Support NeoGAF

[THE:MILKMAN]

As has been said fall 2019 looks tricky, but could spring 2020 be pulled off?

I guess anything is possible and nothing can be ruled out right now. What is confusing me is if the consensus leads to 2020 being the most likely launch date then why is Matt able to give hints so far out?

I would think there would only have been initial discussions between say Cerny's team and those select dev teams (16 internal, 16 third party for PS4 I believe) at this point i.e. general specs and feedback about what devs would like.

I may be reading his posts very wrong but Matt gives me the impression PS5 is further along than that?

 

[Negotiator]

I can't remember who, but someone did the math, If the PS5 uses true EUV 7nm chips @ about the same size as the Xbox one X Apu and clocked higher would be a little over 19TF.

That's insane if true. I'll gladly wait until 2021-2022 for a true generational leap (10x, 7nm+/5nm).

 

[TheThreadsThatBindUs]

It wouldn't be that simple. For one it assumes the best case scenario for 7nm when there is nearly always a delay (Even Intel have delayed their recent node shrinks) Sony can't one day in late 2019 flick a start PS5 production switch when Apple finally "allow" others to take some 7nm wafers.

For a late 2019 launch they would need to start up production in mid 2019 to ramp up and produce a couple of million units or more for worldwide launch and all the distribution logistics that come with this.

If Sony go with GF or Samsung then all articles are saying these will be coming later than TSMC for 7nm so in effect about the same time for availability without Apple taking the first wafer runs.

Actually is it likely that Sony would want to release PS5 on a brand new node just released? Have they done such before? PS3 and PS4 both launched on pretty mature nodes, right?

I'm pretty sure that TSMC, GloFo and Samsung have separate production lines for their low power 7nm mobile products and 7nm high performance computing devices.

You keep talking about Apple getting first dibs on volume production, and yet Apple's first movers on the newest production process (so in this case 7nm) will be their mobile SoCs on the low power 7nm production lines.

Neither Sony nor Ms will be competing with Apple for capacity on any of the big three contract fabs, they'll be competing with other relatively higher margin, relatively lower volume parts going into things like servers and enterprise computing domains. As soon as the 7nm high performance lines are ready to begin ramping up in capacity, the ink would have already dried on their deals with the console manufacturers who will be offering them guaranteed volumes at reasonable margins for the next 3-5 yrs.

Like I said before, availability on the 7nm is not the critical path for a 2019 launch for either Sony nor MS.

 

[Shin]

For a late 2019 launch they would need to start up production in mid 2019 to ramp up and produce a couple of million units or more for worldwide launch and all the distribution logistics that come with this.

If Sony go with GF or Samsung then all articles are saying these will be coming later than TSMC for 7nm so in effect about the same time for availability without Apple taking the first wafer runs.

I need to stop you there, AMD has an agreement with GloFo that lasts till 2020 with some clause in it.
They "AMD" invested in the 7nm node with GloFo which will give them priority to make use of the new node when it's available.
It now makes sense as to why GlobalFoundries press release read the way it does with them taking customer designs already.
AMD is going to use GloFo for Zen2 which is due next year and on 7nm (as per their roadmap) and that's very well on track.

I guess anything is possible and nothing can be ruled out right now. What is confusing me is if the consensus leads to 2020 being the most likely launch date then why is Matt able to give hints so far out?

It's not surprising if they have a target year, I'm fully expecting 2020 for all the right reasons, node 7nm/7nm EUV, Navi, Zen3, they tend to dictate when a new gen starts, PS4 selling well.

I should note that there's a AMD Conference Call on the 25th of July, 2Q earnings, I'm sure something about their semi-custom business will be mentioned.
Which seems to be booming btw since Apple is making use of AMD's semi-custom designs for 2018 and also 2019 IIRC and we got Hot Chips next month as well.

 

[THE:MILKMAN]

I'm pretty sure that TSMC, GloFo and Samsung have separate production lines for their low power 7nm mobile products and 7nm high performance computing devices.

You keep talking about Apple getting first dibs on volume production, and yet Apple's first movers on the newest production process (so in this case 7nm) will be their mobile SoCs on the low power 7nm production lines.

Neither Sony nor Ms will be competing with Apple for capacity on any of the big three contract fabs, they'll be competing with other relatively higher margin, relatively lower volume parts going into things like servers and enterprise computing domains. As soon as the 7nm high performance lines are ready to begin ramping up in capacity, the ink would have already dried on their deals with the console manufacturers who will be offering them guaranteed volumes at reasonable margins for the next 3-5 yrs.

Like I said before, availability on the 7nm is not the critical path for a 2019 launch for either Sony nor MS.

I didn't know they had separate lines to be honest but the issue I read was that Apple took all the wafers anyway (50,000/month at the start articles say).

I'm intrigued by your last line, though. Do you think that PS5 could launch on existing tech i.e 16nm?

I need to stop you there, AMD has an agreement with GloFo that lasts till 2020 with some clause in it.
They "AMD" invested in the 7nm node with GloFo which will give them priority to make use of the new node when it's available.
It now makes sense as to why GlobalFoundries press release read the way it does with them taking customer designs already.
AMD is going to use GloFo for Zen2 which is due next year and on 7nm (as per their roadmap) and that's very well on track.

Forgive me, Shin. I'm not getting the point or question you are asking here.....!?

 

[Shin]

What I meant is Apple that won't eat up eat up all the Fabs and the availability of 7nm, at least not GlobalFoundries as that will go to AMD.
Not so strange since AMD used to own GloFo until they sold off the rest of their stake (last year?), they got strong ties still though.
We could say that it is because AMD invested in 7nm through GloFo that they seem to have gotten shit together this time, AMD needs that node or it will mess with their financials (Zen2 launch).

That's insane if true. I'll gladly wait until 2021-2022 for a true generational leap (10x, 7nm+/5nm).

I suppose that holds true as the lithography technology adds to the equation, though they are still having difficulties with it so who knows.

 

[THE:MILKMAN]

What I meant is Apple that won't eat up eat up all the Fabs and the availability of 7nm, at least not GlobalFoundries as that will go to AMD.
Not so strange since AMD used to own GloFo until they sold off the rest of their stake (last year?), they got strong ties still though.
We could say that it is because AMD invested in 7nm through GloFo that they seem to have gotten shit together this time, AMD needs that node or it will mess with their financials (Zen2 launch).

Oh sure...My mentioning of GF at all was strictly about availability of 7nm and all articles I've read have TSMC as clear favourites to have 7nm products in consumer hands. I think it is correct that Sony and Microsoft have only used TSMC at launch for PS4/Pro and Xbone/S but they could change for sure if GF do come through on time but history is against them from I have read.

All these foundries say they are on track/ahead of schedule and its still nearly always 18-24 months before products are actually in consumer hands for any given (full at least) node!

 

[TheThreadsThatBindUs]

I didn't know they had separate lines to be honest but the issue I read was that Apple took all the wafers anyway (50,000/month at the start articles say).

I'm intrigued by your last line, though. Do you think that PS5 could launch on existing tech i.e 16nm?

Nope... 7nm certainly.

Just that manufacturing capacity in time for the target launch won't be an issue; at least not for a late 2019 launch at the earliest.

 

[watomsk]

Has Patcher ever been right?

 

[Shin]

I wonder if they'll stick with the current font since they changed it during PS3 and using it on PS4.
They should totally bring back the PS2 font, though we probably won't see consoles with such big imprint anymore.

 

[Bitch Pudding]

Has Patcher ever been right?

He predicted Nintendo Switch to launch in 2017...

 

[Necron]

I wonder if they'll stick with the current font since they changed it during PS3 and using it on PS4.
They should totally bring back the PS2 font, though we probably won't see consoles with such big imprint anymore.

I agree. It would be a good idea too - just turn the "2" around. I love that font.

 

[THE:MILKMAN]

Nope... 7nm certainly.

Just that manufacturing capacity in time for the target launch won't be an issue; at least not for a late 2019 launch at the earliest.

You sound really confident about this! Me not so much having read what happened with the progress and timeline of 16nm. A fair comparison would be TSMC's 16nm FF+ node as used for the GTX 1080 which is a bit smaller than what PS5's APU is likely to be.

Risk production ~November 2014 - GTX 1080 launch = May 27th 2016 or 18 months.

http://www.androidauthority.com/tsmc-16nm-finfet-and-ic-mems-567189/

If PS5 is going to be on the 7nm+ (EUV?) node then I'd be really surprised if any foundry could supply the 100's of thousands of chips a month a console requires right from the get go?

And that is if all the stars align and they come in on time for once!

 

[Souldestroyer Reborn]

Retail: $449
Total BOM Cost: $499
Manufacturing Cost: $9.00
Shipping Cost: $9.00
BOM + Manufacturing + Shipping: $518

CPU/GPU: $190.00 Navi/Zen3 APU
RAM: $100.00 - 16 GB GDDR6
Additional RAM: $20.00 4GB DDR4L
Power Supply: $20.00
Optical Drive: $33.00 UHD
Hard Drive: $37.00
Mechanical / Electro-Mechanical: $35.00
Other (Electronic Content): $40.00

Console Sub Total: $475
Controller: $18.00
Box Contents: $6.00

I used PS3 and PS4 BOM and adjusted it accordingly to expected specs and parts and prices.
From that the only thing we don't know is the price of a 4GB DDR4L module and the APU itself.
Smaller memory jump compared to previous generations due to cost, but beefier CPU & GPU.

Pretty much what I expect, although i think it will be dedicated CPU/GPU. Non-APU parts if they can get thermals down with 7nm, which i think they possibly can even with just the node shrink.

You sound really confident about this! Me not so much having read what happened with the progress and timeline of 16nm. A fair comparison would be TSMC's 16nm FF+ node as used for the GTX 1080 which is a bit smaller than what PS5's APU is likely to be.

Risk production ~November 2014 - GTX 1080 launch = May 27th 2016 or 18 months.

http://www.androidauthority.com/tsmc-16nm-finfet-and-ic-mems-567189/

If PS5 is going to be on the 7nm+ (EUV?) node then I'd be really surprised if any foundry could supply the 100's of thousands of chips a month a console requires right from the get go?

And that is if all the stars align and they come in on time for once!

I may be wrong, but I don't believe there will be as big of an issue moving to 7nm as, this time, many of the current tooling is able to be used for the new process.

 

[Averon]

Shouldn't the PS5 have a shorter and easier R&D cycle given that Sony will just iterate off what they did with the PS4? They do not have to start mostly from scratch like they did going from the PS3 to PS4.

 

[TheThreadsThatBindUs]

You sound really confident about this! Me not so much having read what happened with the progress and timeline of 16nm. A fair comparison would be TSMC's 16nm FF+ node as used for the GTX 1080 which is a bit smaller than what PS5's APU is likely to be.

Risk production ~November 2014 - GTX 1080 launch = May 27th 2016 or 18 months.

http://www.androidauthority.com/tsmc-16nm-finfet-and-ic-mems-567189/

If PS5 is going to be on the 7nm+ (EUV?) node then I'd be really surprised if any foundry could supply the 100's of thousands of chips a month a console requires right from the get go?

And that is if all the stars align and they come in on time for once!

The fatal flaw in your logic is that you're comparing a console APU with a high end desktop GPU binned part.

Read up on chip binning and you'll understand why you can't really infer much about the timeframe from a process going into risk production to full volume production on a console chip, by looking at the launch date of a single binned desktop GPU (one of the highest end GPUs at that). You'd have better luck looking at the launch dates of the lower end chips of the same microarchitecture, which would normally launch earlier than the full chip, due to the time it takes to get yields up high enough to produce chips in sufficient volumes.

Even then the timeframes will only give you a general idea, as there are different considerations in manufacturing desktop parts to console parts (namely business considerations).

 

[heathen]

A lot of people seem convinced that Sony would wait for 7nm+ / EUV, but I don't see why they would. They're more likely to use 7nm, launch a year earlier, and save what marginal future gains can be had from FinFET for slim/pro revisions (or for a PS6 if things look really dire).

 

[Justinh]

Has Patcher ever been right?

Sometimes...

http://kotaku.com/revisiting-michael-pachters-2013-predictions-for-2016-1790520436

 

[TheThreadsThatBindUs]

A lot of people seem convinced that Sony would wait for 7nm+ / EUV, but I don't see why they would. They're more likely to use 7nm, launch a year earlier, and save what marginal future gains can be had from FinFET for slim/pro revisions (or for a PS6 if things look really dire).

I agree with you.

 

[THE:MILKMAN]

The fatal flaw in your logic is that you're comparing a console APU with a high end desktop GPU binned part.

Read up on chip binning and you'll understand why you can't really infer much about the timeframe from a process going into risk production to full volume production on a console chip, by looking at the launch date of a single binned desktop GPU (one of the highest end GPUs at that). You'd have better luck looking at the launch dates of the lower end chips of the same microarchitecture, which would normally launch earlier than the full chip, due to the time it takes to get yields up high enough to produce chips in sufficient volumes.

Even then the timeframes will only give you a general idea, as there are different considerations in manufacturing desktop parts to console parts (namely business considerations).

Hey I bow to you or anyone with better insight. I'm just reading a tonne of articles about the nodes, processes and the timeline. As for the Nvidia GPUs they released in power order GTX 1080=May 2016 through to GTX 1050=October 2016.

 

[Shin]

Pretty much what I expect, although i think it will be dedicated CPU/GPU.

This is more accurate and cross referenced with XB1S BoM:

Retail: $449
Total BOM Cost: $505
Manufacturing Cost: $9.00
Shipping Cost: $9.00
BOM + Manufacturing + Shipping: $523

CPU/GPU: $189.00 Navi/Zen3 APU
RAM: $100.00 - 16 GB GDDR6
Additional RAM: $15.00 4GB DDR4L
Power Supply: $16.00 Internal
Optical Drive: $33.00 UHD
Hard Drive: $43.00 2TB Seagate
Mechanical / Electro-Mechanical: $35.00
Other (Electronic Content): $50.00 Vapor chamber

Console Sub Total: $481
Controller: $18.00
Box Contents: $6.00

A lot of people seem convinced that Sony would wait for 7nm+ / EUV, but I don't see why they would. They're more likely to use 7nm, launch a year earlier, and save what marginal future gains can be had from FinFET for slim/pro revisions (or for a PS6 if things look really dire).

I think this is correct since EUV brings it's own set of problems and products listed to use that are scheduled for 2020, so I don't see this being an option.
The risk is too big and the window is too small for them to make use of it, 7nm will do just fine they'll probably use 5nm for revision/Pro which should be mature by then.

 

[onQ123]

I think it's time to add co-processors & accelerators.

The engineers already know that devs will be anchored by the Xbox One for awhile because of the One Microsoft model that MS is going for so they should be able to build the hardware around the software.


Let a ray-tracing co-processor handle the lighting , shadows & so on & the GPU & CPU wouldn't even need to be so powerful.

 

[Welfare]

This is more accurate and cross referenced with XB1S BoM:

Retail: $449
Total BOM Cost: $505
Manufacturing Cost: $9.00
Shipping Cost: $9.00
BOM + Manufacturing + Shipping: $523

CPU/GPU: $189.00 Navi/Zen3 APU
RAM: $100.00 - 16 GB GDDR6
Additional RAM: $15.00 4GB DDR4L
Power Supply: $16.00 Internal
Optical Drive: $33.00 UHD
Hard Drive: $43.00 2TB Seagate
Mechanical / Electro-Mechanical: $35.00
Other (Electronic Content): $50.00 Vapor chamber

Console Sub Total: $481
Controller: $18.00
Box Contents: $6.00

Sony won't eat that big of a loss on hardware, so either that becomes a $499 device or the specs take a hit for $399. $449 would be reserved for a larger HDD SKU to a $399 launch, it won't be the actual price of the base PS5.

 

[Shin]

Sony won't eat that big of a loss on hardware, so either that becomes a $499 device or the specs take a hit for $399. $449 would be reserved for a larger HDD SKU to a $399 launch, it won't be the actual price of the base PS5.

I think they both are eating $50 loss as is on base models, parts become cheaper overtime but I don't think $150 - 200 cheaper in just 3 years.
If no vapor chamber ($20-), 1TB HDD ($15-), external PSU ($6+), the rest is fixed so I don't expect any changes there ($35 saved).
The CPU is about $50 of the APU price, they could settle for less graphical power that would lower the APU cost ($189 seems way too high IMO).
I would have to look into their financials to see if they are tucking away PS4 hardware loss under Network Services which I think is the case.

The component prices I listed gives us this:

3.2GHz 8 Core Zen3
16GB GDDR6
4GB DDR4L BG/OS
2TB SSHD
12TF GPU

Once IHS releases their research on XBOX BoM we'll have a clearer image, Matt did say that $499 might be a possibility (opinion?).
I don't know about you guys but $499 justifies such a console, memory isn't what people would want, but that seems to be a balanced system.
Kinda crazy when you think about it, be it a 9, 10 or even 12TF machine, they are getting it for only $110 - 190, that's seriously cheap AF.

 

[Souldestroyer Reborn]

Sony won't eat that big of a loss on hardware, so either that becomes a $499 device or the specs take a hit for $399. $449 would be reserved for a larger HDD SKU to a $399 launch, it won't be the actual price of the base PS5.

I believe that at launch Sony was taking around a $50 loss per console.

I'm sure there was an article stating that a game and a year of ps+ made them even/a profit.

As for above, thanks for the updated BoM.

Personally I'd be happy with a $500 console, IF, it offered a substantial improvement, i.e. VRR, 4K, UHD playback.

For 6/7 years worth of entertainment I can't complain about that.

 

[Angelus Errare]

Pretty much what I expect, although i think it will be dedicated CPU/GPU. Non-APU parts if they can get thermals down with 7nm, which i think they possibly can even with just the node shrink.



I may be wrong, but I don't believe there will be as big of an issue moving to 7nm as, this time, many of the current tooling is able to be used for the new process.

Def not happening

 

[Renekton]

A lot of people seem convinced that Sony would wait for 7nm+ / EUV, but I don't see why they would.

Without EUV the design cost would be too damn high.

 

[Welfare]

I believe that at launch Sony was taking around a $50 loss per console.

I'm sure there was an article stating that a game and a year of ps+ made them even/a profit.

As for above, thanks for the updated BoM.

Personally I'd be happy with a $500 console, IF, it offered a substantial improvement, i.e. VRR, 4K, UHD playback.

For 6/7 years worth of entertainment I can't complain about that.

The hardware itself was not that big a loss. Sony reported PS4 hardware was profitable 7 months after launch.

If Sony are making hardware that costs over $500 to produce and ship, it will be $499. Loss leading consoles aren't a thing anymore where Sony and MS are willing to lose so much money per unit. They will be as cheap as possible to produce and the MSRP will be close to the total manufacture price.

 

[Shin]

I believe that at launch Sony was taking around a $50 loss per console.

I'm sure there was an article stating that a game and a year of ps+ made them even/a profit.

As for above, thanks for the updated BoM.

Personally I'd be happy with a $500 console, IF, it offered a substantial improvement, i.e. VRR, 4K, UHD playback.

For 6/7 years worth of entertainment I can't complain about that.

Me neither and I forgot to mention that they can save $15 by excluding a UHD drive (though I rather they not I'm banking on PS5 having one).
Because I need a Region 1 and Region 2 UHD player, XBOX will be my Region 1 (along with 4 games I plan on buying for it).
I think will have a UHD drive though mainly because the need of larger capacity (BDXL / 100GB), compression can only do so much.

More important than specs, as far as PlayStation is probably concerned is growing PS+ subscription, a higher price point on the SKU won't help with that.
Backward compatibility is already a step towards that move and Sony isn't the platform holder that tends to push specs, Ken Kutaragi has long left Sony.

 

[Marlenus]

Sony did really well with a $399 price point so unless XBO-X does well at the $499 price point I do not see Sony targeting anything higher.

AMD will be launching their Zen + Vega APUs by the end of the year so using a bigger / faster version of that is the same as what they did with the jaguar APU.

AMD are also due to release ryzen 2 on 7nm in 2018 (iirc) so even if there is some delay 7nm should be available by end of 2019 and deffo by 2020.

 

[Bitch Pudding]

Sony did really well with a $399 price point so unless XBO-X does well at the $499 price point I do not see Sony targeting anything higher.
.

Sony's PS4 did well in an environment where their main competitor launched a weaker console for $499 at basically the same time.

Having a one year head start over the competition allows other price strategies. E.g. launching a $499 console which can be offered cheaper ($399) once the competition enters the playing field.

 

[mrklaw]

I think it's time to add co-processors & accelerators.

The engineers already know that devs will be anchored by the Xbox One for awhile because of the One Microsoft model that MS is going for so they should be able to build the hardware around the software.


Let a ray-tracing co-processor handle the lighting , shadows & so on & the GPU & CPU wouldn't even need to be so powerful.

The GPU is already a co processor with compute allowing the CPU to be less powerful.

I wonder if anyone will go for discrete CPU+GPU to give them more die space to work on

 

[Marlenus]

Sony's PS4 did well in an environment where their main competitor launched a weaker console for $499 at basically the same time.

Having a one year head start over the competition allows other price strategies. E.g. launching a $499 console which can be offered cheaper ($399) once the competition enters the playing field.

I don't see what 1 year of advancement between 2019 and 2020 will bring to a competitor. Maybe Zen 3 instead of Zen 2 but GPU wise it will still be on 7nm so die size will be a limiting factor so unless one company is prepared to make a larger APU performance will be roughly similar.

2021 has the slim chance of being on the next node but if Sony launch in 2019 I can't see MS giving them a 2 he headstart.

You also need to look at real wage growth and can people justify paying more than $399? I think Xbo-x will be a great barometer of this but being a mid life refresh it is not apples to apples.

The GPU is already a co processor with compute allowing the CPU to be less powerful.

I wonder if anyone will go for discrete CPU+GPU to give them more die space to work on

I can see it being an MCM connected via Infinity Fabric. Sony / MS could essentially choose an off the shelf Zen die and an off the shelf Navi die and 'glue' them together. Depends if that is cheaper than making a custom monolithic core.

 

[anothertech]

The component prices I listed gives us this:

3.2GHz 8 Core Zen3
16GB GDDR6
4GB DDR4L BG/OS
2TB SSHD
12TF GPU

I don't know about you guys but $499 justifies such a console, memory isn't what people would want, but that seems to be a balanced system.
Kinda crazy when you think about it, be it a 9, 10 or even 12TF machine, they are getting it for only $110 - 190, that's seriously cheap AF.

Agree. And interesting, this is almost my exact prediction for s $499 console at the end of 2019. Only difference is I think HBM will have made enough headway by that time it may be used in place of gddr6 for that extra bandwidth.

There's a few technologies coming in the near future not yet fully revealed that are going to need more bandwidth than gddr can give us at that time. I'm betting Sony is going to have ps5 fully ready and capable for those tech in the next 2.5 years.

 

[Maroon Gamer]

I wonder if they'll stick with the current font since they changed it during PS3 and using it on PS4.
They should totally bring back the PS2 font, though we probably won't see consoles with such big imprint anymore.

Best. Font. Ever.

 

[Makoto-Yuki]

Shouldn't the PS5 have a shorter and easier R&D cycle given that Sony will just iterate off what they did with the PS4? They do not have to start mostly from scratch like they did going from the PS3 to PS4.

i would think so. they could just make what is essentially a "PS4 Pro+" but call it "PS5" instead and let it have its own exclusive games. That way we get PS4 Backwards Compatibility and next generation level of graphics/performance.

PS5 will use the same architecture as PS4. PS3 used Cell so like you say they will have created PS4 from scratch. PS4 is a good base to build on for the PS5.

 

[Marlenus]

Agree. And interesting, this is almost my exact prediction for s $499 console at the end of 2019. Only difference is I think HBM will have made enough headway by that time it may be used in place of gddr6 for that extra bandwidth.

There's a few technologies coming in the near future not yet fully revealed that are going to need more bandwidth than gddr can give us at that time. I'm betting Sony is going to have ps5 fully ready and capable for those tech in the next 2.5 years.

I think $499 for those specs is too much. It is essentially double what we have now (after XBO-X) in 2 years time for the same money. I also think it will have at a minimum a unified pool of 32GB HBM.

If you think about it in 2012 we had the 7xxx series launch and a $399 2013 console used the cut down 2nd tier GPU. The Navi roadmap suggests a 2018 release meaning it's possible a 2019 console could use a cut down 2nd tier Navi GPU (assuming full product stack) and come in at $399.

The 7850 also had 2GB of ram and PS4 ended up with 8GB, I would expect a 2nd tier Navi to have 8GB of ram so 4x that is 32GB. OTOH PS2 - PS3 and PS3 - PS4 both had 16x memory increases which would indicate 128GB. Might be overkill and if it was that much it would not be HBM.

I think it depends how well XBO-X sells as to whether Sony opts for $399, $449 or $499 as their target price point and the price point will obviously pact the overall spec of the console.

 

[Renekton]

XoX is not a reliable indicator for pricing, it's a refresh that shares the momentum and issues of the X1 platform since 2013.

 

[THE:MILKMAN]

I think $499 for those specs is too much. It is essentially double what we have now (after XBO-X) in 2 years time for the same money. I also think it will have at a minimum a unified pool of 32GB HBM.

If you think about it in 2012 we had the 7xxx series launch and a $399 2013 console used the cut down 2nd tier GPU. The Navi roadmap suggests a 2018 release meaning it's possible a 2019 console could use a cut down 2nd tier Navi GPU (assuming full product stack) and come in at $399.

The 7850 also had 2GB of ram and PS4 ended up with 8GB, I would expect a 2nd tier Navi to have 8GB of ram so 4x that is 32GB. OTOH PS2 - PS3 and PS3 - PS4 both had 16x memory increases which would indicate 128GB. Might be overkill and if it was that much it would not be HBM.

I think it depends how well XBO-X sells as to whether Sony opts for $399, $449 or $499 as their target price point and the price point will obviously pact the overall spec of the console.

Matt has already told others 32GB or more RAM isn't happening and shouldn't be expected. Also devs won't need that much even though more is always nice. The main reasons PS4 ended up with 8GB was because double density chips came out in time and the use of clamshell mode.

Even if devs told Sony they'd be dead sticking with 4GB my guess is the main reason would be that Durango always had 8GB and a 8 and 4GB console split would have caused all sorts of problems for devs. The same happened with the 360 IIRC.

My PS5 spec guess as of today assuming 2019 or later launch on 7nm:

10-12TF and 8 core custom Ryzen (mobile/cut down) APU
16GB RAM
2TB storage (I'm hoping for some innovation here but just don't know what or how!)
UHD drive (For the capacity more than anything)
$399/£349 in today's money

 

[Marlenus]

XoX is not a reliable indicator for pricing, it's a refresh that shares the momentum and issues of the X1 platform since 2013.

If it sells well relative to PS4 Pro then it means a $499 console is something consumers can stomach, provided the specs make it worthwhile.

I would consider selling well relative to PS4 Pro as around half as many units.

 

[Council Pop]

I mean that makes this gen 6 years.
More than plenty.
Consoles can't go another generation of 7+ years with how much has changed

That's not how it works. Generations are determined by commercial factors, there's no 'time limit'. Console gaming is probably more mainstream than it's ever been, and the PS4 is selling incredibly. There's no reason to push an upgrade in the next few years. I'd say 2020 is the absolute earliest- I'd expect 2021.

 

[Theonik]

XoX is not a reliable indicator for pricing, it's a refresh that shares the momentum and issues of the X1 platform since 2013.

This is true but Sony needs to be convinced that 499 is the correct pricepoint for a system intended for a mainstream market. MS succeeding would help drive that point but any failure of MS even if it's not because of pricing will be used as confirmation internally that Sony is right.

 

[Rolf NB]

I could see them going slightly asymmetric on the CPU again to save die space.

Low single-threaded performance is the biggest pain. People had to jump through lots of hoops (like multi-threaded command buffer building) not because the aggregate CPU performance was too low, but rather because there is no fast "main" CPU thread to rely on for the parts of the code that remain hard to parallelize.

8C/16T Ryzen seems like utter overkill for a console with decent GPU compute facilities. A 2C/4T Ryzen at 3.5GHz+ would already make short work of current PS4 CPU workloads while also reducing pressure to parallelize everything for native PS5 code..

OTOH there's increased value of BC which would lead to considerations that you need at least 8 threads (plus / minus OS reserves) for games fully utilizing the PS4 CPU. Which makes 4C/8T seem like the easy, obvious choice.

But what if it was 2C/4T Zen + any number between 4 and 8 Cats to "help" with the "more mundane" things and to ensure BC? Or even just 1C/2T Zen plus Cats?

 

[Theonik]

There isn't much point in a hybrid design with Zen and cats. You would still consume similar die area but basically be in a CBE nightmare scenario where only some cores can be used for certain things and developers need to both parallelise their code better but deal with these deficiencies. At this point it depends on whether 7nm matures on time or not.

 

[Midas]

Are there any disadvantages with using single chip for GPU and CPU? Except for price and space?

 

[anothertech]

I think $499 for those specs is too much. It is essentially double what we have now (after XBO-X) in 2 years time for the same money. I also think it will have at a minimum a unified pool of 32GB HBM.

If you think about it in 2012 we had the 7xxx series launch and a $399 2013 console used the cut down 2nd tier GPU. The Navi roadmap suggests a 2018 release meaning it's possible a 2019 console could use a cut down 2nd tier Navi GPU (assuming full product stack) and come in at $399.

The 7850 also had 2GB of ram and PS4 ended up with 8GB, I would expect a 2nd tier Navi to have 8GB of ram so 4x that is 32GB. OTOH PS2 - PS3 and PS3 - PS4 both had 16x memory increases which would indicate 128GB. Might be overkill and if it was that much it would not be HBM.

I think it depends how well XBO-X sells as to whether Sony opts for $399, $449 or $499 as their target price point and the price point will obviously pact the overall spec of the console.

Nah, just double the tf gpu Xbonx power, but essentially it would actually be better to compare it as

7x the og ps4 gpu along with die shrink +
20x the CPU power +
HBM2/3 100x the bandwidth +
4x the HDD space/r/rw speed

That's all exaggerated but Easily worth that extra $100 over Xbonx. Not to mention $499 in holiday 2019 won't look as bad as $499 in 2013.

 

[Theonik]

Are there any disadvantages with using single chip for GPU and CPU? Except for price and space?

The main disadvantage is size and heat are limited by what you can fit on a single chip economically vs two chips. Of course if you start with two chips you can always save money down the line by integrating them into a single chip like what Sony did with PS2.

Now, there are advantages to an APU as well, under most circumstances it is much cheaper provided you don't want to push the size too hard. (a big APU might cost more than two smaller chips) You also get much lower latency between components than you can on a bus, though a bus between chips also costs money to implement in terms of the motherboard.

 

[Averon]

i would think so. they could just make what is essentially a "PS4 Pro+" but call it "PS5" instead and let it have its own exclusive games. That way we get PS4 Backwards Compatibility and next generation level of graphics/performance.

PS5 will use the same architecture as PS4. PS3 used Cell so like you say they will have created PS4 from scratch. PS4 is a good base to build on for the PS5.

Yeah, that is what I thought. The PS4 had a 5 year development cycle, with Sony going from Cell, Nvidia GPU tech, and split RAM config to AMD tech with unified memory config. I could Sony starting PS5 development in earnest this year and be ready for launch by 2020, a 3 year development cycle. If Sony gambles on HBM or wait for Zen 3, I can see another year added to R&D, making it a 4 year development cycle. Still shorter than the PS4's.

 

[Renekton]

I could see them going slightly asymmetric on the CPU again to save die space.

But what if it was 2C/4T Zen + any number between 4 and 8 Cats to "help" with the "more mundane" things and to ensure BC? Or even just 1C/2T Zen plus Cats?

Not feasible given that Zen is not designed like ARM's BIG.little.

Even if it is somehow possible, it's an awkward Frankenstein solution.

 

[Theonik]

Big.Little only really makes sense in mobile where it is specifically to lower idle power consumption. On a normal desktop APU that's not a concern and it's just a waste of diespace. If some background tasks need to be offloaded for that reason, you are better off doing what Sony did with PS4 and have a dedicated ARM CPU for idle, though that's still expensive and didn't work out that well on PS4 on hindsight.