Sony PS6, Microsoft neXt Series - 10th gen console speculation [2020]

[QPlayer]

Regarding nextgen consoles, two factors will be decisive. One of them is their pricing, which will make them deviate from the category that is traditionally still affordable for the general public, and even higher. In other words, there will certainly be expensive console variants, necessarily, because technological developments can only be followed with more expensive components today. Of course, in addition to this, there will also be variants with lower performance, but those are not the ones that someone will go for if they want to present the maximum graphics skills of the new consoles. In my view, they will therefore copy the segmented hardware philosophy of the PC market more than ever.

Another significant factor will be the effect of image and framerate tuning techniques on visual performance. Today, everyone still talks about the raster performance when looking at the graphics capabilities of a machine, but this will soon change. The latest VGA generation will illustrate this well. AI-enhanced frames and improved image quality will be more decisive than ever. Once they reach the minimum input lag, from then on nobody will care that that computer is actually only calculating 10-20% of the graphics on the screen and the rest is AI "magic".

 

[raytracingfan]

Path tracing is also not gonna feel like such a crazy upgrade when ray traced GI becomes pretty common this generation already, as it's already starting to be. It'll be a nice, but modest improvement. That alone wont remotely justify calling something 'next generation', unless we're seriously watering down that term to be so much less than it used to mean. We've gotten quite good improvements to lighting quality every generation for a while now, but that's generally been accompanied by sizeable improvements in many other aspects of the visuals as well. Without those other sizeable improvements, you're just putting lipstick on a pig. Again, there's still so much room for improvement in visuals outside things that ray/path tracing can help with. Those things cant just be neglected.

Ray tracing can improve more than just lighting. Once RT is used for primary visibility, depth-of-field and motion blur effects superior to the current post processing effects can be implemented. Occlusion culling is no longer necessary with RT primary visibility. Order-independent transparency can be handled elegantly. And with RTX Mega Geometry (or the inevitable AMD/cross-IHV equivalent), Nanite-equivalent geometric detail can be achieved without the software rasterizer.

 

[Charlietus]

If you were expecting this gen to usher in perfectly realistic graphics, I dont know what to tell you. Of course there will be compromises. Next gen machines will also require games to be full of visual compromises.

Before launch I didn't expect consoles to be this bad at ray tracing. So you have developers that clearly want to use more ray tracing, and the hardware that isn't great. Reminds me of deferred lighting in the 7th gen. And just like that gen, 9th gen is an inbetween transition where the next consoles will be much better for developers.

But what's the price point to reach those ambitions? On the playstation side, I wouldn't be surprised to see 2, maybe three consoles all at launch.
The biggest bottleneck for the PS5 is the price, so Sony is going to come prepared I assume.

 

[Boss]

Ray tracing can improve more than just lighting. Once RT is used for primary visibility, depth-of-field and motion blur effects superior to the current post processing effects can be implemented. Occlusion culling is no longer necessary with RT primary

For all the many cross gen games we had for a long time, sure, but there's a significant leap in overall presentation values with plenty of the actual next gen titles we've started getting. Just because we dont have direct 1:1 screenshots to compare with non-existent last gen versions of the games doesn't mean that big leap isn't there. In fact, if you ARE showing direct comparison shots of games with old gen versions, then it's kind of implicitly NOT next gen, is it?

Path tracing is also not gonna feel like such a crazy upgrade when ray traced GI becomes pretty common this generation already, as it's already starting to be. It'll be a nice, but modest improvement. That alone wont remotely justify calling something 'next generation', unless we're seriously watering down that term to be so much less than it used to mean. We've gotten quite good improvements to lighting quality every generation for a while now, but that's generally been accompanied by sizeable improvements in many other aspects of the visuals as well. Without those other sizeable improvements, you're just putting lipstick on a pig. Again, there's still so much room for improvement in visuals outside things that ray/path tracing can help with. Those things cant just be neglected.

As we get further and further into the gen, the graphics will improve. We'll start to see projects with more recent versions of UE5 which will look better when compared to cross gen games. However, i don't think that graphics or raytracing will be the big sell people on here think it will be. In fact, I think we may even see a longer cross-gen period next gen than we saw this gen. Despite the increase in image quality, power, and frame rate, people still refused to upgrade to the ps5 as quickly as they did from ps3->ps4.

You could argue that covid hampered the progress and that's true. However, from current trends, we can see that people are playing more old games than they are new games. If the sell next gen is, more raytracing paired with static, non interactive set dressed worlds, it's going to be a seriously tough sell. The games themselves have fallen into the hollywood trap of releasing visually appealing but thoroughly uninteresting content. It just doesn't sell. More attention needs to be paid to other aspects of the games like the gameplay, physics, etc. That's where Sony and co should look to innovate/differentiate themselves from the pack.

Finally, consoles need to get back to their roots of enabling new experiences through custom dedicated programmable hardware. That doesn't necessarily mean a departure from x86 but, a new line of thinking. Sony and nintendo tried to accomplish this with their controller features with varying levels of success. They need to put more resources in that style of innovation from peripherals to dedicated blocks on their chips, etc.

 

[rntongo]

Before launch I didn't expect consoles to be this bad at ray tracing. So you have developers that clearly want to use more ray tracing, and the hardware that isn't great. Reminds me of deferred lighting in the 7th gen. And just like that gen, 9th gen is an inbetween transition where the next consoles will be much better for developers.

But what's the price point to reach those ambitions? On the playstation side, I wouldn't be surprised to see 2, maybe three consoles all at launch.
The biggest bottleneck for the PS5 is the price, so Sony is going to come prepared I assume.

When has Sony ever released 2-3 models at the same time and how would Software support work for this? Seeing how obsessed Cerny is with making development easier and cutting down the time to triangle this doesnt add up for me

 

[Seanspeed]

Ray tracing can improve more than just lighting. Once RT is used for primary visibility, depth-of-field and motion blur effects superior to the current post processing effects can be implemented. Occlusion culling is no longer necessary with RT primary visibility. Order-independent transparency can be handled elegantly. And with RTX Mega Geometry (or the inevitable AMD/cross-IHV equivalent), Nanite-equivalent geometric detail can be achieved without the software rasterizer.

I think you're heavily overestimating what RT can realistically do, or in a better manner than other techniques with many of these things, and it's quite impossible to ignore that your reasoning might be incredibly biased based on your screenname....

 

[Seanspeed]

Before launch I didn't expect consoles to be this bad at ray tracing.

Well this will be a bit harsh, but saying that just tells me you're not terribly great at understanding what specific hardware should be capable of, and thus what kind of hardware we might need for the kind of visual leaps you want to see.

It was very apparent when RDNA2 and the new consoles were announced, that the RT support would not be strong. In fact, most people were quite surprised there was any kind of hardware accelerated support for RT at all. Beyond that, even among the generally lower expectations of what RDNA2 hardware could do within these console limitations, I think there's been a number of titles that have exceeded those expectations and have shown that with some effort, getting RTGI to work even at 60fps in true next gen titles is possible. UE5 as well has made progress towards making Lumen work at 60fps on consoles. Clearly, if 30fps were still being prioritized, a fair bit more could be done still. RDNA2 is not exactly good at RT, but we've already seen it can work on these consoles and we'll undoubtedly see it more and more going forward in the next few years.

You seem to have bizarrely had very high expectations of these consoles and I dont quite know what else to say about that.

But what's the price point to reach those ambitions?

Well we're getting heavily into territory the mods dont want us to get into, even though it's invariably all interwoven together. You're very right to ask that question, though. In fact, I really think moreso than trying to talk specs, the question about the financial aspects should actually be the main priority, as everything about the tech and specs follows from that. We're clearly in a new age where we cant just make the same basic assumptions about improvements in processes and performance-per-dollar like we used to.

 

[Seanspeed]

In fact, I think we may even see a longer cross-gen period next gen than we saw this gen.

No argument there at all. With the kind of limited hardware improvements I think are realistically possible with a new console at any kind of reasonable price point in say 2028, I'm not sure there will be many games that will be justifiably exclusively next gen that couldn't be ported down to XSX/PS5. Especially if AI is doing a lot of 'enhancement' work, that leaves plenty of room to have un-enhanced versions.

 

[Charlietus]

Well this will be a bit harsh, but saying that just tells me you're not terribly great at understanding what specific hardware should be capable of, and thus what kind of hardware we might need for the kind of visual leaps you want to see.

You seem to have bizarrely had very high expectations of these consoles and I dont quite know what else to say about that.

My expectations before launch (and we didn't really have that much of a clear picture honestly) was the capability to enable one (1) ray traced effect in most games at decent resolutions (1080p and higher) at 60 fps. And I don't think that those expectations were that unrealistic.

 

[Shifty Geezer]

Well we're getting heavily into territory the mods dont want us to get into, even though it's invariably all interwoven together. You're very right to ask that question, though. In fact, I really think moreso than trying to talk specs, the question about the financial aspects should actually be the main priority,

Feel free to hold those discussions in the Industry forum. This thread discusses what you'd pick for a given BOM, and doesn't try to justify that BOM - just say, "I think they'll spend this much money and get this hardware."

If you want an attempt to predict what the console companies will spend on hardware to inform your discussion in this thread on how that money will be spent, you want a new thread in Industry, "Next-gen console pricing - what do you think they'll cost and what hardware margins will they have" or somesuch.

 

[Seanspeed]

My expectations before launch (and we didn't really have that much of a clear picture honestly) was the capability to enable one (1) ray traced effect in most games at decent resolutions (1080p and higher) at 60 fps. And I don't think that those expectations were that unrealistic.

We knew that RDNA2's hardware ray tracing implementation was quite barebones.

There was not much real dedicated RT hardware, they were just repurposing the TMU to handle a small portion of the RT workload in terms of calculating actual RT 'hits' on pixels, but everything else about the RT process was still up to the main shader cores. It was a fairly weak setup.

I dont know what else to say other than that you definitely should not have expected so much from these consoles in terms of RT and there were definitely plenty of us who understood this beforehand.

 

[Seanspeed]

Feel free to hold those discussions in the Industry forum. This thread discusses what you'd pick for a given BOM, and doesn't try to justify that BOM - just say, "I think they'll spend this much money and get this hardware."

If you want an attempt to predict what the console companies will spend on hardware to inform your discussion in this thread on how that money will be spent, you want a new thread in Industry, "Next-gen console pricing - what do you think they'll cost"

It's like if somebody posted asking how to play a certain song on guitar, but a mod intervened and demanded that we have one topic about what we should do with the left hand, and a separate topic about what we should do with the right hand.

 

[MfA]

wonder if it'll be possible to run CP77 path traced in VR with a 5090.

Unless Intel brings out a Vulkan version of GFFE or NVIDIA does extrapolation the latency will be painfull. NVIDIA are probably hesitant to do it because it prevents a drop in replacement of existing DLSS, it needs some extra game engine support. Less than "ray reconstruction" though, maybe a bit of stubborness and laziness too

PS. at some point when they are good and ready NVIDIA will probably build single frame to stereo conversion into DLSS too.

 

[rntongo]

As we get further and further into the gen, the graphics will improve. We'll start to see projects with more recent versions of UE5 which will look better when compared to cross gen games. However, i don't think that graphics or raytracing will be the big sell people on here think it will be. In fact, I think we may even see a longer cross-gen period next gen than we saw this gen. Despite the increase in image quality, power, and frame rate, people still refused to upgrade to the ps5 as quickly as they did from ps3->ps4.

You could argue that covid hampered the progress and that's true. However, from current trends, we can see that people are playing more old games than they are new games. If the sell next gen is, more raytracing paired with static, non interactive set dressed worlds, it's going to be a seriously tough sell. The games themselves have fallen into the hollywood trap of releasing visually appealing but thoroughly uninteresting content. It just doesn't sell. More attention needs to be paid to other aspects of the games like the gameplay, physics, etc. That's where Sony and co should look to innovate/differentiate themselves from the pack.

You bring up valid points. People are still going to be buying PS5s in 2028-2030, this may be the most profitable console Sony has ever sold. Especially as you have stated when the second half of the gen produces titles that more efficiently utilize the hw and showcase the creativity of the studios.

But with the next gen, Baseline rasterization performance has to improve, but most of the improvements next gen will come from improved hw acceleration for AI and Raytracing, thats where there is more room for growth. Coupled with a larger initial investment in the memory subsystem and you get a true generational leap. This is why I think raster performance will be equivalent or slightly better than a a 6 year old GPU 7900XTX or 4090. For the CPU, I think they'll settle for a 16 core x86 CPU to keep the costs reasonable. Compared to the current gen these could be clocked at or above 5Ghz.

To achieve the generational leap coupled with simpler SDLC and short time to triangle, they really need to increase the amount of RAM like way over 32GB, over the lifetime of the next gen this makes it affordable.

CPU: 16 core CPU 5Ghz
RAM: 48 GB RAM using GDDR7(4GB modules)36-42Gbps per pin, 1.7-2TB, memory bandwidth, 384 bit bus.
GPU: dedicated hw for raytracing and AI as well as equivalent or slightly faster raster performance to RTX 4090/7900XTX i.e ~96-100 CU GPU

 

[Shifty Geezer]

It's like if somebody posted asking how to play a certain song on guitar, but a mod intervened and demanded that we have one topic about what we should do with the left hand, and a separate topic about what we should do with the right hand.

If that forum was crazily split into one forum for left hand and one forum for right hand, that's how that discussion would have have to be split. There's no (obvious) logic for such a distinction though, unlike the split between business and technical talk which is necessary to stop those who can't talk tech from polluting technical discussions. If not enforced this way, this thread will become another thread talking about MS's subscription strategy and the better value of PCs.

 

[Charlietus]

We knew that RDNA2's hardware ray tracing implementation was quite barebones.

There was not much real dedicated RT hardware, they were just repurposing the TMU to handle a small portion of the RT workload in terms of calculating actual RT 'hits' on pixels, but everything else about the RT process was still up to the main shader cores. It was a fairly weak setup.

I dont know what else to say other than that you definitely should not have expected so much from these consoles in terms of RT and there were definitely plenty of us who understood this beforehand.

Let's leave behind this discussion about expectations of current consoles. Then, what do you expect for the next ones?

If Sony and AMD (will Microsoft contribute too?) can't deliver on the expectations set by the Cerny's seminar, than it's going to be pretty difficult to spark enthusiasm to sell the consoles.

It's not that they should make a AI-ray accelerated something console, it's what they have to do. They don't have much choice, that's what investors want. Right now AI is the magic word. It's fine to doubt AMD, but when consoles were involved, they delivered a pretty good architecture for the time (RDNA2).

Much more than in prior generations, the power level of the consoles will be decided by what TSMC or maybe Samsung will be able to deliver at a good price. So unless there is some miracle process node that delivers performance/watt while being cheap, we can already say that we are not getting a 5090, maybe not even a 5080, at 500$.

 

[MfA]

Wonder if they'll just scale up xdna 2 by 10x for image based AI workloads.

 

[VFX_Veteran]

This is just speculation on your part, the only difference is that it is an outdated speculation based on previous trends. But we know very well that the previous cheap console philosophy is no longer profitable for large companies. If they want innovation and "this will be the biggest technological leap" promised by Microsoft, then they will sell the next console more expensively with more advanced components, especially if it will actually be a unique PC. The traditional console concept is becoming less and less important.

MS will compete differently, the next Xbox could easily be considered only the brand's flagship, similar to Nvidia's 5090 VGA. Thus, we can expect powerful and modern hardware from them with more expensive pricing.

The major problem with these type of threads is that people assume that the hardware will be made during the same year it's released. This has happened over and over again. This generation was a huge letdown for the amount of speculation people were "wishing" up from the ether.

My speculation is just a speculation also, however it's not a wish. It's based on what we have seen from AMD so far and what we can reasonably assume in 3yrs time. They aren't making high-end hardware anymore. And if they are planning for a console release in 2028, the hardware is 90% chosen and being worked on right now with today's tech -- not 2028.

 

[raytracingfan]

And if they are planning for a console release in 2028, the hardware is 90% chosen and being worked on right now with today's tech -- not 2028.

I suppose this depends on what you mean by "today's tech". If you mean the next-gen consoles are going to use tech already in development today - like Zen 6 and UDNA, then I agree. If you mean they'll be using tech that's already/about to release like Zen 5 and RDNA 4 that's too pessimistic - the current-gen consoles have Zen 2 and RDNA 2 and released a year after Zen 2 and the same year as RDNA 2. They definitely won't feature 2028 flagship-level performance though, not even in comparison to the Radeon lineup (whether AMD returns to the high-end with UDNA or not). Best case scenario is 9800X3D CPU performance, raster performance slightly beating 7900XTX, RT and ML performance slightly beating 9070XT.

 

[VFX_Veteran]

I suppose this depends on what you mean by "today's tech". If you mean the next-gen consoles are going to use tech already in development today - like Zen 6 and UDNA, then I agree. If you mean they'll be using tech that's already/about to release like Zen 5 and RDNA 4 that's too pessimistic - the current-gen consoles have Zen 2 and RDNA 2 and released a year after Zen 2 and the same year as RDNA 2. They definitely won't feature 2028 flagship-level performance though, not even in comparison to the Radeon lineup (whether AMD returns to the high-end with UDNA or not). Best case scenario is 9800X3D CPU performance, raster performance slightly beating 7900XTX, RT and ML performance slightly beating 9070XT.

AMD doesn't have competing GPU hardware for RT today compared to last-gen's Nvidia boards. The 4090 was such a huge leap from 3x to 4x, I don't see the next-gen consoles having that equivalent. Raster performance isn't really going to be a big factor in 2028 as most studios will have full RT game engines and will have moved away from raster pipelines. The CPU won't really matter much as long as it's within reasonable speeds to not bottleneck the GPU. Those are my rationale thoughts on the matter.