Sony PS6, Microsoft neXt Series - 10th gen console speculation [2020]
- Thread starter Megadrive1988
- Start date
Swinging out on a tangent, what's the real limiting factor for future games? Is it actually rendering power, or is it money to make games that can fill more power? Let's imagine a perfectly virtualised streaming tech, fetching JIT resources from SSD. Doesn't matter how much RAM is needed to buffer that, let's say the consoles have it. The limiting factor will be games in the terabyte domain. We can't get assets sufficient to saturate what UE5 and Nanite can render now.
Is the future something we talked about for PS360, runtime procedural generation? ML texture, model, and animation generation? If we don't establish those protocols now, could a next-gen hardware focus on that hardware without it being overlooked?
I think we at a an asset quality limit already. I think the future needs better lighting, which we're getting towards, and then systems which I'm guessing will be ML based. Prediction? Similar hardware to what we have, scaled up a bit in the CPU, stronger RTRT, and then the additional silicon budget all on ML processing! ML is the only avenue left to go and it's virtually untapped so far.
Is the future something we talked about for PS360, runtime procedural generation? ML texture, model, and animation generation? If we don't establish those protocols now, could a next-gen hardware focus on that hardware without it being overlooked?
I think we at a an asset quality limit already. I think the future needs better lighting, which we're getting towards, and then systems which I'm guessing will be ML based. Prediction? Similar hardware to what we have, scaled up a bit in the CPU, stronger RTRT, and then the additional silicon budget all on ML processing! ML is the only avenue left to go and it's virtually untapped so far.
Well you knever know. We still have a long way till next gen. Maybe until 2027 the earliest I believe
MS seem to have been more forward thinking in this area with the inclusion of DP8a/4a hardware.
I think it's gonna be necessary to push this generation out for a long time. The X360/PS3 gen extended standard generation lifetime from 4-5 years to 7-8 years, and I think this might be the time to go further and start thinking 9-10 years.
The sort of leap we got this gen in terms of CPU and I/O will never be repeatable, so I think we've gotta push things out as much as we can to get the best viable increases possible. And remember with costs of manufacturing on new process nodes becoming ever more expensive, it may be necessary to stay a little bit farther behind here when it comes to 'affordable' hardware like consoles.
Plus we're already like two years into the lives of XSX/PS5 and it still doesn't feel like we've actually started at all with 'next gen' gaming. I also suspect that with the lengthening of development times, developers and publishers would quite appreciate being able to produce more than one or two big AAA games in a generation before having to think about retooling for 'next gen' again.
I think we will have a Zen 16 cores CPU, Mark Cerny told in a Wired video some devs wanted 16 cores maybe GG or Id Software and GG physics engine and Id software game engine scale until 16 cores. The GPU will at least be 40 Tflops and a much better raytracing and tracing performance in general, some progress for ML at least on PS side. And I continue to think we will stay with 16 GB of RAM maybe some GDDR7 but with a faster SSD PCIE 5 for example. In UE 5 we have virtualization of geometry and texture but developer can go further, why have animation of NPC not visible on screen or sound the player can't hear like for example a dialog between two NPCs on an open world city? We can imagine keep in memory animation of the player or sound generated by the player and ambient sound because it will be there all the time and stream everything else.
In the CPU part some devs asked 16 cores for PS5
www.tweaktown.com
In the CPU part some devs asked 16 cores for PS5
Samsung announces next-gen GDDR7 memory, with insane 36Gbps bandwidth
Samsung teases that it is working on 36Gbps GDDR7 memory for next-gen GPUs in the data center, HPC, mobile, gaming, and automotive markets in the future.
www.tweaktown.com
Yeah we can. People do this every generation. The limiting factor is gpu power, cpu power, and ram just like every generation. Every engine and every game has an approach they'd rather take if they had more powerful targets.
CPU and GPU power will limit the framerate, resolution, lighting quality and quality of antialiasing but now assets quality is limited by game size. We can have into a current game game 8k texture like in movie with one texel per pixel or scenery with 1 polygon per pixel and later all geometry. Brian Karis did a tweet saying he thinks Nanite will ne available for skinned geometry next generation of consoles but maybe foliage and improved nanite terrain will be there this generation.
This generation we have strand hair with Analytical AA by compute. On asset quality side we will probably reach the ceiling next generation consoles.
Well their point is that growing asset quality and quantity to actually saturate something like UE5 or some other next gen engine is going to start requiring incredibly large file sizes. Of course there will still be processing limits at some point, but how do we manage having 'cinema quality' assets in high quantities all over the screen and at greater draw distances, while keeping file sizes manageable? Especially with the switch to SSD's which will mean you get much less storage for the same price?
So this can become just as big a problem, but one potentially much harder to solve.
I mean
PS1 - 1994
PS2 - 2000
That is 6 years
PS2- 2000
PS3 - 2006
That is 6 years
Ps3 - 2006
Ps4- 2013
That is 7 years
Ps4 2013
PS5 - 2020
That is 7 years
The age of consoles only really increased by a year for sony which is closer to the norm anyway
The Nes was 1983 - Super ness 1990 - 7 years , N64 was 96 so 6 years , Gamecube was 01 so 5 years , wii 06 so 5 years again , wii u 2012 so 6. years switch 2017 so 5 years.
It really seems to be the market leader can extend out the generation with a successful device while the ones who are doing poorly release systems faster. We can get into the nitty gritty about which month of the year its .
As for leaps of technology I dunno. We could see xillian tech in the newer amd processors allowing for vastly better performance esp ai. Huge amounts of 3d cache are really speeding up the AMD chips with that x3d cache. Amd is also rumored to be doing BIG/Little cores at 3nm with zen 2 as the little and newer zen 5/6 cores as the bigger. Having 8 large zen cores with 8 small could also boot performance on the cpu side. From the DF videos I'm seeing if I understand correctly the zen 2 cpus may be just as much of a limitation on ray tracing as the gpu is itself. But heck having a few little cores that just handle the OS could provide benefits above what we have now
AMD had pretty poor ray tracing peformance with rdna 2. With rdna 3 we can see large increases (will find out tomorrow) but before another generation we might see RDNA 4 and even RDNA 5 or whatever they may be named. Infinity cache could also be a way to vastly increase performance. MS did spend a lot of their hot chips presentation with the series consoles talking about how poorly ram bandwidth has scaled. IF we are in for another 4k generation they might be able to put enough IC on a chiplet to have really fast performance.
Is it not an argument that because good ray-tracing performance will save developers a lot of lightning work, and what else good ML/A.I performance can save developers time and money, a new generation consoles will release sooner rather than later?
That'd depend entirely on how well RT and ML could scale in hardware. Let's say the target is 2x the best of the best possible now, 4090, to makea clear visual difference to this gen, how long will it take to get that into a console-sized bit of silicon? There's the possibility of novel hardware improving these over just scaling up from what we have now, but I'm not seeing anything that suggests these will scale better than normal rasterisation has done. The argument then shifts to how much ML + RT is minimum to make a generational improvement over what we have now.
That could be an issue in the future but we still don't have a real 4k generation of hardware. The ps5 and xbox series are not really able to hit 4k 30fps as it stands now without upscaling or dynamic resolution. So I think the major thing for consoles is to actually get hardware out there that can do 4k 60fps without that.
For machine learning texture /model and animation generation , would that not be better off happening off line before a game is shipped ? That way you can have an artist or animator do a pass on everything to make sure its okay ?
We don't know what ML will be able to offer. 
Assuming it can do magical things, it could provide an alternative avenue for More Power by working different. See for example the ML physics simulations requiring a fraction of the power of traditional solvers.
As for a true 4k machine, I don't think that'd constitute a next-gen, looking not much different to this gen only in slightly improved clarity. This gen gets away with it somewhat by the experience, largely thanks to the SSD and OSes. If we tweak the definition of 'next gen' to be 'a new console that everyone would want to spend money to upgrade to', in the past that's been through more power and better (looking) games, or a fancy new controller. This gen is as much user experience as better looking games. What will the must-have USP of next-gen be? I'd need to see a substantial improvement on PS5 to upgrade just on visuals. Just using UE5 demos so far, that'd come not from lighting or geometry as UE5 is amazing, but the people an animation in that is lacking. That's one area, creating believable 'people' in these world.
Assuming it can do magical things, it could provide an alternative avenue for More Power by working different. See for example the ML physics simulations requiring a fraction of the power of traditional solvers. As for a true 4k machine, I don't think that'd constitute a next-gen, looking not much different to this gen only in slightly improved clarity. This gen gets away with it somewhat by the experience, largely thanks to the SSD and OSes. If we tweak the definition of 'next gen' to be 'a new console that everyone would want to spend money to upgrade to', in the past that's been through more power and better (looking) games, or a fancy new controller. This gen is as much user experience as better looking games. What will the must-have USP of next-gen be? I'd need to see a substantial improvement on PS5 to upgrade just on visuals. Just using UE5 demos so far, that'd come not from lighting or geometry as UE5 is amazing, but the people an animation in that is lacking. That's one area, creating believable 'people' in these world.
Everyone's wants are different as you can easily see from the pc market. Some want pure resolution , some want super high frame rate and then some want as many advance effects like raytracing as possible.We don't know what ML will be able to offer.
As for a true 4k machine, I don't think that'd constitute a next-gen, looking not much different to this gen only in slightly improved clarity. This gen gets away with it somewhat by the experience, largely thanks to the SSD and OSes. If we tweak the definition of 'next gen' to be 'a new console that everyone would want to spend money to upgrade to', in the past that's been through more power and better (looking) games, or a fancy new controller. This gen is as much user experience as better looking games. What will the must-have USP of next-gen be? I'd need to see a substantial improvement on PS5 to upgrade just on visuals. Just using UE5 demos so far, that'd come not from lighting or geometry as UE5 is amazing, but the people an animation in that is lacking. That's one area, creating believable 'people' in these world.
I think for a console its hard to be capable of everything for everyone. My assumption will be that we get more of the same from this generation but better. You will likely see Resolution Mode which gives you full 4k and then a frame rate mode that picks a lower resolution to offer 120fps modes. Then you'd see a mode to enable ray tracing. Of course it could simply be that next generation there is no traditional vs raytracing option and all games are raytraced. if we assume a 2026/7 release for next gen consoles you might see RDNA 5 or 6 or a fully new design from AMD. RDNA 2 was 2020 , 3 is 2022 , 4 2024? , 5 2026? , 6 2028 ?. Based off last gen the ps4 pro came out 3 years after (2013 vs 2016) and the ps5 came out in 2020. So you have 3 and 4 year gaps in releases. For sony a ps5 pro would possibly come out next year. That might firmly be an RDNA 3 console with Zen 4 as an option. All in all not a bad mid gen console refresh. 2027 would following last gen be a ps5 which could put it using RDNA 5. Looking at MS the xbox one x came out 4 years after the xbox one (2013 vs 2017). So it could either be a RDNA 3 or RDNA 4 refresh with zen 4 or 5 by that point. It would be interesting for MS if they do go with a new RDNA design vs more of zen2/rdna2. If for example Zen 5/RDNA4 is ready in time they could have a xbox series s like console that is more powerful than the series x was.
In regards to a refresh I don't think they would be powerful enough to offer a lot of machine learning tools as we already know what RDNA 3 consists of and its not a large enough leap in capabilities. RDNA 4 may bring more stuff but I am doubtful that either Sony or MS would use them for a mid cycle refresh. It could be possible however depending on how they want to extend out the generations.
I still think what you are describing will be done off line and not on the consoles themselves. Something like azure or geforce cloud would be much better at it than a single console. Imagine as a developer you could simply buy time on azure or geforce and have a bunch of 4090s or insert top of the line hardware for that year running through your game making new textures and animations. It would be way faster than what can be put in a box and it wouldn't have to be done in real time. The only downside is storage space but I think that is an easily solved for case in the future by bringing back carts
The next gen consoles would imo be the ones to do what you are talking about. But who knows what the new constraints on micron shrinks will do to the industry.
I am afraid that next gen gaming will not include consoles going forward. Today's next gen PC graphics are tomorrow's console standard visuals. The push of first party console games onto the PC which makes PC gaming more attractive, the growth of the PC market in general and the ever-growing significant performance gap, make PC the perfect testbed for modern cutting-edge graphics.
These 70-90 Tflops gpus have alot of head room especially when the push for higher resolutions has slowed and 150+ fps gaming isn't seeing a major push from users. It gives devs a ton of performance to play around with. Allowing brute force features with years of lead time before needing to spit out more elegant efficient solutions for next generation tomorrow consoles that offer similar performance of PC gpus of today.
LOL. Im guessing it will look akin to the PC scene of the late 90s and early 2000s when PC exclusive devs were plenty and were concentrated on the performance of your average PC gpu and not concerned with cross gen support with console gaming.
Silent_Buddha
Legend
The rapidly increasing cost associated with PC gaming is going to significantly hinder PC gaming uptake going into the future.
I can however, see PC gaming again becoming the target platform for graphics intensive game development as it was in the late 90's and early 2000's.
It'll be nice to have a return to cross-gen game development being the norm as it was in the 90's and early 2000's versus this odd reaction (to me) of console gamers somehow thinking that cross-gen is a bad thing. If anything it forces developers to have far more flexible coding skills in that they will be targetting the most cutting edge hardware and then scaling it down to work well on much less powerful hardware.
It's been a sad state of affairs, IMO, with console first game development where instead of targeting the best hardware and scaling it down developers have been targeting mid-range (start of a console gen) or low-end (end of a console gen) and then attempting to scale up. The complete opposite of what was prevalent on PC two decades ago.
If the trend had continued to be target the best and scale down to the worst then all of those "bullshot" graphics that we saw during the launch of the PS4/XBO would have been delivered to the gaming public. Even if they might not have run on at max graphical settings on the majority of PC hardware at release, those settings could be reduced on mainstream hardware and in one or two generations of hardware and they'd run just fine ... just like back in the 90's and early 2000s. Ah well, c'est la vie.
I do hope that it goes back to that rather than developers being bullied by complaints from the console market that they should only target mediocre hardware (current gen consoles) rather than having a scaling centric game development pipeline that instead targets the best hardware and scales down. Maybe then I'd get super excited by game graphics again?
Regards,
SB
It is not financially feasible to build cutting edge games just for PC anymore.
I know people point to something like Star Citizen, but it's very much an exceptional case and should not be used as an argument to what other developers/publishers can expect.
This also isn't like the old days where some smaller group of very clever developers could build some super advanced-for-the-time rendering features in their bespoke engine and blow people away. There's no low hanging fruit to tackle anymore and the limiting factor for making an amazing looking game isn't the tech as much as it is manpower to properly utilize the tech. It's the main reason why AAA development has ballooned in costs and why AA has been slowly dying.
I know people point to something like Star Citizen, but it's very much an exceptional case and should not be used as an argument to what other developers/publishers can expect.
This also isn't like the old days where some smaller group of very clever developers could build some super advanced-for-the-time rendering features in their bespoke engine and blow people away. There's no low hanging fruit to tackle anymore and the limiting factor for making an amazing looking game isn't the tech as much as it is manpower to properly utilize the tech. It's the main reason why AAA development has ballooned in costs and why AA has been slowly dying.
