Next Generation Hardware Speculation with a Technical Spin [2018]

Status
Not open for further replies.
Yep, silicon base - massive arrays of single-transistors and row/columns. The capacitor design is the trickier part.
 
Last edited:
its a bit like the orig One and the One S... The same machine but more performing... So they can delay the PS5 to 2021-23 and keep on with the PS4 success
 
Are wafers also needed to create memory?
Yes. The basic technology is the same; sequencial rounds of applications of photoresist, exposure to light, acid etching; vapor deposition of metal and so on as with other semiconductor manufacturing. DRAM and pure logic manufacturing differ in the specifics, though how exactly I could not tell you. :D
 
A One S kind of improvement seems reasonable, but not to the extent of having much effect on its Floppage.

As far as I'm concerned, they need to get it cheap and quiet, and then it's a perfectly acceptable budget 4K console.
 
if -as I know- 7nm will bring consistent wattage reduction but not (at least initially in 2019) transistor budget increased.................
 
I don't see how that article supports that transistor budgets won't increase with a switch to 7nm.
Density improvement of TSMC 7nm is roughly 3x over 16nmFF. Risk production started a year ago and volume production starts now (take a wild guess at who the customer might be :)). The article you referred to takes up concerns with EUV, (by someone with a vested interest in e-beam) which though valid, doesn't seem to be stopping the industry for the immediate future up to and including 5nm entering volume production in a couple of years.
 
Can you please stop posting such ARM suggestions now (and by please, I mean I'll remove future OT posts and possibly issue a reply ban). That is clearly related to nVidia's existing use of ARM in mobile and automobile etc. There's no direct correlation between that and consoles, and the ARM in console discussion needs things like price and performance of actual silicon to be worthy of a tech discussion. So presently your just shilling for ARM and diluting the discussion.
I think ARM in consoles are a real possibility - in fact, it is already there, it is the ISA used by the Switch.
Whether it will actually be used further is a business decision by Sony and Microsoft.
This is a technical thread, so while technology choices are obviously determined by business concerns (so separating the two are tricky), the focus should be on what technological advantages/consequences shifting to ARM from x86 would have. Remember we are only a generation away from when all consoles used PowerPC as their CPU ISA, so a change is definitely possible, and arguably easier than the shift from the PPC cores to x86.
Now, I happen to believe that inertia and an attractive low-margin deal with AMD will make staying with x86 the preferred path. One thing I could see shifting the priorities is if for instance Sony wants to go low power/portable. Ian Cutress at Anandtech, just made a preliminary benchmark of the latest (last?) Atom in a smartphone.
 
No-one's saying ARM isn't possible nor shouldn't be discussed. What's not happening here is constant links to PR materials about businesses forming deals. The reasons behind choosing ARM have already been discussed. What's needed now is actual silicon to compare against the x86 offerings, rather than unsubstantiated claims of performance superiority. If the next ARM is amazeballs and trumps Zen, great. It's an option. But constant, religious like prophesising that it will be great doesn't belong in a fact-based intellectual discussion that relies on evidence and data-points.
 
When properly implemented, EUV is supposed to significantly increase yields and chip quality (i.e. more chips from the waffer will behave as "higher-binned" chips).
The end result is not to get higher-performing high-end chips, but to decrease the production costs per mm^2.
 
Shifting is a business decision but there are also technical difficulties. It is more complicated for Sony to shift than it is for MSFT. MSFT have top notch software ready to translate X86 into ARM, all they need is enough performances headroom on their new system. It is unclear which volume the XBX will reach but the relevant system for BC is the original XBone (for both the CPU and GPU) it should be possible to emulate the system (with benefits) on a completely new architecture (cpu and gpu). Porting the 360 emulator on a new system should not be much of an issue either (looking at software solutions available to MSFT).
MSFT can offer a pretty consistent environment even if they shift to another ISA. The other thing is there are not market leader and to try to regain ground it is more sensical for them to break the mold and possibly offer an incomplete BC/FC at launch IF that allow them to set their system aside from competition in a significant manner. That is tough if both are using AMD tech. As an aside MSFT has created a higher stepping stone for them to climb onto to attract costumers (the XB1X has impressive specs).

Sony is in a different spot now, they are market leader, they ant to keep their costumers, the PRO is doing OK, OK enough so I believe they can keep with pretty incremental upgrade. They are delivering on the software side, their brand value is still stronger outside of US, MSFT has (and try) to make a dent through hardware (misplaced this gen Kinect and not the GPU or a more basic part of the system), it shows on price MSFT has to launch pricier system to compete: original XB1 an the XB1X are priier than the system they are competing with, and not only that MSFT cold be loosing slightly more too.
Imho Sony just has to keep unrolling their strategy while avoiding creating complication for themselves at the hardware level. For their next system Zen cores /CCX should available for their design, so are Navy GPU. On the memory side for the system as whole, HBM1&2, GDDR5x and quite possibly GDDR6.
There are IPs around to build a follow-up to the PS4 PRO and more importantly the PS4 (a massive part of their user base). Building a much better system for 399€ should not be too complicated.

As for a handheld/portable systems, whereas I want to see one to compete against Nintendo Switch it could prove indeed an complicated affair both for Sony and MSFT. Nintendo as its fans and only one (new) system to promote.
If they are done with their bad habit the Switch 2 would be likely to prove a serious issue for wannabee competitors. Then their is the unknown fate of the DS line, a well design and affordable system could also be the last blow to wannabee competitors. It see two options:
-The Switch 2 could use a newer (OTC) Nvidia SOC, significantly more power, etc. As for the new DS? (If any) The Tegra X1? Something that could attract NV and Nintendo (no R&D, easy BC for Nintendo), I believe it is not a good solution for a low end / low power handheld but it could do for sure. Sure a minor revision would be great but Nv does not work for cheap so...
-Nintendo gives-up on the home console functionality which create quite hurdle at the hardware level which translate into costs. They get people from the Tv usage but my believe is that they loose equally (or more) from the price overhead on the system viewed and used as a handheld only. They may be better off with something akin the (defunct) Vita Tv. In that case to claim the market for themselves a custom SOC from Nvidia would be a much better choice than Nvidia OTC shell designs which are two powerful and power hungry to fuel a replacement for both the Switch and DS line. Nintendo could give aces to Nvidia cloud solution (where that is relevant, it is getting relevant in a lot of densely populated areas).

EDIT usual grammar/spelling edit...
 
Last edited:
No-one's saying ARM isn't possible nor shouldn't be discussed. What's not happening here is constant links to PR materials about businesses forming deals. The reasons behind choosing ARM have already been discussed. What's needed now is actual silicon to compare against the x86 offerings, rather than unsubstantiated claims of performance superiority. If the next ARM is amazeballs and trumps Zen, great. It's an option. But constant, religious like prophesising that it will be great doesn't belong in a fact-based intellectual discussion that relies on evidence and data-points.

Not to mention there hasn't been a single moment for the last 17 years that we haven't had an ARM-powered gaming console in the market, and it's not like we haven't seen a fair share of set-top-box consoles (oh look it's a constantly docked Switch with 4K60 video output!) for quite a while, so I really don't get all this sudden fanfare surrounding ARM cores in consoles.
Is it only because ARM solutions are naturally evolving into wider cores because they have enough transistor/die-area to spare?


The Switch 2 could use a newer (OTC) Nvidia SOC, significantly more power, etc. As for the new DS? (If any) The Tegra X1? Something that could attract NV and Nintendo (no R&D, easy BC for Nintendo), I believe it is not a good solution for a low end / low power handheld but it could do for sure. Sure a minor revision would be great but Nv does not work for cheap so...
I'm hoping - but not counting - on Nintendo having to pay the billions that nvidia will inevitably ask just for designing a custom SoC to follow-up on the Switch with backwards compatibility, plus another billion to continue using their vertical stack of software development.

I wonder if NIntendo ever approached Qualcomm for the Switch, as the Adreno line has been the undisputed leader of power efficiency in mobile GPUs for several years and they're focused on the ULP SoC market.
Right now, nvidia couldn't care less about ULP SoCs, and Nintendo is stuck with a super expensive supplier who does not license their GPU IP and will cost them an arm and a leg to develop a follow-up.

Nintendo could give aces to Nvidia cloud solution (where that is relevant, it is getting relevant in a lot of densely populated areas).
Nintendo - who can't for the love of them - get almost anything right regarding online stuff, to pioneer the cloud-powered consoles?
Naaaaah...
 
I'm hoping - but not counting - on Nintendo having to pay the billions that nvidia will inevitably ask just for designing a custom SoC to follow-up on the Switch with backwards compatibility, plus another billion to continue using their vertical stack of software development.
Well NIntendo said they were going with a significant partnership with NV. I also know that PR is PR and everything has to sound good. Nintendo may become more and more reliant on Nvidia tech (including software). May be they already are (and so will pay, think so companies that had ok-ish in house solution that got tricked into the SAP environment condemned to milked ad eternam with no way out of that ecosystem).
I wonder if NIntendo ever approached Qualcomm for the Switch, as the Adreno line has been the undisputed leader of power efficiency in mobile GPUs for several years and they're focused on the ULP SoC market.
Right now, nvidia couldn't care less about ULP SoCs, and Nintendo is stuck with a super expensive supplier who does not license their GPU IP and will cost them an arm and a leg to develop a follow-up.
That's interesting. Clearly Qualcomm had good enough SOC may be the broadband part of the SOC raised the bill too high for Nintendo? May be Nvidia is not that expensive if you go with what they are willing to sell.) I would not underestimate the software side of things. Nintendo may have lost it, they may have no choice but to turn to a more expert partner. Nvidia is "expensive" sure but if they were better ways for many companies to spend their money they would and it would reflect on NV accountability, I would no expect crazy discounts but they will sure get a deal if it makes money and does not alter the workforce focusing on their cores company goals.
Nintendo - who can't for the love of them - get almost anything right regarding online stuff, to pioneer the cloud-powered consoles?
Naaaaah...
... Well that will be through, the companion app... on your phone... thanks to Nv proprietary app...lol
 
Last edited:
Well NIntendo said they were going with a significant partnership with NV. I also know that PR is PR and everything has to sound good.
Significant partnership == nvidia selling Nintendo a 2 year-old SoC that objectively failed in the markets it was built to compete in.
And with nvidia also being a software company they sold Nintendo their game/engine optimization libraries. For sure, the same they already had developed for those Tegra-exclusive PC->Android ports (a market which nvidia abandoned in the meantime).

Brilliant move by nvidia if you ask me. Terrible for Nintendo on the long run, though.

Nintendo may become more and more reliant on Nvidia tech (including software). May be they already are (and so will pay).

As it is right now, nvidia just has Nintendo by the balls.
If Nintendo wasn't so obsessed with backwards compatibility they'd be fine. But it'll be mighty hard for Nintendo to break their chains from nvidia now.


That's interesting. Clearly Qualcomm had good enough SOC may be the broadband part of the SOC raised the bill too high for Nintendo?
Qualcomm makes mobile SoCs without broadband, like APQ8084 (Snapdragon 805) and the APQ8064 (Snapdragon 600).
Not that it would hurt to have a small Cat.4 modem in the SoC for a 4G version of the console, though.

Or if Nintendo was in the business of straight out purchasing older chips, they could simply use the Snapdragon 820 that would probably have achieved significantly higher sustained performance than the TX1, at least in mobile mode.
Earlier last year we had ~$200 chinese phones with the S820, so Qualcomm was probably selling it for cheap at the time.

Though I imagine Qualcomm could be able to make e.g. a SoC with 6 or 8 Kryo cores at 1.5GHz, 2*32bit LPDDR4 1866MHz (30GB/s), the same Adreno 530, small low-end Hexagon DSP for pictures/video and a Cat.4/5 modem at less than 100mm^2 using 14LPP.
Or they could just go all-out with a wider GPU of the Adreno 500 family and quad-channel LPDDR4 with a slightly larger die area.

At the rate at which Qualcomm pumps out different SoCs every year, I doubt they would find it hard to make a custom order for Nintendo.
 
Significant partnership == nvidia selling Nintendo a 2 year-old SoC that objectively failed in the markets it was built to compete in.
And with nvidia also being a software company they sold Nintendo their game/engine optimization libraries. For sure, the same they already had developed for those Tegra-exclusive PC->Android ports (a market which nvidia abandoned in the meantime).

Brilliant move by nvidia if you ask me. Terrible for Nintendo on the long run, though.



As it is right now, nvidia just has Nintendo by the balls.
If Nintendo wasn't so obsessed with backwards compatibility they'd be fine. But it'll be mighty hard for Nintendo to break their chains from nvidia now.



Qualcomm makes mobile SoCs without broadband, like APQ8084 (Snapdragon 805) and the APQ8064 (Snapdragon 600).
Not that it would hurt to have a small Cat.4 modem in the SoC for a 4G version of the console, though.

Or if Nintendo was in the business of straight out purchasing older chips, they could simply use the Snapdragon 820 that would probably have achieved significantly higher sustained performance than the TX1, at least in mobile mode.
Earlier last year we had ~$200 chinese phones with the S820, so Qualcomm was probably selling it for cheap at the time.

Though I imagine Qualcomm could be able to make e.g. a SoC with 6 or 8 Kryo cores at 1.5GHz, 2*32bit LPDDR4 1866MHz (30GB/s), the same Adreno 530, small low-end Hexagon DSP for pictures/video and a Cat.4/5 modem at less than 100mm^2 using 14LPP.
Or they could just go all-out with a wider GPU of the Adreno 500 family and quad-channel LPDDR4 with a slightly larger die area.

At the rate at which Qualcomm pumps out different SoCs every year, I doubt they would find it hard to make a custom order for Nintendo.
Exactly - Nintendo is in a position where they have options if their relationship with nVidia turns sour. That’s crucial, even if they are likely to have something forward looking in their contract.
It would be very, very interesting to see what the adreno guys could achieve if allowed to stretch their legs a bit (Hi sireric!). And a lovely twist of fate if they could compete with AMD.
Not that I think it will happen. But it is possible, and compared to a hard lock-in, that makes a difference.
 
I honestly can't see Nvidia getting one over on Nintendo. I mean hell, Nintendo already canceled on them once and switched to the pica chip in the 3ds :p
 
I'm not holding my breath for next gen console on 2019. PS pro and Xbox one x sales make it look like there is no demand for slightly more powerful console. Sony probably had plans for 2019 in case Xbox one x had turned the tide for Ms.

My guess for Sony is vanilla PS4 will keep selling for a real long time. Come next gen PS4 will be the cheapest option and ps pro will be phased out. I'm expecting late 2020 at earliest for launch. I would expect modest flop increase but I also expect more capable hw in terms of what can be greatly accelerated compared to current gen.
Neither do I but I believe it would be a good move if they can back it with the proper hardware offer; the ps4 will 6 years old, the PRO three years.
Especially if the stick with AMD and they suspect MSFT won't, they are better off jumping on MSFT and keep having them running behind them.
They have time to work on having the original PS4 software running on the new hardware and that is ll they need to have people making the jump.
 
Status
Not open for further replies.
Back
Top