those UE5 games feature high scene complexity (AKA aren't releasing on last generation or started development w/ last generation technology) ?

Several. Silent Hill 2, Until Dawn, Gears E-Day, Black Myth: Wukong, Legend of Ymir, ..etc.

Just because I don't see much value in their technology doesn't mean I don't see value in Juan's statement which was the principal basis I was bringing up

You took his statement out of context, he only made it in the context of his engine embracing path tracing instead of virtual geometry.

 

[Shifty Geezer]

You took his statement out of context, he only made it in the context of his engine embracing path tracing instead of virtual geometry.

I disagree with that. His main opening statement is "[in response to virtual geometry being added to FOSS] I don't think this is that useful compared to traditional LOD."

However, if the creator of Godot isn't a major authority on high-end game engines or rendering or ray-tracing, unlike, say, Carmack, it's just a POV. Furthermore, Linietsky even states Godot isn't going for a cinematic renderer, so he's not even looking into it and this is just his current thinking. Once he starts trying to combine realistic lighting with super-high detailed geometry, he might start thinking differently and finding other solutions.

 

[Lurkmass]

Yes it is. One game on the Switch is hardly anything significant.

For all of Sony's talk about "believing in generations", there's seemingly a severe lack of exclusive current generation defining AAA games by the other major publishers with dismal results so far ...

Bandai Namco - 1 (Tekken 8)
Capcom - 1 (Dragon's Dogma 2)
EA - 6 (Need for Speed Unbound/Dead Space/EA Sports WRC/rest being sports games not on PC)
Koei Tecmo - 2 (Wild Hearts/Rise of the Rōnin)
Square - 3 (Forspoken & FF XVI/VII Rebirth)
Ubisoft - 3 (Avatar: Frontiers of Pandora/Skull and Bones/XDefiant)
Warner Bros - 2 (Gotham Knights/Suicide Squad: Kill the Justice League)

Some of what were listed above are still using UE4 and there's been absolutely NOTHING from the likes of Take-Two Interactive, Sega, or Tencent. Even Microsoft are better proponents for next generation entertainment than Sony are ...

WHERE are the current generation AAA games with current generation technology ?! Maybe that's why the AAA game industry have been taking a beating as of recent since they're only interested in slapping on RT with last generation quality content so that they can release it on those systems or reuse last generation technological foundations ...

Several. Silent Hill 2, Until Dawn, Gears E-Day, Black Myth: Wukong, Legend of Ymir, ..etc.

Bold of you to list games that don't even have release dates yet, no gameplay footage, or don't even specify 'hardware' RT ...

You took his statement out of context, he only made it in the context of his engine embracing path tracing instead of virtual geometry.

Respectfully, I did not as he exactly implicates the mutual exclusion between the choice of virtual geometry and ray tracing ...

 

[trinibwoy]

I meant that statement to be in terms of game integration ...

Isn’t RT far ahead of any other next gen tech in terms of engine and game integration?

There's a major difference between tech that's been extensively tested with more flaws being discovered over time in comparison substitutes that haven't been tried yet ...

Agreed. We should be very skeptical of PowerPoint slides.

@Bold That's not how feature development works. Hardware vendors absolutely NEED to show us HOW they can realistically get there!

Yes of course. It will take both software and hardware. I think we’re well beyond the prototyping phase for RT. We’re in the evolution and refinement phase. Alternatives like software Lumen are in the same phase but have a lot less potential for providing a comprehensive and general purpose solution IMO.

 

[DavidGraham]

I disagree with that. His main opening statement is "[in response to virtual geometry being added to FOSS] I don't think this is that useful compared to traditional LOD."

Then he goes on to express his views that realistic lighting + RT are enough.

https://twitter.com/x/status/1809638038135329081

Linietsky even states Godot isn't going for a cinematic renderer

He is.

https://twitter.com/x/status/1809936888788054429

Bold of you to list games that don't even have release dates yet, no gameplay footage, or don't even specify 'hardware' RT ...

Only Gears has no gameplay footage or a release date, the rest have release dates, gameplay footage and confirmed RT. Gears also specify "hardware" RT.

 

[Shifty Geezer]

He is.

https://twitter.com/x/status/1809936888788054429

That discussion is a theoretical. He states the engine has been modernised which is a prerequisite for attempting a cinematic renderer. He then goes into the hypothetical, "okay, let's say everything ready to do a cinematic renderer". "We should, we need to plan, we could..." Not, "we are, we chose, we will be..." He then reiterates that AAA-quality cinematic renderers and assets isn't Godot's audience, it isn't a priority, they aren't catering to AAA devs, and any Godot design choices will be around what their indie dev users need. Godot is not trying to compete with Epic and will not be chasing solutions to match Epic's but instead doing their own thing with their own priorities.

 

[Lurkmass]

Isn’t RT far ahead of any other next gen tech in terms of engine and game integration?

That doesn't mean we're not seeing some pushback against it and do you think it won't see more resistance w/ competing technologies over time in the future ?

Right now there's not a lot of persuasive uses for compressed geometry format rendering in mesh shading when you can be easily rasterizer limited by the hardware since an occupancy as little as ~25% w/ mesh shaders is enough to be bottlenecked on AMD. Procedural rendering is another possible use case for mesh shading but you can encounter state change bottlenecks so I'm hopeful that changes with the introduction of GPU-driven PSO swapping w/ the mesh nodes extension for Work Graphs. Perhaps Work Graphs were the missing component all along to see more uptick in virtual geometry or mesh shading to stem the proliferation of RT ...

Yes of course. It will take both software and hardware. I think we’re well beyond the prototyping phase for RT. We’re in the evolution and refinement phase. Alternatives like software Lumen are in the same phase but have a lot less potential for providing a comprehensive and general purpose solution IMO.

The same process applies for extensions or feature enhancements too and it's not invariably any easier either!

Would you define stagnant API specs and feature (DXR) for years now to be in an "evolution and refinement" phase ?

 

[trinibwoy]

That doesn't mean we're not seeing some pushback against it and do you think it won't see more resistance w/ competing technologies over time in the future ?

I wouldn’t call it pushback. Competing technologies would be great but right now they’re all paper tigers and don’t actually claim to be equivalent to RT in compatibility or quality. So not really competing. At best they’re cheaper alternatives that provide “good enough” results with lots of caveats.

Perhaps Work Graphs were the missing component all along to see more uptick in virtual geometry or mesh shading to stem the proliferation of RT ...

Why do you want to stem the proliferation of RT?

The same process applies for extensions or feature enhancements too and it's not invariably any easier either!


Would you define stagnant API specs and feature (DXR) for years now to be in an "evolution and refinement" phase ?

Yes I would. That stagnant api is still producing more tangible results in more engines and more games than anything else. So while it would be nice to see things moving faster it’s still the best we’ve got.

 

[Remij]

Figured this could go here since there's a lot of RT discussion itt.

https://twitter.com/x/status/1813318841746981370

 

[Dictator]

Figured this could go here since there's a lot of RT discussion itt.

https://twitter.com/x/status/1813318841746981370

Great to see AMD responding to a number of the problem cases for current hardware and APIs.

 

[fehu]

Can this work with nanites?

 

[Frenetic Pony]

Figured this could go here since there's a lot of RT discussion itt.

https://twitter.com/x/status/1813318841746981370

Hmm, the DGF (hardware compression support for meshlets w/RT support) thing looks like a proposal for something that could show up in RDNA5/Xbox whatthefuckever. Still doesn't allow skinned geometry, but still kinda neat.

 

[Remij]

Can this work with nanites?

https://twitter.com/x/status/1813393591030702460

Click that and read his responses.

 

[cheapchips]

https://twitter.com/x/status/1813393591030702460

Click that and read his responses.

His replies, so you don't have to click though:

 

[Lurkmass]

7 CONCLUSION AND FUTURE WORK

We propose DGF, a block-based format for efficiently storing dense geometry data. Its design and structure have been optimized for ray tracing specific use cases. Although it offers a lower density than competing displacement-based lossy compression approaches, it is able to support any mesh topology, and offers higher density than other meshlet-based lossy compression formats. The ability to quickly encode any topology allows for an easier integration into existing asset creation pipelines. Even though software-based intersection of DGFs is currently slower than a simpler uncompressed leaf data representation, future direct decoding support in hardware would completely offset this current limitation, and unlock substantial memory footprint and bandwidth reductions. In the future, we are interested in simulating different variants of direct hardware decoding support for DGF and in improving the encoding algorithm

Yikes! Their software implementation (intersection shaders) of this feature is slower (check table 5) than ray tracing against uncompressed geometry by a factor of up to ~2.4x and this doesn't really help address the acceleration structure update cost either. It's going to be a shaky proposition to persuade the hardware team to implement support if it doesn't really improve performance ...