Next Generation Hardware Speculation with a Technical Spin [2018]

[borntosoul]

The software has to catch up with the hardware too, you could release 2019 but what meaningful difference would you have to today’s games? Gaming budgets would have to increase at the same rate as hardware ......so the way l see it is that all these factors would come together in 2020+ , that’s to creat games noticeably different to what we have at the moment.

 

[Xbat]

I'm not sure why people have it in their minds that these companies schedule hardware releases when certain silicon technologies and nodes become available like that's the reason they release new hardware and a whole new ecosystem. It's very much a forum/enthusiast mindset.

Well it does play a part, for instance hypothetically if 7nm was going to be ready by 2019 but 5nm was going to be up and running and feasible by 2020 I'm sure Sony or Microsoft wouldn't want to launch in 2019 with 7nm when possibly the other could launch a year later on 5nm and have quite a power advantage at the same price.

Latest Sony financials indicate they expect to ship less consoles this year and therefore even less in 2019 which points to that they starting to saturate the market at there current price.

On another note of crazy speculation I'm starting to think that maybe Microsoft will be the first mover in 2019, the onus is on them to try and gain market share from Sony. Where Sony still has some big first party exclusives planned for 2019 on PS4 and I can't think of any of there big first party studios that will have anything ready for 2019 or even 2020 for that matter.

They could market the PS5 as the place to play the big PS4 exclusives still coming out at native 4k and locked 60fps, that shouldn't be a problem at all for PS5. Would also explain why they are shipping so few PS Pro's.

 

[anexanhume]

I'm not sure why people have it in their minds that these companies schedule hardware releases when certain silicon technologies and nodes become available like that's the reason they release new hardware and a whole new ecosystem. It's very much a forum/enthusiast mindset.

They have to commit to solutions, and we’re just debating the optics of how those decisions are made. Sony had to delay PS3 a year in part due to the RSX, if the rumors are accurate.

 

[milk]

Are you saying that, instead of releasing a PS5 for 499 in 2019, Sony should not have any new products for a few years, so that they can release a PS5 for 399 in 2021?

I think he is saying SONY would rather not release yet another new next-gen console for 499 while the current gen hasn't been cost-cut enough to tap the 199/249 market so much. There are consumers to be gained there too. While having new premium products is good, it also can caninilize some of your long tail, in which there is also some good money to be had, and brings more consumers into your platform and ecosystem.

 

[Shifty Geezer]

I think he is saying SONY would rather not release yet another new next-gen console for 499 while the current gen hasn't been cost-cut enough to tap the 199/249 market so much. There are consumers to be gained there too. While having new premium products is good, it also can caninilize some of your long tail, in which there is also some good money to be had, and brings more consumers into your platform and ecosystem.

Historically, I don't think that's true. Software sales for the old platforms drops off a cliff from the second year of the new platform. However, there's never been anything comparable to significant, late-in-the-generation price drops. Price drops happened very progressively before PS3, hitting $200 very early (year 1 or 2) and dropping to $150 and below after that. PS360 never dropped below $250 (ignoring gimped Arcade in year 3). So we've no precedent for a $400+ machine hanging on to $250+ for five/six years and then dropping to $150-200 as the next gen arrives. That price point may be impossible, of course. Looking at hardware sales, assuming there's 70 million ish customers interested in a <$200 console, there might be good money in sustaining these machines. Guaranteed forwards library compatibility would make them a safer investment too.

 

[Deleted member 13524]

I'm not sure why people have it in their minds that these companies schedule hardware releases when certain silicon technologies and nodes become available

Because it's an historically accurate assertion.
The PS3 and the X360 were launched 1 year after the 90nm process was available for volume production at IBM foundries. The PS4 and XBone were launched less than 2 years after the 28nm was available for volume production at TSMC. The PS4 Pro, PS4 Slim and Xbone S launched 1 year after the 16FF process was available for volume production at TSMC. XboneX launched 1 year after the others, but in that case Microsoft was probably just making sure they'd have a significantly higher performance console than the Pro.

The availability/maturity of a new process will probably determine the ~12 month release window for the next generation of consoles. Just as it did for the last ~15 years.


Within a larger window, other events that dictate the time period for launching new consoles are how fast are the cheap PC solutions catching up, the peak of annual console sales and the timing for the "visual apex" attained with the current console hardware.
As mentioned before, the peak for console sales has already happened for the PS4 during 2017, and Raven Ridge already brought XBone performance for $99 with the Ryzen 3 2200G.

The apex of PS4's capabilities may have already happened with God of War 4 or even Horizon Zero Dawn.
If not, it will definitely be seen in TLoU2 (Naughty Dog does have ICE in their midst..) or Death Stranding during 2019 but the upgrade in visuals will be minimal IMO. Both consoles were designed to reach that peak earlier than in previous generations.
After that, Sony will have to keep pushing for better looking games, and that can only be possible by breaking compatibility with (at least) the 2013 PS4 and targeting higher performing hardware.

 

[dten]

Because it's an historically accurate assertion...

It's not based on nodes. Engineers don't run companies, business people do. It's based on the market and the node just opens the door. That's why you get anywhere from 4 to 8 years between releases and as much as 4 node shifts. In the history of consoles, it is often the distant second place that launches first. Or PS3/360 when it was essentially even, dropping game sales and the increasing distance, especially RAM, to PC forced their hand after 7-8 years. These guys are about making money and want to milk their consoles for as long as it can be sustained.

 

[Tkumpathenurple]

The market plays a role. The state of competition informs the role to play in the market.

The technology plays a role. The available nodes and cost thereof inform the price and performance.

Is there any chance that we could nip this particular debate in the bud? Because it's going to go in circles like people arguing over whether the tyre or the axel is more important to a car.

 

[dten]

The market plays a role. The state of competition informs the role to play in the market.

The technology plays a role. The available nodes and cost thereof inform the price and performance.

Is there any chance that we could nip this particular debate in the bud? Because it's going to go in circles like people arguing over whether the tyre or the axel is more important to a car.

This is exactly right that there are many facets. It's not so much arguing whether tyre or the axel is more important, as I agree that will go in circles, but not forgetting that they're both needed in a well rounded analysis.

We can guess with some ballpark accuracy what is possible in 2019, 2020, 2021 or whatever. But picking actual years which is what the last few pages seems to do is far more complicated than silicon and power. Just pointing that out.

 

[Deleted member 13524]

It's not based on nodes. Engineers don't run companies, business people do.

Did you happen to miss the second half of the post you replied to?

 

[dten]

Did you happen to miss the second half of the post you replied to?

Yes, I was just replying to that sentence because that was my original assertion. The points about the market situation actually back that up.

 

[Deleted member 13524]

Did you happen to miss the second half of the post you replied to?

Yes

Okay then.
(see what I did there?)

 

[MrFox]

This is exactly right that there are many facets. It's not so much arguing whether tyre or the axel is more important, as I agree that will go in circles, but not forgetting that they're both needed in a well rounded analysis.

We can guess with some ballpark accuracy what is possible in 2019, 2020, 2021 or whatever. But picking actual years which is what the last few pages seems to do is far more complicated than silicon and power. Just pointing that out.

Of course, all aspects of the question remain fluid. I guess we have been doing both for quite a while, first picking a year as an exercise (testing rumors of the day) to figure out what would be possible, and rinse, repeat, for another year. Once we figured out roughly what can be done in 2019, 2020, 2021, we can then argue which one is sufficient to make a business case for a next gen at that time.

Based on past generation transitions, next generation timing is somewhat tied to the ability to produce a significant improvement at a maimstream console price point. If there is no significant technological improvement, the current generation will continue to be milked until something happens. Big jumps are mostly tied to node transitions, new cpu/gpu architectures, new paradigms (from sprites to 3d, gpgpu, new distribution models or media, etc), or all that together.

2019 is very unlikely. 4.2/6TF is not going to look much worse than 8-10TF. That makes the case for waiting a couple more years. But OTOH, the cpu could be mid-gen's achille's heel. So a much better cpu, paired with only a 10TF gpu, would be a good argument for a next gen transition, then gpu upgraded to 24TF or something for a later mid-gen. New hardware with ray tracing helpers could be magical for (hybrid) GI. I think it's not crazy, just unlikely.

At the other extreme we have a 2022 launch on 5nm+ or something. Waiting too long is dangerous, sales dip when the generation lingers, and competitors with fresh new hardware get the upper hand. Unless there is a major tech ready only in 2022, it creates an up hill battle.

I see 2020 and 2021 as good candidates as long as they get the significant tech improvement on time. So my biggest question is... Will they?

 

[Silent_Buddha]

After that, Sony will have to keep pushing for better looking games, and that can only be possible by breaking compatibility with (at least) the 2013 PS4 and targeting higher performing hardware.

Why? The architechture of the PS4 and PS4-P are extremely similar, even having virtually the same memory pools.

A game made to exploit PS4-P to 100% or as near as possible to 100% of its capabilities will still have absolutely no problem running on PS4 at reduced settings.

Regards,
SB

 

[Deleted member 13524]

A game made to exploit PS4-P to 100% or as near as possible to 100% of its capabilities will still have absolutely no problem running on PS4 at reduced settings.

A game that makes heavy use of FP16 on the Pro will be forced to look like crap on the PS4 because of the large differences in ALU throughput and bandwidth (PS4 probably still uses full FP32 register size for FP16 calculations so there's zero benefit).
We're talking pixel shaders and A.I. inferencing, for example.
I'm afraid the PS4-Pro may never be stretched to its max capabilities because of that.

 

[borntosoul]

Which company with billions at stake would not......
1) Have road map information from suppliers giving details on process technology.
2) Have solid information years in advance on upcoming tech and how powerful it may be.
3) Have already spoken to software companies years in advance of launch.
4) Have a bunch of ultra smart hardware guys trying to meet a given performance target based on expected costs.
5) all these things are work in progress years in advance and are tweeked to suit.

The pro came out for two reasons...VR and 4K and that’s it.
There must be a compelling reason to bring out a new generation otherwise it is pointless.

 

[Pixel]

I'm not sure why people have it in their minds that these companies schedule hardware releases when certain silicon technologies and nodes become available like that's the reason they release new hardware and a whole new ecosystem. It's very much a forum/enthusiast mindset.

I wouldn't be surprised if you are confusing what alot of forum members are saying. Of course they don't schedule hardware releases when certain silicon technologies and nodes become available. But in the age of diminishing returns and the breakdown of Moore's law they are keenly aware they need to wait until it becomes viable to bring a large jump in hardware performance, and the processes technology is intrinsically tied to that.

It DOESNT mean SNY/MS will jump on the latest and greatest nodes as fast as like Apple or Samsung do for flagship Smartphones and Nvidia does for deep learning $3000-8000 GPUs. I doubt most of the forum members here think that they are going to time the release of PS5 6 months after TSMC hits volume production of 7nm finFET. The forum mindset is that they need to wait at least until they can create a powerful enough console that will entice another 160+ million gamers to buy the PS5/Xbox2. That isn't happening on TSMCs 12nm finFET processes.

Check out this interview with Ex-Sony Head Andrew House where he briefly mentions future processes technologies being one of a the key things regarding future consoles.

 

[Shifty Geezer]

1) Have road map information from suppliers giving details on process technology.
2) Have solid information years in advance on upcoming tech and how powerful it may be.

Paper specs. 7800 looked okay on paper but in reality is stank, sapping up a lot of Cell's power to help with graphics. Cell looked great on paper but devs had trouble maximising it. You never really know just how good the hardware is when final silicon is put into practice.

3) Have already spoken to software companies years in advance of launch.

Plenty of software devs went into PS3 gen not knowing how powerful it was, creating visions based purely on hype. See MGS concept trailer for example. You can give some peak numbers but it's hard to look at that and appreciate just what shaders you can implement or not. Another example, Xenos had tessellation in hardware but it never really did anything. Yoiu'd see that spec on paper, think what you could do with, then in reality find it's too limited and you don't use it.

The pro came out for two reasons...VR and 4K and that’s it.

The Pro came out because there was a market for a better PS4, and the marketing wasn't 'VR' or explicitly '4K', and the reasons given were to offer a higher tier to stop bleeding to PC. Unless you can present hard numbers correlating 4K/VR with Pro sales, it's safe to assume it'd likely sell the same numbers without the existence of 4K TVs and VR.

 

[borntosoul]

Paper specs. 7800 looked okay on paper but in reality is stank, sapping up a lot of Cell's power to help with graphics. Cell looked great on paper but devs had trouble maximising it. You never really know just how good the hardware is when final silicon is put into practice.
Plenty of software devs went into PS3 gen not knowing how powerful it was, creating visions based purely on hype. See MGS concept trailer for example. You can give some peak numbers but it's hard to look at that and appreciate just what shaders you can implement or not. Another example, Xenos had tessellation in hardware but it never really did anything. Yoiu'd see that spec on paper, think what you could do with, then in reality find it's too limited and you don't use it.
The Pro came out because there was a market for a better PS4, and the marketing wasn't 'VR' or explicitly '4K', and the reasons given were to offer a higher tier to jstop bleeding to PC. Unless you can present hard numbers correlating 4K/VR with Pro sales, it's safe to assume it'd likely sell the same numbers without the existence of 4K TVs and VR.

A few points.....

Cell was different, Sony took a big risk and Sony’s ethos at the time was to creat a cpu that had lots of untapped potential, that could get better over the life of the consol as developers learnt better ways its code on it- it was a Janpenese Thing. As we know the ps4 was the opposite, easier and more eccesable and born from the mistakes of cell.
They’ve been there done that and learnt from their mistakes - that is why Sony will have a much better idea of how the ps5 will perform from both hardware and software point of view - hardware will be similar style to what came before and devs will be much more up to speed then what they ever were.

In saying what l said it doesn’t mean everything is fool proof.

The market for a better ps4 was to market a system that had some 4K ability and VR. Otherwise there was no point. The ps4 sales were still strong and it was basically an opportunity to cater to the 4K t.v market that was about to boom.

 

[Silent_Buddha]

A game that makes heavy use of FP16 on the Pro will be forced to look like crap on the PS4 because of the large differences in ALU throughput and bandwidth (PS4 probably still uses full FP32 register size for FP16 calculations so there's zero benefit).
We're talking pixel shaders and A.I. inferencing, for example.
I'm afraid the PS4-Pro may never be stretched to its max capabilities because of that.

Yes it would look like a PS4 game on PS4. I wouldn't say PS4 games look like crap, though. But if you think PS4 games look like crap, then that's your opinion.

That still wouldn't mean not maximizing the potential of a PS4-P.

Regards,
SB