So we are getting
*Improved textures
*Some ray traced reflections at a low-medium quality
FauxK @30fps
Not bad, but not amazing either.
I think its about on par with the gears5 visual update.
Yeah I can see your point there. I imagine the LOD is probably shortened in the frame rate mode, so he might not have thought about it lol.
The ray tracing is actually very high quality, especially for a console title and I think using the term FauxK is underselling it quite a bit.
Considering the scope and effects in use, I’d say it’s more impressive than the Gears 5 update. That update is basically porting the PC ultra settings with some slight improvements like better GI and particle effects.
1 Like
I often get burned out on open world games after ‘completing’ them. I’m the type to play em super thoroughly but then never wanna touch em again. I’d say soft spot (his words to me) suggests (in my view) that one might be willing to look past some flaws or highlight some good stuff in ways that could be seen as justifying the excitement. This is something literally everyone interested in gaming does though. If they were just bland robots, it’d be painful to watch stuff like this. I’m not supportive of ppl nitpicking John’s comments.
Is the RT very high quality? I dunno. I’d disagree with that tbh. Lower res and missing TONS of objects in the reflection is not what I’d consider very high quality.
I should have been more clear, when I was referring to low-medium quality Raytracing I was referring to low-medium quality raytraced PC reflections, some of the RT reflections in mm are low res, and fauxk just means dynamic 4k or sub 4k, mm is not native 4k. Mm on PS5 is about what I would expect from a RTX 2060, which is what I expected.
I mean for next gen consoles. In that context, considering it’s an open world game where RT is much more taxing than a linear game, I think it’s impressive and a good quality RT. Alex at DF did an entire video covering how well optimized it was.
The optimization doesn’t make it high quality. Faithful, accurate reflections are (for reflection RT I mean).
1 Like
To each their own. It may be early to really determine what is and not good quality RT considering this is only the first taste of RT on consoles. I do wonder how the RT implementation in Watch Dogs will compare since that’ll be the closest in terms of world design.
1 Like
I am quite bullish on RT quality for this coming gen. More so than most are. 
I’ve caught that. A bit. Aside from white papers or tech sheet comparisons (that would be too in depth for laymen to extrapolate the RT capabilities) can you summarize your basis for the bullishness (and why you lean towards one hardware set capabilities over the other)?
Feel free to maybe summarize here and DM me if too far off topic?
1 Like
Yeah I’ve gotten that impression lol. I do hope you’re right but what I’ve read and seen leads me to be a bit more reserved with my expectations. Hopefully that leads to me being wrong and the RT results end up better than what I’m expecting. That’s a win, win situation in my opinion.
2 Likes
Specifically on Xbox, which has machine learning inference capabilities, you can do some rather impressive RT denoising. That means less rays needed per pixel and a lot better RT performance. It’s almost like DLSS but for RT. There is also the option to use ML to help handle texture work, which frees up the CU for more RT operations (since RT is handled in texture unit of the CU iirc).
More generally, cross gen games won’t show off meaningful RT much. Their lighting engines are built to mimic RT as is (and have been for decades) so they can’t throw that out until they have a good fresh, ground up next gen only game engine. This will be one of the major growth areas this gen wrt improving the engines, pipelines and art design.
1 Like
I’ve read where Series X has the capability for ML and INT8, INT4 but I’m unclear how that’s processed. Do the RT cores handle the ML? If so, how would that help RT by taking away the HW acceleration? Or, is the denoising you mentioned responsible for this visual improvement?
Has there been fairly reliable evidence that Sony hasn’t implemented any form of machine learning or are we judging more on the absence of info, particularly in light of MS being so vocal.
I never did a deep dive learn for regular shading and rasterization, and the RT on top of (or tied into) these methods is still zooms fairly high over my head. 
It’s generally being inferred due to the fact we don’t have full details from Sony on their SoC, historically this is no different than any console before Sony have always delivered their info via Mark Cerny at Dev. conferences and via interviews. A quick google for the DF takes show that we got the most detail for PS4 Pro but still not a full breakdown of the chip. Links for those interested PS4
PS4 Pro
Interesting how the guarded approach to information reveal and conservative design have carried over for PS5, although in the PS4 Pro interview the number of “early” GCN features in the chip is specifically called out and detailed which puts the assumption that PS has no ML features (as they’ve not been mentioned) on a stronger footing. Time will tell I guess as Devs speak up through the generation.
Since this is a bit off topic (but still somewhat relevant to RT on the PS5) I’ll put this under a summary tag.
Summary
There are no RT cores like on the Nvidia cards. The CUs are capable of RT and ML (on top of all of the functions they normally do). Because of this I’m far less bullish on RT implementation on consoles. I don’t think the raw RT ability on console will compare to even last year’s PC hardware.
That said, as TavishHill pointed out, there are opportunities in leveraging ML to make RT more and more effecient.
Much like ML 2 years ago, RT is relatively unexplored and new.
Nvidia just released a paper and where they’ve used a denoising algorithm to create situations where you can have a million real-time dynamic direct lights (from emissive geometry to boot!) and shadows on existing hardware. It is imperfect but it demonstrates the untapped potential of this technology.
Plus, we are seeing innovations in software-based RT. Unreal Engine’s Lumen is done in software, and we’re seeing a few games dabbling in it.
I personally don’t think Minecraft DXR on the Series X compares very well to Minecraft RTX on Nvidia hardware. For now. But when you mix these efficiencies and the usage of software-based RT, we may very well see something better on the Series X in 2 years than what we do today with Minecraft RTX.
I think Spider-Man looks great and I was really impressed with that mall sequence. I wonder if there will be a case of diminishing returns for the general public, but for me that was pure eye candy. RT helps objects sit in the scene naturally and for the whole scene to look natural. And in my eyes it performed beautifully in this demo.
To me this looks like a second-year title where the developers are accustomed to the hardware moreso than a launch title. Part of it has to do with the fact that Insomniac was able to do a lot with the aging PS4 hardware, making one of the best looking games of the past generation, but they also seem to know what they’re doing in regards to implementing new features like RT.
One thing that I’m looking forward to that has not been touched on, for technical reasons, is how Insomniac will implement 3D audio via the Tempest engine. We’ll only really know once we play it on our PS5s with a set of headphones calibrated to our heads. There is so much opportunity to take audio to the next level when you have a level situated in a city. Ray-traced audio along with Tempest, could mean another level of immersion when you are swinging around in the open air. Insomniac went hard when it came to leveraging the SSD in Ratchet, that I have a feeling that they’re going to do it with the audio in this game.
2 Likes
I also quite liked the part where they show Spidey from up close, the attention to detail on the hoody, the backpack etc, that was indeed very nicely detailed. I hope we see this from many studios going forward. Awesomely detailed characters. Know where some games the character model is very detailed but some of the clothing almost seems textures or pasted on? That needs to be gone.
Gears 5 armors were very nicely detailed as well. I am sure we’re gonna see great stuff from both sides.
2 Likes
ML is veering slightly off topic for this thread, but its is handled by the CU’s. It helps by letting RT run lower quality (fewer rays per pixel) which results in either low res reflections of sparse pixel coverage (i.e. noise). So you end up with noisy parts (i.e. gaps in the reflection imagery), but filling those in is akin to super sampling techniques where you can use ML to add in the missing pixels. And yes, Sony has confirmed no ML in PS5.
More on topic, an old school mate of mine is a senior lighting artist on Spiderman. He was actually super skeptical that next gen would have any RT at all just prior to Insomniac being bought out, which is kinda interesting.
2 Likes
Looks pretty great. Insomniac are some of the best developers out there, so no surprise. What is mildly surprising is that the RT looks better than almost any game I’ve seen to date.