I would love path traced unreal.
BTW - every single aspect of the rendering can be tweaked in VKQuakeRT, check out my twitter thread on it:
Digital Foundry Article Technical Discussion [2022]
- Thread starter BRiT
- Start date
- Status
BTW - every single aspect of the rendering can be tweaked in VKQuakeRT, check out my twitter thread on it:
Wonder why that has not happened yet, Unreal was and is the more graphical impressive game at the time?
Quake's source code is publicly available, so it's easier for large scale reworks like this. Although, there are homebrew rendering options for UE1, including modern DX (DX10 I think) and openGL ports. So maybe there is hope.
vjPiedPiper
Newcomer
I originally posted about this in the next-gen console thread ( PS6/ next MS box),
but since I consider DF to currently be the in public experts on spatial and temporal
up-scaling techniques I would love to hear what they think about my idea...
Basically any chance game engines might start rendering, partial frames, that only include accurate
information for the up-scaling part of the rendering process. I understand existing game engines already contribute
a lot of accurate info already, but as I understand it, no render engine provides anything in the way of temporal up-scaling data?
My comment...
I wouldn't be surprised if we start seeing game engines outputting "mid-frames",
which are just motion data, and other low level info - perhaps some raw texture data,
and then a temporal up-scaling system to use this as input to generate intermediate frames.
Basically, instead of having the GPU calculate all you motion vectors, optical flow, and Depth data from 2 different frames,
the GPU, creates that data for you, giving a much better quality temporal upscale as the result.
Eg. GPU calculates complete render and rasterizes/RT's for the entire frame to completion. = Frame 0
GPU calculates just the updated geometry, and possibly some basic lighting information, then provides a temporal up-scaling
system the Motion vectors, optical flow, updated geometry and light(ing) locations. = Frame 1.
This may in fact place MORE burden on the CPU, but in could over time, develop a much better system for the resulting image stream.
Essentially applying some of the original Render data to the temporal up-scaling system,
or applying AI earlier in the render pipeline.
In Engine super TAA I guess?
Thoughts?
Looking through the current DLSS SDK on guithub,
aint that helpful as it's still 2.4, not sure how much if any stuff wold be helpful for DLSS3 temporal stuff wold be exposed..?
Might be a good side project for anyone thats interested...
but since I consider DF to currently be the in public experts on spatial and temporal
up-scaling techniques I would love to hear what they think about my idea...
Basically any chance game engines might start rendering, partial frames, that only include accurate
information for the up-scaling part of the rendering process. I understand existing game engines already contribute
a lot of accurate info already, but as I understand it, no render engine provides anything in the way of temporal up-scaling data?
My comment...
I wouldn't be surprised if we start seeing game engines outputting "mid-frames",
which are just motion data, and other low level info - perhaps some raw texture data,
and then a temporal up-scaling system to use this as input to generate intermediate frames.
Basically, instead of having the GPU calculate all you motion vectors, optical flow, and Depth data from 2 different frames,
the GPU, creates that data for you, giving a much better quality temporal upscale as the result.
Eg. GPU calculates complete render and rasterizes/RT's for the entire frame to completion. = Frame 0
GPU calculates just the updated geometry, and possibly some basic lighting information, then provides a temporal up-scaling
system the Motion vectors, optical flow, updated geometry and light(ing) locations. = Frame 1.
This may in fact place MORE burden on the CPU, but in could over time, develop a much better system for the resulting image stream.
Essentially applying some of the original Render data to the temporal up-scaling system,
or applying AI earlier in the render pipeline.
In Engine super TAA I guess?
Thoughts?
Looking through the current DLSS SDK on guithub,
aint that helpful as it's still 2.4, not sure how much if any stuff wold be helpful for DLSS3 temporal stuff wold be exposed..?
Might be a good side project for anyone thats interested...
Inuhanyou
Veteran
Anyone who thinks these consoles are not strong enough to last 8 years is too pessimistic. This is far from a PS3/360 generation scenario where devs were actively taking entire modes out of games because they could not be downscaled properly.
Ps4 still holds up well to this day despite being 9 years old. And ps5 was much stronger compared to ps4 was
Hearing DF/Richard talk about the 2028 consoles doesnt sound too great.
Ray tracing is the main limiting factor i think for current gen, and could warrant a mid-gen refresh. RT is generally already too much for the consoles and thats cross-gen, imagine by 2027.... ML is also abit on the backseat hardware-wise, which is another technology where things can be improved alot. Regarding raw raster performance, this is easier to scale and come by with, however its 3060/RX6600XT level performance that has to last all the way to 2028. Dunno man, we havent even seen any game making use of current gen tech due to cross-gen.
Then as Rich mentions its looking very hard for Sony/MS to double the performance for eventual next gen machines at console prices. Things are already too expensive atm. Dire times (ahead).
Ray tracing is the main limiting factor i think for current gen, and could warrant a mid-gen refresh. RT is generally already too much for the consoles and thats cross-gen, imagine by 2027.... ML is also abit on the backseat hardware-wise, which is another technology where things can be improved alot. Regarding raw raster performance, this is easier to scale and come by with, however its 3060/RX6600XT level performance that has to last all the way to 2028. Dunno man, we havent even seen any game making use of current gen tech due to cross-gen.
Then as Rich mentions its looking very hard for Sony/MS to double the performance for eventual next gen machines at console prices. Things are already too expensive atm. Dire times (ahead).
Inuhanyou
Veteran
Rich is not talking about doubling current gen hw for next gen but the cost of a mid gen refresh doubling being pointless and hard to sell to the consumer especially with current economies of scale (he's using PS4 pro as an example which was double PS4). He also mentions good point with Xbox series S and X where it's just kind of silly to have 3 consoles at once of the same family to optimize for when MS and Sony are just fine now releasing their games on PC to take advantage of the market that wants higher graphics than what console offers.
Which is fair. Devs have had to restrict their games to ps4 limits for 9 years and games are still relatively fine. Devs will learn how to best make use of the tech this time around for both 60fps gaming and whatever they want to add for 30fps. Consoles will always be the economic and convenient option and not meant to "keep up" all the time with power and cutting edge features.
There are far less barriers that will scale against the consoles. You mentioned RT but it's something that PC being scalable can have as an option the consoles don't need to have or can substitute with a much less taxing approximation that looks worse but still is ok.
Taking out some parts of RT consoles can't handle is inevitable but it's not a must have feature to sell an extra new console on when most people can't even find the original in stock at a decent price and most devs have not even started taking advantage of the hw available to them already.
Inuhanyou
Veteran
But then you have to think about what can they justify that hardware with and what exactly are they going to put in it. Growth in hardware is slowing down dramatically and process nodes which would justify the costs of such transitions are not appearing to have the same effect they once did. It's the same reason MS released series S.
I could see a chiplet slimline PS5 but largely speaking we are hitting the limits of whats possible and cost effective at such a short amount of time.
Sony for pro said the main justification was that they wanted 4k gaming and to stop people from going to pc. At this point resolution is dead and they seem to be leaning into pc market more than ever as long as it nets them software sales.
The thing is.... what does faster mid-gen consoles actually get you nowadays? Let's face the reality... the things holding back game visuals and presentation are budgets and time.
Bumping resolution doesn't make a game next gen. Improving performance is always nice, but there's enough power in these consoles to last 5+ more years easily. The types of games that are really pushing the state of the art take 5 or 6 years to develop. We're finally just now, 2+ years after the release of these consoles... getting around to actual next gen exclusive games. Sony Santa Monica for example is working on another IP alongside God of War (Cory Barlogs new project) and is likely 3+ years away still.
So if you think of it in terms of time and projects... Sony's teams will get around 1 or 2 more titles each out on the platform on average. That gives them a "current gen" game, plus a game that improves on that and refines the tech. Then they start developing for the NEXT console.
Bumping resolution doesn't make a game next gen. Improving performance is always nice, but there's enough power in these consoles to last 5+ more years easily. The types of games that are really pushing the state of the art take 5 or 6 years to develop. We're finally just now, 2+ years after the release of these consoles... getting around to actual next gen exclusive games. Sony Santa Monica for example is working on another IP alongside God of War (Cory Barlogs new project) and is likely 3+ years away still.
So if you think of it in terms of time and projects... Sony's teams will get around 1 or 2 more titles each out on the platform on average. That gives them a "current gen" game, plus a game that improves on that and refines the tech. Then they start developing for the NEXT console.
Inuhanyou
Veteran
Right! You get exactly what I'm saying! It doesn't make much logical sense for a upgraded console of either machine in this type of environment!
Well, I felt the same with PS4 pro and one x but that still happened. So I'm not claiming to be Nostradamus...but it makes even less sense now than it did then. We have good GPUs and CPUs, good amount of ram and bandwidth. RT performance and quality is not as impressive as on PC but still doable if a dev wants to do it.
And on top of that, dynamic res and FSR2 available on consoles is now really allowing flexibility for the GPU that wasn't around in previous gens. If games stay around minimum 1080p as a baseline, thats fine as long as upscaling techniques improve and dynamic res is aggressively applied.
I was shocked when Alex said, contrary to the "current gen is already too weak" crowd, that plague tale could have run at 60fps if taken down to 1080p instead of 1440p. Add VRR on top of that and it's perfectly great.
The most important thing you said is the most true. These consoles are powerful enough where their hardware is not holding back developers from making 99 percent of games they want to make. It's budget, time and resources. There is no Jaguar CPU in there forcing devs down to the wire, and these consoles are the exact opposite of the Nintendo Switch.
These consoles at this point could run star citizen if they wanted. It started development for hw far weaker than consoles are now despite being designed for cutting edge at that time, and it's still the most ambitious and high scale project I've seen. Whether the devs are actually planning to complete it or not
Last edited:
Again they don't understand the economy Sony is not satisfied if someone goes from PS ecosystem to PC because of PS5 underpowered. Sony biggest revenue and profit comes from DLC, microtransaction and third party games sales royalties from the PSN.
First party sales on PC is a bonus. When you see the Sony result this is probably around 1% of Sony revenue and profit. They earn less money from first party games on PC than sales of first party games on PS5/PS4 per unit* and first party games earn less money for Sony than third party royalties.
If they need one PS5 pro, they will do it.
You need to understand than supply constraint will disappear and probably before 2024. Like I said they will have shipped 42.3 million PS5 31rst march 2023 and between 65.4 million(official minimum) to 72.3 million(rumor) the 31rst march 2024. If with such number the PS5 is supply constraint, there is a hope for Sony it will reach PS2 number and I don't believe it.
*Average price sales is lower on PC than PS5/PS4 for first party games and they need to pay royalties to Steam or Epic Games store. And they sold more first party games on PS5&PS4 than on PC.
Last edited:
if they don't release new consoles before 2027-28, be it mid gen refreshes or new gen, we may end up with mobile chips offering better RT performance than the consoles
This is the reason to offer mid gen, RT performance.
Inuhanyou
Veteran
As long as battery life is a pressing concern that's probably super hyperbole. Architecturally, the switch 2 will have greater RT performance relative to series S, but as it will be far weaker than the consoles to begin with it's an unbalanced comparison. switch 2s heavy lifting will be in utilizing dlss. Something devs already have an equivalent to in FSR for other consoles, to not even mention the other upscaling techniques devs have been using.if they don't release new consoles before 2027-28, be it mid gen refreshes or new gen, we may end up with mobile chips offering better RT performance than the consoles
Last edited:
Inuhanyou
Veteran
If Sony and MS didn't see PC as a real driver of growth they wouldent invest in the marketplace to begin with.
Again, PS5 is not underpowered and won't be considered underpowered for a long time. Certainly not by 2024, only a year and 1 month away. And if they don't launch in 2024, there's largely no point in doing a refresh as consumers will not take kindly to a refresh only 2 or so years prior to the actual next generation.
Does AMD have some special new architecture we haven't heard about planned in a year and a few months that will drastically increase RT to make a new console worth it? Will process nodes shrink enough or have enough wafers to go around with all the companies trying to get them to justify a new system investment? Are consumers who are still getting used to paying 500 bucks or more for consoles(more in Europe where their stock is also still dire) willing to pay out 100 or even 200 dollars more for a new machine like that when their current one is doing relatively fine to justify Sony making it? Let's just say I'm skeptical
Last edited:
In RDNA3, there is traversal acceleration and ray bounding box sorting. This is probably enough to push midgen console in RT.
EDIT: And TSMC will be more than happy to have a new customer for 5nm process node, currently orders are decreasing. This is not a problem anymore and one of the reason Sony was able to go from 18 million PS5 this current fiscal year to 23 million.
Last edited:
Both of these are not confirmed at all yet - Ray traversal May be in reference to AMD specific Ray traversal programmability that goes above DXR. Aka what one can so in the consoles.
At least we know they improve ray traversal to be able to traverse less node.
And they have ray sorting programmability for different use case here shadow ray and reflection
And the most important it helps with heavy RT workload where current gen console and AMD RDNA 2 GPU suffer a lot and up to 1.8x is not bad. I don't expect mid gen console to be at Nvidia level but to be much better than PS5 and Xbox Series.
Last edited:
- Status
Similar threads
