Predict: Next gen console tech (9th iteration and 10th iteration edition) [2014 - 2017]

[HBRU]

PS4 Pro's physical memory bandwidth is based -as I know- on 256 bit GDDR5-7000... dont know actually GDDR6 frequency (but I guess there will be compatible type with GDDR5-7000), and voltages should be quite compatible too... so memory upgrade toghther with a minor boost in other frequencies should be quite possibile and an easy win for Sony. Hope also MS think about not launching Scorpio with GDDR5

 

[Prophecy2k]

PS4 Pro's physical memory bandwidth is based -as I know- on 256 bit GDDR5-7000... dont know actually GDDR6 frequency (but I guess there will be compatible type with GDDR5-7000), and voltages should be quite compatible too... so memory upgrade toghther with a minor boost in other frequencies should be quite possibile and an easy win for Sony. Hope also MS think about not launching Scorpio with GDDR5

This all seems a little pie in the sky. There is no easy "win" for Sony releasing yet ANOTHER higher end PS4, further fragmenting the user base and making the lives of existing PS4 devs even more difficult (i.e. increasing the number of SKUs they need to test their software on).

Further, Scorpio is a known quantity. It will be as it has been announced to be. There will be now last minute updates as the design has been likely set in stone for over a year now and the console is almost certainly being manufactured en mass as we speak, ready for its Fall 2017 launch.

Given the actual purpose of these "stop-gap" consoles (i.e. Scorpio and PS4Pro), in extending the lifetime of the existing platforms, there's little need for higher memory quantities or memory bandwidth in either machine (well... you could make an argument about bw for PSPro); since both will be supplied with nothing but up-rendered versions of existing PS4/XB1 games for the rest of their respective lives on the market.

PS5/XBNext will likely launch in 2019-2021, so there's absolutely no need to do another mid-gen refresh at any point before then, given the existing business model.

I'm not sure why as a gamer you would want that either?

 

[msia2k75]

256-bit PS4pro bus, with just a memory upgrade, will double its memory bandwidth -using GDDR6- going from today's 218 GB/s to 436 GB/s .... and this without much silicon changes... also the rest can quite easily overclocked to reach and go beyond Scorpio TFLOPS....

I don't see any interest in this.

Just build a proper PS5. Which will include GDDR6 (in conjunction of DDR5 or not remain to be seen)... now a certainty, Ryzen CPU and most likely a GPU with 4096sp (with a couple desactivated).

 

[HBRU]

Well why ? Of course I would prefer a retrocompatible beefed up PS4-PRO-ULTRA than a non retrocompatible PS5, same Sony I guess ....

 

[Prophecy2k]

I don't see any interest in this.

Just build a proper PS5. Which will include GDDR6 (in conjunction of DDR5 or not remain to be seen)... now a certainty, Ryzen CPU and most likely a GPU with 4096sp (with a couple deactivated.

With 4096sp (64 CUs) GPU, your hypothetical PS5 would need to be clocked above 1500 MHz in order to hit even Vega 10 performance (12.5 TFLOPs single precision). I don't think a console APU would be clocked that high.

Vega is going to be manufactured on a 14nm FinFett process I believe, so a 7nm PS5 APU could, in theory, go bigger in terms of CU count and use a lower clock-speed to achieve similar performance to Vega 10, say 72 CUs with a 1360 MHz GPU clock-speed.

 

[iroboto]

With 4096sp (64 CUs) GPU, your hypothetical PS5 would need to be clocked above 1500 MHz in order to hit even Vega 10 performance (12.5 TFLOPs single precision). I don't think a console APU would be clocked that high.

Vega is going to be manufactured on a 14nm FinFett process I believe, so a 7nm PS5 APU could, in theory, go bigger in terms of CU count and use a lower clock-speed to achieve similar performance to Vega 10, say 72 CUs with a 1360 MHz GPU clock-speed.

I almost feel the minimal requirement is 7nm. If they attempt it at 14/16nm I would be pretty confident that we'd be looking at another mid gen refresh. And I don't necessarily know if mid-gen refreshes are going to be common place or not. Part of me says 'yea', but the other aspect is that 4K arrived right in the middle of this generation so the reasoning was sound, and I don't see 8K coming for a while, as it's going to take some time to transition to 4K.

 

[DieH@rd]

SK Hynix plans to launch denser capacities later on with up to 16 Gb or 2 GB chips in the second half of 2018.

That's good news. 32GB nextgen console now more real [I still don't believe that anyone will give the cash for HBM].

One midgen console upgrade is enough. Next HW from both MS and Sony should be new platforms that devs will have to hit with their new software.

 

[Rodéric]

Weren't they supposed to abandon cycles and only release upgraded consoles ?

 

[iroboto]

Weren't they supposed to abandon cycles and only release upgraded consoles ?

I don't think they neither confirmed or denied this point. I doubt they are willing to commit to schedule, might be better to play each hardware generation by ear.
MS committed to the idea of making sure generational-less games continued forward. So Destiny 2 (and it's expansions) will continue to be playable all the way through XBO to XBOX 4 or something, that's where a lot of the confusion was and a lot of us took that concept as no cycles, more updated consoles.

I think what they failed to say, was really that your games will work from XBO forward forever. Which doesn't necessarily imply that XBO will be dead when gen 2 comes out, but that doesn't mean XBO need to be the base spec of the games.

I do see a future where XBO and XBO 4 players play together, and I see exclusives for XBO 2, XBO 3, XBO 4, but developers can target games as far back as XBO if they wanted to for instance. Indies for instance would benefit greatly from having this larger cross generational install base.

 

[Prophecy2k]

I almost feel the minimal requirement is 7nm. If they attempt it at 14/16nm I would be pretty confident that we'd be looking at another mid gen refresh. And I don't necessarily know if mid-gen refreshes are going to be common place or not. Part of me says 'yea', but the other aspect is that 4K arrived right in the middle of this generation so the reasoning was sound, and I don't see 8K coming for a while, as it's going to take some time to transition to 4K.

Agreed.

That's good news. 32GB nextgen console now more real [I still don't believe that anyone will give the cash for HBM].

One midgen console upgrade is enough. Next HW from both MS and Sony should be new platforms that devs will have to hit with their new software.

Couldn't agree more.

Weren't they supposed to abandon cycles and only release upgraded consoles ?

The only people who said this were gamers and speculating journos.

Both MS and Sony have neither mentioned nor intimated towards this at all. Tbh, it's pretty obvious why, given how much it flies in the face of what consoles fundamentally are in comparison to other iterative computing platforms like PC and mobile.

I never understood why some folks even want this especially when endless, regularly upgraded specs, will simply mean the highest end hw forever gets held back by legacy hw. To me it serves no-one:

- Those wanting the best hw don't ever get software designed to take full advantage of it
- Those who prefer the existing generation model get shafted by hw upgrades feeling a whole lot less meaningful
- Devs suddenly need to target multiple spec. configs as opposed to a small number of fixed specifications (increased dev time/cost)
- Platform holders have to invest more in platform software dev and maintenance, SDKs, APIs to support more hw configs
- Everybody sees worse software performance relative to a specific hw config. due to PC-esque software abstraction layers
- With increased difficulty in die-shrinking hw configs, with fewer potential future node-shrinks available, less and less room for future improvement means existing performance is sacrificed for very little potential future gain.

Edit:

I think the question MS and Sony will be asking themselves and AMD/Nvidia/GF/TSMC/Samsung etc, is beyond 7nm - 5nm, how does computing performance progress further?

Whatever technology arise between now and the maturity of the 5nm process will dictate this. Whatever they are, they might be so fundamentally incompatible with existing console APU designs that moving to an iterative upgrade model now would be meaningless since ensuring software compatibility for hw beyond the 5nm node would be impractical.

 

[iroboto]

beyond 7nm - 5nm, how does computing performance progress further?

I assume if the diminishing returns become overwhelming, they would likely be a move to adding fixed function hardware to speed up parts of the pipeline.

 

[DieH@rd]

7nm - 5nm, how does computing performance progress further?

As Tim Sweeney have said in 2012

 

[iroboto]

As Tim Sweeney have said in 2012

funny that I forgot about this. This is certainly a possibility.

 

[AlNom]

Funny how the piano blocks the power density bit.

 

[HBRU]

The huge PS4 installed base (Ps4 that looks also quite a robust, long lasting, hardware) will not allow PS5 to arrive in near o middle future SORRY and we will have another PS4 upgrade (10nm or 7 nm)... a TRUE 4K PS4.... 6 to 8 TFlops). I belive in an entry level PS4-PRO-SLIM and another expanded & expensive PS4-PRO-ULTRA (with 16 gbytes GDDR6... something similar to Scorpio as cooling.... maybe uses same silicon as PS4-PRO but overclocked).

 

[Prophecy2k]

As Tim Sweeney have said in 2012

I wonder, has the cooling issue for 3D stacked ASICs been solved?

I also worry whether 3D stacking will ever be cost effective enough to take off. You're inherently using more silicon and the manufacturing process will always be significantly more complex. 3D stacking for ASICs may end up as one of those technologies that is always too expensive to achieve mass market adoption and thus enjoy economies of scale to drive down prices.

On the other hand, maybe mobile will be the catalyst, especially since those devices inherently carry significantly larger price premiums for the end product. Still, for mobile, I'd expect the cooling problem to be even more pronounced, as mobile devices have to rely on passive cooling; which I can't see having remotely sufficient capability in terms of dissipative heat flux to get all of that heat off the chip.

The huge PS4 installed base (Ps4 that looks also quite a robust, long lasting, hardware) will not allow PS5 to arrive in near o middle future SORRY and we will have another PS4 upgrade (10nm or 7 nm)... a TRUE 4K PS4.... 6 to 8 TFlops). I belive in an entry level PS4-PRO-SLIM and another expanded & expensive PS4-PRO-ULTRA (with 16 gbytes GDDR6... something similar to Scorpio as cooling.... maybe uses same silicon as PS4-PRO but overclocked).

Die shrinks aren't free, HBRU. It's questionable whether Sony will even be able to cost-effectively redesign the PS4 APU for the 7nm node (nobody is looking at 10nm at the moment, I believe). I doubt we'll see a 7nm PS4, and if we do, it'll almost certainly be a design targeted at cost reduction, to bring down the retail price of the PS4 in order to access the wider market. A $100-$150 PS4 would be a very attractive proposition in the wider global markets which are more price-sensitive.

A PS4 Pro-'Pro' would render the original PS4 Pro as completely redundant, especially since that console's main selling point is to play your games at 4k. Not only that, it would also devalue the inevitable PS5, since a 12-14TFLOPS PS5 won't sound anywhere near as impressive anymore next to a hypothetical 8-10 TFLOPS PS4 Pro-'Pro'.

It's easy to see why Sony isn't going to put out yet another upgraded PS4 revision with higher specs that the Pro.

 

[London Geezer]

A PS4 PRO2 would be a disastrous decision.

Even releasing a PS4 Pro with the same specs but a UHD drive later on would annoy a hell of a lot of people.

Pretty sure Sony knows better than that.

 

[HBRU]

The worst decision Sony can make is take a different way from MS backward compatibility.... that would even be worst for Sony having a huge installed base. I know Cerny sayd he belives in the console generations... but probably PS4 owners not so much.

 

[Tkumpathenurple]

The PS2 was a different generation to the PS1, yet was still able to play every PS1 game. Same with the 60GB PS3 and PS2, and similar with the XBoxOne and X360. Therefore, there can be generational breaks that maintain backwards compatibility.

So, I agree that backwards compatibility is a must, and I'm glad Microsoft are pushing it because it will hopefully force Sony to follow suit. That said, gamers and Sony need a backwards compatible PS5, not yet another PS4, slim revisions excluded - a 7nm, cheap PS4Pro Slim would be a worthwhile release, certainly if the PS5 will be backwards compatible and they wish to draw people from around the world into the PlayStation ecosystem.

Personally, I'd like to see the PS5 launch with two tiers. Both would need to come with a UHD Blu-ray drive, so all PS5 games may have scope beyond the 50GB limitation of Blu-ray. Beyond that, the lower end PS5 would need to be significantly more powerful than the Scorpio (not vastly, but visibly so) but still cheap enough to manufacture that it could launch at $300-$350 and still turn a slight profit with wiggle room to cut the price when competition demands.

The higher end PS5 is where my wish may become entirely unfeasible. Hopefully not. I think it would be a great move if the higher end console was essentially two lower end consoles strapped together: two APU's and two pools of memory that can be utilised as one. It would mean that any multiplayer game could be split screen, two different games could be played simultaneously (or an app and a game,) or a single game could be played with extra bells and whistles. Two HDMI ports and the capability for Remote Play would make it a multiple person gaming hub that can dial everything up for a single player.

It would certainly be more expensive, but I think it would be mightily appealing to gaming couples, families, and roommates. Does anyone else agree, or are there technical/economic reasons why this wouldn't be a good idea?

 

[London Geezer]

I'm not sure how I feel about a theoretical PS5 launching alongside a PS5 Pro, and having a two tier situation from day 1.
Doesn't sound too hot, personally. Then again if it happened, I'd of course go for the Pro anyway so who cares.

The problem is that games would still be coded for the minimum denominator, which would be the base PS5.
Pro support would need to be much stronger and stable than what we get today, which is a bit all over the place in my opinion. Some games have zero support, some others have token features, practically invisible compared to the base PS4, while others go full on.