Current Generation Games Analysis Technical Discussion [2024] [XBSX|S, PS5, PC]

[Lurkmass]

Very excellent point. The methods used here were 'workarounds' for PS4 to get lighting not available by other methods like RTRT. Those techniques don't necessarily map well to the hardware (and middleware) being ported to. And here we get an important reminder that what's on screen does not tell you what's happening to get it there. A game may look worse and run worse than an equivalent on a platform without being incompetent. Indeed, it might be an astounding piece of code for its original purpose.

I don't know if I'd describe it to be a 'workaround' as opposed to it being "intentional by design" since Sucker Punch were well aware that the hardware that they were targeting at the time didn't have HW RT and right on point about there being astounding piece of code because I certainly wouldn't think of inventing a new BRDF model to simulate velvet-like cloth materials or doing subsurface scattering in a forward shading pass ...

Just because a console was an AMD x64 SOC, doesn't mean the software developed for it is an ideal design to transition to a Windows PC. As the consoles become more like the PC, the delta between software design and PC hardware should diminish, resulting in PC ports performing proportionally better. Native PS5 exclusive games should port better.

Yes, as hardware between consoles and PC get closer so too should the software/API gap but I start to wonder if there will ever be 'parity' in this regard between consoles/PC since next generation platforms could potentially undo this progression because console vendors felt like they wanted to introduce exotic hardware features out of the left field for whatever purpose ...

 

[Shifty Geezer]

I don't know if I'd describe it to be a 'workaround'

I use 'workaround' in quotes as it's clearly a design choice, solving a solution, but it's a different choice to that which would be made now on current GPUs, so equivalent to working around the absence of RTRT etc.

Yes, as hardware between consoles and PC get closer so too should the software/API gap but I start to wonder if there will ever be 'parity' in this regard between consoles/PC since next generation platforms could potentially undo this progression because console vendors felt like they wanted to introduce exotic hardware features out of the left field for whatever purpose ...

I think exotic hardware gets used less, and even moreso now everyone is multiplatform. PS5 games and PS6 will be designed with PC in mind, probably even concurrently.

 

[pjbliverpool]

The PS5 version has little to no visual improvements. It’s perfectly fair to compare to PS4 performance.

But then the question should really be why isn't the PS5 performing better since we can assume the PC is using the PS5 native codebase as a starting point.

Anyway, this is all a storm in a teacup until we see how the setting scale at PS5 equivalents. I really hope DF do a scene and resolution matched performance comparison this time as well as it would be really interesting to see exactly how the PC compares performance wise. If the PS5 version doesn't use DRS then it should be pretty directly comparable. There must also be plenty of areas where the 4K mode drops below 60 on the PS5 or else there would be literally no point at all in the performance mode existing. So this seems like a perfect use case for comparison.

 

[Lurkmass]

I think exotic hardware gets used less, and even moreso now everyone is multiplatform. PS5 games and PS6 will be designed with PC in mind, probably even concurrently.

I feel like if there is a convincing advantage to make use of this said hardware, console centric developers (even if they are multiplatform) will abuse it to their own benefit as we currently see with Sony especially when we're talking about a competitive market where some may desire to get this leg up against other players ...

 

[Charlietus]

But then the question should really be why isn't the PS5 performing better since we can assume the PC is using the PS5 native codebase as a starting point.

Anyway, this is all a storm in a teacup until we see how the setting scale at PS5 equivalents. I really hope DF do a scene and resolution matched performance comparison this time as well as it would be really interesting to see exactly how the PC compares performance wise. If the PS5 version doesn't use DRS then it should be pretty directly comparable. There must also be plenty of areas where the 4K mode drops below 60 on the PS5 or else there would be literally no point at all in the performance mode existing. So this seems like a perfect use case for comparison.

The PS5 version doesn't have any drops in the 60 fps 4k mode. It's also running with check board rendering, so not a native 4k.

 

[trinibwoy]

The PS5 version doesn't have any drops in the 60 fps 4k mode. It's also running with check board rendering, so not a native 4k.

Source for checkerboard? DF claims they couldn’t find any dips below 4K.

 

[techuse]

But then the question should really be why isn't the PS5 performing better since we can assume the PC is using the PS5 native codebase as a starting point.

Anyway, this is all a storm in a teacup until we see how the setting scale at PS5 equivalents. I really hope DF do a scene and resolution matched performance comparison this time as well as it would be really interesting to see exactly how the PC compares performance wise. If the PS5 version doesn't use DRS then it should be pretty directly comparable. There must also be plenty of areas where the 4K mode drops below 60 on the PS5 or else there would be literally no point at all in the performance mode existing. So this seems like a perfect use case for comparison.

It was a minimal effort PS5 port. Pouring a lot of work into optimization when native 4k/60 was never possible anyway made little sense. CB 4k/60 was probably a very easy win.

 

[Charlietus]

Source for checkerboard? DF claims they couldn’t find any dips below 4K.

They say so in the DF video

 

[Shifty Geezer]

But then the question should really be why isn't the PS5 performing better since we can assume the PC is using the PS5 native codebase as a starting point.

When we say PS5 native, we're not talking about a bespoke PS5 rewrite, but a port. The engine is still the same is PS4 working the same way to produce the same visuals, just compiled for PS5 instead of being a PS4 executable running in BC mode.

 

[trinibwoy]

They say so in the DF video

Oh so no dips below 4K checkerboard. So what’s PS5 actually rendering at?

 

[pjbliverpool]

The PS5 version doesn't have any drops in the 60 fps 4k mode. It's also running with check board rendering, so not a native 4k.

Then why does the 60fps performance mode at 1800p CB exist at all?

 

[Charlietus]

Then why does the 60fps performance mode at 1800p CB exist at all?

I don't know, even John is confused about it. I played the whole game in resolution mode and I didn't see any drops either.

 

[Charlietus]

Oh so no dips below 4K checkerboard. So what’s PS5 actually rendering at?

Isn't 4k check boarding 1440p native to 4k? Also it's not clear how heavy check boarding even is, so comparisons with PC aren't that straightforward.

 

[Globalisateur]

Then why does the 60fps performance mode at 1800p CB exist at all?

From memory the PS4 pro version was using 1800 CB and it was increased to 4K CB for the PS5 port. It could explain why this setting still exist. The same thing happened for the Shadow of the Tomb raider (except they removed the old settings on PS5).

Isn't 4k check boarding 1440p native to 4k? Also it's not clear how heavy check boarding even is, so comparisons with PC aren't that straightforward.

It's 1536p ish. But like you said checkerboard rendering has a cost (last we know in Dark Souls remastered was 30% of the internal native rendering) so it should be higher than that.

 

[techuse]

Isn't 4k check boarding 1440p native to 4k? Also it's not clear how heavy check boarding even is, so comparisons with PC aren't that straightforward.

It's higher than 1440p. It's half the pixels of native 4k.

CB 4k - 4147200 pixels
1440p - 3686400 pixels

 

[see colon]

Isn't 4k check boarding 1440p native to 4k? Also it's not clear how heavy check boarding even is, so comparisons with PC aren't that straightforward.

No, it's more like 2160i, but instead of skipping even or odd lines on the vertical axis, it's skipping every other pixel on every line.

 

[raytracingfan]

I think exotic hardware gets used less, and even moreso now everyone is multiplatform. PS5 games and PS6 will be designed with PC in mind, probably even concurrently.

I think the lesson of tensor cores and DLSS is that developers are happy to make use of exotic hardware as long as the extra effort they need to do so is minimal. For this console generation we've already seen TLOU Part 1 have a rocky launch because the way it used the PS5's hardware decompression wasn't ported over well to PC. We've already seen from other PS5 ports that it is possible to do a better job, but it does require more effort in porting.

 

[Shifty Geezer]

I consider Tensor as standard, not exotic. Can't be exotic if everyone else has it. Exotic will be stuff unique to your hardware that doesn't port, which in the past enabled great things but now just gets in the way and is ignored.

 

[Flappy Pannus]

Isn't 4k check boarding 1440p native to 4k? Also it's not clear how heavy check boarding even is, so comparisons with PC aren't that straightforward.

Checkerboarding halves the horizontal pixel count. So a 4K checkerboarded image is 1920x2160 native, so 4,147,200 pixels. 2560x1440 is 3,686,400 pixels, so a little less.

Like all reconstruction methods though, checkerboarding has a cost, and it's relatively high from what I understand compared to something like DLSS. So the actual native rendering performance is probably closer to a native res like 1800p.

Gonna be interesting to see DF's video, from their coverage of the PS5 version there were basically no rendering improvements aside from the slightly higher CBR and realtime cutscenes. Some decent opportunity for LOD improvement at least. GOT doesn't have exactly have any egregious pop-in, but room for improvement.

 

[Shifty Geezer]

Like all reconstruction methods though, checkerboarding has a cost, and it's relatively high from what I understand compared to something like DLSS.

Yes in terms of processing, but no in terms of hardware. Hence it's possible on a PS4 where DLSS isn't.