Next Generation Hardware Speculation with a Technical Spin [2018]

[anexanhume]

The situation is kinda different now though, isn't it?

Back then, Sony had just one PS4 model; now they have two already. With a "PS5" (let's call it, and we'll also assume it's fully backwards compatible too because anything else would be nonsense); then they'd have 3 consoles in their stable; that's kinda crowded, don't you think?! So which one are they going to discontinue, Pro, after so short a run on the market? It'd look bad and would annoy those who bought it.

They're NOT going to discontinue PS4 vanilla; that's their entry level model, the most affordable one with the highest sales potential. So they'd be inbetween a rock and a hard place I'd say. 3 consoles would cannibalize sales from each other, but there's no good strat for eliminating one console.

If you instead wait until at least fall of 2020; by then vanilla PS4 will start looking rather long in the tooth, putting it out to pasture might not be such a bad idea then, and PS4 Pro could thusly be moved down a notch in the pecking order. There's now room at the top for a new heavyweight champion; "PS5".

The Pro will always be appreciably more expensive to produce than the slim. They’ll want the lowest base price possible for a legacy console. I don’t see how the Pro lives after PS5 is launched. It’s not going to appreciably affect cross gen games, and once everyone has moved onto PS5, the power difference won’t matter at all IMO. People who want power will buy PS5. People who want cheap access will buy super slim or whatever.

 

[Grall]

The Pro will always be appreciably more expensive to produce than the slim.

Not necessarily true; if it was some sort of universal constant then today's computers and consoles would cost more money than the total worth of entire nations. Computing power goes down in price over time, even today, and if you keep old baseline models alive for too long they start to become an anchor dragging down games/software development. We saw that happening during the overly long and drawn-out PS360 era.

 

[MrFox]

Not necessarily true; if it was some sort of universal constant then today's computers and consoles would cost more money than the total worth of entire nations. Computing power goes down in price over time, even today, and if you keep old baseline models alive for too long they start to become an anchor dragging down games/software development. We saw that happening during the overly long and drawn-out PS360 era.

I think the difference will become smaller over time, but the Pro is practically the same hardware except it has a bigger psu and a bigger SoC. The main SoC will always keep it more expensive than the slim. And with cost per transistor not dropping very fast, it might keep a $50 premium on the BOM until one of them is discontinued. Not sure which one, to be honest, I think they might keep both for a while. There are pros and cons to every choices.

 

[anexanhume]

Not necessarily true; if it was some sort of universal constant then today's computers and consoles would cost more money than the total worth of entire nations. Computing power goes down in price over time, even today, and if you keep old baseline models alive for too long they start to become an anchor dragging down games/software development. We saw that happening during the overly long and drawn-out PS360 era.

I simply meant that the die will always need to be bigger and harder to yield, the memory slightly more expensive, it has the system RAM, and the cooling and power solutions must be slightly larger. They will asymptotically approach some base price with a delta of $20 or so to produce, I would venture.

I think the difference will become smaller over time, but the Pro is practically the same hardware except it has a bigger psu and a bigger SoC. The main SoC will always keep it more expensive than the slim. And with cost per transistor not dropping very fast, it might keep a $50 premium on the BOM until one of them is discontinued. Not sure which one, to be honest, I think they might keep both for a while. There are pros and cons to every choices.

Woops, we basically had the same reply simultaneously. I don’t think the Pro will be discontinued straight away. They’ll probably let the sales in the market determine when to stop producing it. It has to compete for shelf space though, and too many variants could lead to customer confusion.

 

[Tkumpathenurpahl]

I reckon we'll see the following:

2018 - PS4Pro Slim. 12nm Zen clocked at 2.1GHz wouldn't howl nearly as much as the current Jaguar cores.

2019 - PS4 Super Slim, and PS4Portable. I'd love a dock and processing unit solution to cover both of these simultaneously, but separates may be simpler for the market.

2020 - PS5 and PSVR2. Vega 64 level performance in the console, and a wireless, 4K headset.

2021 - PS5 portable. Not full PS5 performance, but something more than the Pro. Able to play all PS4 games - with PS4Pro settings if selected and applicable - and all PS5 games in a manner similar to the Switch's portable mode.

 

[Xbat]

I don't see the Pro or the base PS4 at the same level or family as the PS5. Yes the PS5 will hopefully be able to play PS4 games but the Pro or base ain't going to play PS5 games. So yeah most people who bought a Pro will upgrade pretty quickly to PS5 and then it's up to Sony to decide if they can manufacture a PS4 with Pro specs as the cheapest console otherwise the market for the it will disappear.

It still comes down to if the PS5 they can release in 2019 is practically the same as the one that could be made in 2020 except for price I don't see why they don't just release it in 2019, It's a big if though .

 

[anexanhume]

Micron is ramping their 10nm DRAM, so that’s good news for GDDR6 supplier diversity.

https://www.digitimes.com/news/a20180425PD204.html

Edit: more TSMC 7nm+ tid-bits. If HVM goes live early next year, that would be a full year before 2020 consoles would need to ramp. Not completely far-fetched

 

[anexanhume]

Sorry for the double post, but it’s been a couple days, and there’s a lot of good info (heh) in Rick’s full post.

TSMC announced it is in volume production with a 7nm process and will have a version using extreme ultraviolet (EUV) lithography ramping early next year.

The only bad news is the advantages with the new process nodes are getting thinner. The new normal for performance gains and power reductions generally fall in a 10-20 percent range, a reality that makes the new packaging and specialty processes increasingly important.

TSMC is in volume production of 7nm chips today with more than 50 tape-outs expected this year. It’s making CPUs, GPUs, AI accelerators, cryptocurrency mining ASICs, networking, gaming, 5G and automotive chips.

The node delivers 35 percent more speed or uses 65 percent less power and sports a 3x gain in routed gate density. By contrast the N7+ node with EUV will only deliver 20 percent more density, 10 percent less power and apparently no speed gains — and those advances require use of new standard cells.

TSMC has validated in silicon what it calls foundation IP for N7+. However, several key blocks will not be ready until late this year or early next year, including 28-112G serdes, embedded FPGAs, HBM2 and DDR5 interfaces.

Looking ahead, TSMC aims to start risk production of a 5nm node late next year, focusing on mobile and high-performance computing chips.

Compared to the initial 7nm without EUV, the 5nm node promises a 1.8x greater density than 7nm. However, it is only expected to reduce power by up to 20 percent or raise speeds by about 15 percent, perhaps 25 percent using a new device option.

Now that TSMC has established its 2.5-D CoWoS package in GPUs and other processors and its wafer-level fan-out InFO in smartphone chips, it is expanding both offerings and adding others.

CoWoS chips will have options for silicon interposers up to twice a reticle's size, apparently stitched in the field, starting early next year. Versions with 130-micron bump pitch will be qualified this year.

The InFO technique is getting four cousins. Info-MS, for memory substrate, packs an SoC and HBM on a 1x reticle substrate with a 2x2-micron redistribution layer and will be qualified in September.

InFO-oS has a backside RDL pitch better matched to DRAM and is ready now. A multi-stacking option called MUST puts one or two chips on top of another larger one linked through an interposer at the base of the stack.

But that’s not all. TSMC introduced two wholly new packaging options.

A wafer-on-wafer pack (WoW) directly bonds up to three dice. It was released last week, but users need to ensure their EDA flows support the bonding technique. It will get EMI support in June.

Finally, the foundry roughly described something it called system-on-integrated-chips (SoIC) using less than 10-micron interconnects to link two dice. Details of the process and its target apps was sketchy for the capability that will be released sometime next year.

 

[Xbat]

So unless I'm missing something the end of 2019 looks viable for a next gen console?

 

[DieH@rd]

So unless I'm missing something the end of 2019 looks viable for a next gen console?

Small chance. 2020 is more realistic.

 

[eloic]

Small chance. 2020 is more realistic.

Small chance which I hope is not taken, honestly. I don't want a rushed, underpowered console.

 

[Deleted member 13524]

Late 2019 is realistic for widely available 7nm, since that's over 1 year after volume production starts (which was during these latest weeks).
That said, there will be no substantial difference between what can be made in late 2019 and 2020. GPU arch will be Navi, CPU arch will be Zen, process will be 7nm DUV for both years.Only thing that could change would be HBM3 but even that is a big question mark for 2020, and HBM2 seems to be scaling up in frequency rather nicely, while keeping voltage down.


Unless console makers will be waiting for 7nm EUV for lower SoC costs, which could make even 2020 out of reach. 2021 would be way too late IMO.

 

[anexanhume]

Late 2019 is realistic for widely available 7nm, since that's over 1 year after volume production starts (which was during these latest weeks).
That said, there will be no substantial difference between what can be made in late 2019 and 2020. GPU arch will be Navi, CPU arch will be Zen, process will be 7nm DUV for both years.Only thing that could change would be HBM3 but even that is a big question mark for 2020, and HBM2 seems to be scaling up in frequency rather nicely, while keeping voltage down.


Unless console makers will be waiting for 7nm EUV for lower SoC costs, which could make even 2020 out of reach. 2021 would be way too late IMO.

Agree. You have to push into 2021 to bring things like post Navi, Zen 3, and 7nm+ into the realm of possibility. HBM3 is interesting, but you can get a lion’s share of the bandwidth with a 384-bit GDDR6 interface.

 

[Xbat]

Small chance which I hope is not taken, honestly. I don't want a rushed, underpowered console.

I don't see how there will be any more improvement by waiting till the end of 2020 other than price and maybe a slightly faster clock speed.

 

[eloic]

I don't see how there will be any more improvement by waiting till the end of 2020 other than price and maybe a slightly faster clock speed.

That's what we may think now. But to get to 2020 there're still two years ahead. Who knows how tech may improve, etc.

 

[Shifty Geezer]

That's what we may think now. But to get to 2020 there're still two years ahead. Who knows how tech may improve, etc.

The hardware companies, hopefully, as they have (public) roadmaps.

 

[Deleted member 13524]

Who knows how tech may improve, etc.

Pretty much everyone..
The general roadmaps for AMD's CPU/GPU architectures and Samsung/GF/TSMC's foundries that cover the next 2 years are public.

 

[eloic]

The hardware companies, hopefully, as they have (public) roadmaps.

Pretty much everyone..
The general roadmaps for AMD's CPU/GPU architectures and Samsung/GF/TSMC's foundries that cover the next 2 years are public.

That much, guys? Disappointing.

 

[Deleted member 13524]

That much, guys? Disappointing.

Publicly traded companies need to have public roadmaps to attract shareholders...

 

[Tkumpathenurpahl]

As much as the roadmaps are laid out, isn't AMD's "next-gen" architecture a bit of a wildcard?