Support NeoGAF

[RedSwirl]

Technically PS5 could end up the only traditional next gen console but that's only if Sony actually make a clean break into next generation. I'm not really sure if they will because if things workout for MS Xbox will basically be a pre-made multimedia gaming computer so when Xbox One Z or whatever they call it comes out it wouldn't really matter if it only sell 1 million units because the games for it is going to be made either way because they are UWP games that will run on consoles & Windows PCs & that would leave PS5 on it's own when it come to trying to start a new generation & it would be really hard to get devs to make a PS5 exclusive in the 1st year when they can just make a game that run on everything from the Xbox One/PS4 on up. If MS is successful with the generation-less console model when PS5 come out it will look like Sony is asking devs to make games twice while MS is letting them make the game once.

This is why I think PS5 will only be PS5 by name if it comes out in the next few years putting a pretty coat of paint over PS4 games & if they are planing to make a clean break into next generation they will be better off waiting as long as they can until people are bored with the PS4/Xbox One games & make a radical change with the PS5 that separate it from the PS4 , Xbox One & PC.

I definitely think Microsoft is gonna end up establishing some sort of precedent in contrast to Sony when it comes to backwards compatibility and the relationship between generations.

If Sony still enforces distinct divisions between PS5 and PS4 games or, God forbid, the PS5 isn't even backwards compatible with PS4, I think you're actually STILL going to see developers basing PS4 and PS5 versions of games off the same code base. I think no matter what at this point the PS5 is going to be technically a PS4 with a new CPU and all the other specs jacked up, there's no reason for Sony not to do that. It's just a matter of whether Sony (or Cerny) thinks certain optimizations and expectations about how some developers will code may make it safer comparability-wise to put a "wall" between PS4 and PS5. There are going to be cross-gen PS4 and PS5 games, just like there were cross-gen PS3 and PS4 games. The difference is instead of being developed for two totally different architectures they'll be developed for the same architecture but at different levels of power, which means you're potentially looking at the PS4 and PS5 versions of games being about as different code-wise as a game running on PS4 and that same game running on PS4 Pro. Why even sell them as different SKUs at that point?

If Sony sticks with the same architecture (which it probably will) AND has BC with PS4 games, why not just let developers make a "PS5 Mode" to enhance those BC games? At that point you sort of eliminate cross-gen games. They'll probably put "PS4 and PS5" on the box or something.

 

[Bitch Pudding]

I definitely think Microsoft is gonna end up establishing some sort of precedent in contrast to Sony when it comes to backwards compatibility and the relationship between generations.

If Sony still enforces distinct divisions between PS5 and PS4 games or, God for bid, the PS5 isn't even backwards compatible with PS4, I think you're actually STILL going to see developers basing PS4 and PS5 versions of games off the same code base. I think no matter what at this point the PS5 is going to be technically a PS4 with a new CPU and all the other specs jacked up, there's no reason for Sony not to do that. It's just a matter of whether Sony (or Cerny) thinks certain optimizations and expectations about how some developers will code may make it safer comparability-wise to put a "wall" between PS4 and PS5. There are going to be cross-gen PS4 and PS5 games, just like there were cross-gen PS3 and PS4 games. The difference is instead of being developed for two totally different architectures they'll be developed for the same architecture but at different levels of power, which means you're potentially looking at the PS4 and PS5 versions of games being about as different code-wise as a game running on PS4 and that same game running on PS4 Pro. Why even sell them as different SKUs at that point?

If Sony sticks with the same architecture (which it probably will) AND has BC with PS4 games, why not just let developers make a "PS5 Mode" to enhance those BC games? At that point you sort of eliminate cross-gen games. They'll probably put "PS4 and PS5" on the box or something.

Yeah, it will be interesting to see how things turn out in the end. That said, it will be especially interesting to see if devs accept PS5 as a new minimum standard for exclusive next-gen games or if they'll try to run those on XOX, too. If the later happens, XOX would be the lowest common denominator for next gen games. And no offense, but I really hope that doesn't happen. I want next-gen games to be able to leave Jaguar CPU-based games behind for good and fully exploit whatever Sony puts into PS5.

 

[LukasTaves]

$499 seems like a better goal than compromising in areas to fit $399.

Nah, 499 only makes sense, and even then might be a failure, if you already have a better priced alternative.

For instance, a $500 Ps5 when at the time we could be looking at a $200 Pro that would play the same games for a period of time might be okay. A $500 Ps5 on its own would be a tougher sell.

 

[RoboPlato]

Yeah, it will be interesting to see how things turn out in the end. That said, it will be especially interesting to see if devs accept PS5 as a new minimum standard for exclusive next-gen games or if they'll try to run those on XOX, too. If the later happens, XOX would be the lowest common denominator for next gen games. And no offense, but I really hope that doesn't happen. I want next-gen games to be able to leave Jaguar CPU-based games behind for good and fully exploit whatever Sony puts into PS5.

I think it really depends on the developer and the goals they have. The biggest difference between a PS5 and something like the PS4 Pro is that developers will have the option of making games built for the ground up for it. I would hope that if it is similar architecture that cross gen games would be a single disc/download and it would just load up a different settings profile for PS5, like with the Pro. That way you could still get a lot of benefit out of the hardware upgrade and then when a game is too stressful for the current consoles, devs can drop support for the older platforms without having to wait too long for new hardware.

The Xbox One X will not be the lowest common denominator because it isn't allowed to have games that don't run on the base Xbox One and even if it was, something like a 60fps version with better draw distances on PS5 would still be a huge upgrade.

 

[Omnistalgic]

Maybe Sony will release the PS5 and Pro simulatenously this time at $399 and $599 so everyone can get their beefy specs, or go with the standard from the get go and they improve on costs and size factor as the gen goes on and make improvements where they can.

I think PS5 would do crazy numbers again that way, since hardcore won't be waiting for a potential PS5P once the new gen gets under way.

 

[Darklor01]

Maybe Sony will release the PS5 and Pro simulatenously this time at $399 and $599 so everyone can get their beefy specs, or go with the standard from the get go and they improve on costs and size factor as the gen goes on and make improvements where they can.

I think PS5 would do crazy numbers again that way, since hardcore won't be waiting for a potential PS5P once the new gen gets under way.

Don't know if you'll see a Pro version again unless they need the power for something like PSVR 2.0

 

[RoboPlato]

Maybe Sony will release the PS5 and Pro simulatenously this time at $399 and $599 so everyone can get their beefy specs, or go with the standard from the get go and they improve on costs and size factor as the gen goes on and make improvements where they can.

I think PS5 would do crazy numbers again that way, since hardcore won't be waiting for a potential PS5P once the new gen gets under way.

The main reason the Pro happened was because of die shrinks and architectural improvements after the release of the PS4. The size of a Pro at launch would be huge.

 

[Bitch Pudding]

The main reason the Pro happened was because of die shrinks and architectural improvements after the release of the PS4. The size of a Pro at launch would be huge.

Yup, I asume they'll introduce a PS5 Pro once they can offer it for the same price as a OG PS5. I mean, when PS5 arrives, it will be the fastest console anyhow, maybe for one, maybe for two years. No reason to make an even faster one. And of course, a $200 premium is most likely not enough for a significant (2x+) inrease in power towards the cheaper one.

 

[AAMARMO]

Man 2019 is going be a crazy year for Sony.

These pictures are fake and are concept art from here https://www.behance.net/gallery/52981873/Redesign-for-Play-Station-1.

 

[GribbleGrunger]

I miss those days of the PS3 too.

+10 year lifecycle

And the PS1, PS2 and PSP. The Vita and the PS4 will be exactly the same.

 

[sonycowboy]

I miss those days of the PS3 too.

+10 year lifecycle

1) I don't see a 2019 release unless the industry just starts tanking or the X1X just steals Sony's lunch in 2018.

2) Why do folks agree that PS3 had 10+ years?

http://www.ps5playstation5.com/ps5-release-date-countdown

Playstation Release Date History
PS1 Release Date Saturday - December 3, 1994
Time Between 5 Years 3 months – 1918 days
PS2 Release Date Saturday - March 4, 2000
Time Between 6 years and 8 months – 2444 days
PS3 Release Date Saturday - November 11, 2006
Time Between 7 Years, 5 days – 2562 days
PS4 Release Date Friday - Nov 15, 2013 (XBox beat - 7 days)
Time Between 7 years & 1 day - 2558 days total - assuming 2020 release date
PS4 Pro Release Thursday - Nov 10, 2016 (XBox X beat - 362 days)
Time Between 2 years, 11 months, 28 days - 1093 days total

PS5 Release Date November 15, 2020 (give or take)

Clearly, they added one year from the PS2, but if Sony continues to hold market advantage, I don't see them releasing in 2019 and cutting off the software licensing profit prematurely before a new console.

 

[truth411]

Hardcore gamers aren't carrying any console for two years and a $100 price drop isn't happening after one year unless the PS5 flops out of the gate. I'm not sure I've seen a recent successful console launch at $499. Xbox One is still trying to recover and the PS3 spent way too long coming back. I'll eat my words if I'm wrong, but I just don't see a $499 launch. Especially not if the console launches in 2020 or 2021, which I believe is more likely than 18/19.

You really think the PS5 could come out NEXT year? Lol.

 

[shandy706]

Oh well, PS5 cancelled and Sony to exit the business.

We had a good run guys.

Maybe Sega or Atari will jump back in.

 

[THE:MILKMAN]

You really think the PS5 could come out NEXT year? Lol.

He clearly says he favours 2020/21! 2018 has only ever been (confidently) stated by that one Japanese analyst. Saying that he has been quiet ever since...

Me personally am not confident 7nm for high performance 300+mm^2 SoC/APU will be ready to go in mid 2019 in 500k-1 million/month quantities.

Reading the below shows a decent timeline for smartphone SoCs that makes sense to me:

For smartphone SoCs, this likely means that we won’t see many major, if any, 7nm platforms launch until early 2019, as this is when most companies choose to launch their latest high-end smartphone and typically follows new flagship chip announcements from Qualcomm. We are likely to see 10nm FinFET and subsequent revisions be the process of choice for mobile SoCs throughout 2017 and 2018.

http://www.androidauthority.com/lg-q6-vs-samsung-galaxy-a5-2017-786805/

I just can't see PS5s SoC being ready for a holiday 2019 launch but on the other hand I can't see Sony waiting until 2020 either....it is a conundrum for sure. All the talk from the media and Sony execs themselves tells me PS5 is sooner rather than later but the tech just doesn't seem like it will be ready in time.

I wonder if a 10nm "half step" as Pachter calls it is a possibility? Like Cerny said: 8TF more RAM, BW and Ryzen Mobile and launch next year or 2019?

 

[Herp Derpington]

Man 2019 is going be a crazy year for Sony.

These pictures are fake and are concept art from here https://www.behance.net/gallery/52981873/Redesign-for-Play-Station-1.

1 USB port? Seems legit.

 

[Darklor01]

....
I just can't see PS5s SoC being ready for a holiday 2019 launch but on the other hand I can't see Sony waiting until 2020 either....it is a conundrum for sure. ....

I kind of identify with this. Thing for me when I think about it is that it seems to be a damned if you do launch early situation.

On the one hand, those ready to upgrade from a vanilla PS4/PS4 Slim and are willing to pay for it based on whatever price it hits at might rejoice. Those who invested in a PS4 Pro not too long ago might not be as happy.

Another issue might be that if they do produce a powerhouse machine, waiting longer after availability aids in reducing cost to the consumer whereas releasing sooner might result in a physically larger box and a higher cost.
If they try to stick to a lower price point launching earlier, the result might be console without as much power as people are imagining the PS5 should have, and it might cause other issues in regards to sales as people continue to bash the system for being under-powered.

For me, the best option for them is to build a machine which is powerful enough to justify purchase for price, even if a bit under-powered at the time they release it because they waited longer to get a smaller physical version and a lower cost on parts. That's just my opinion though.

 

[Theonik]

1 USB port? Seems legit.

2 more in the front. Looks painful to manufacture though.

 

[ChaosZeroX]

Guess it won't be then.

First post nailed it

 

[EmiPrime]

Man 2019 is going be a crazy year for Sony.

These pictures are fake and are concept art from here https://www.behance.net/gallery/52981873/Redesign-for-Play-Station-1.

Whoever did that render must really like the Wondermega.

 

[Shin]

These consoles take really long to research and develop, though has there been anything groundbreaking this generation?
MS already working on their next console apparently: http://gamingbolt.com/in-theory-wou...flops-gpu-deliver-true-photorealistic-visuals

 

[RedSwirl]

I think it really depends on the developer and the goals they have. The biggest difference between a PS5 and something like the PS4 Pro is that developers will have the option of making games built for the ground up for it. I would hope that if it is similar architecture that cross gen games would be a single disc/download and it would just load up a different settings profile for PS5, like with the Pro. That way you could still get a lot of benefit out of the hardware upgrade and then when a game is too stressful for the current consoles, devs can drop support for the older platforms without having to wait too long for new hardware.

The Xbox One X will not be the lowest common denominator because it isn't allowed to have games that don't run on the base Xbox One and even if it was, something like a 60fps version with better draw distances on PS5 would still be a huge upgrade.

I have a hard time right now imagining what MIcrosoft will do in response to say, a PS5 with a new CPU in 2020. XOX is only three years old at that point. Put out another new model with a CPU upgrade in 2021 and establish it as the new mandatory minimum? Can we even have year-long gaps between competing consoles anymore?

 

[geordiemp]

I definitely think Microsoft is gonna end up establishing some sort of precedent in contrast to Sony when it comes to backwards compatibility and the relationship between generations.

Not really, just sony stuck with Cell last gen so impossible to do BC. However, PR people like to talk gen no gen rubbish. Both next gen X86's will play current gen, we will just get a whole lot of marketing blurb differences.

I wonder if a 10nm "half step" as Pachter calls it is a possibility? Like Cerny said: 8TF more RAM, BW and Ryzen Mobile and launch next year or 2019?

No point, fabs can do 7 nm, the cost of the process is 30 % more steps and more processing costs, the big issue is parasitic RC, so 10 nm is the same hassle factor so its not really worth it. Look up double patterning techniques on why 7 or 10 is similar costs and work.

So 7 nm is a shrink but its costs are higher to make, people getting carried away with moores law, think we will see Ryzen + 8 -12 TF and $ 499 maybe 2019

 

[RCU005]

1) I don't see a 2019 release unless the industry just starts tanking or the X1X just steals Sony's lunch in 2018.

2) Why do folks agree that PS3 had 10+ years?

http://www.ps5playstation5.com/ps5-release-date-countdown

Playstation Release Date History
PS1 Release Date Saturday - December 3, 1994
Time Between 5 Years 3 months – 1918 days
PS2 Release Date Saturday - March 4, 2000
Time Between 6 years and 8 months – 2444 days
PS3 Release Date Saturday - November 11, 2006
Time Between 7 Years, 5 days – 2562 days
PS4 Release Date Friday - Nov 15, 2013 (XBox beat - 7 days)
Time Between 7 years & 1 day - 2558 days total - assuming 2020 release date
PS4 Pro Release Thursday - Nov 10, 2016 (XBox X beat - 362 days)
Time Between 2 years, 11 months, 28 days - 1093 days total

PS5 Release Date November 15, 2020 (give or take)

Clearly, they added one year from the PS2, but if Sony continues to hold market advantage, I don't see them releasing in 2019 and cutting off the software licensing profit prematurely before a new console.

I think that time bewteen consoles will get longer since it takes longer to make games now. It's also more expensive, so I don't see Sony releasing a PS4 in 2019. There are games that have been in development since 2015, and still not done (God of War).

 

[sugarman]

Not really, just sony stuck with Cell last gen so impossible to do. However, PR people like to talk gen no gen rubbish.

And PS1/PS2?

 

[geordiemp]

And PS1/PS2?

Probably no market demand and games are very dated. Yes you get retro guys but I am sure Sony done their stats and costs / benefit. Ps4 BC will be desirable in 2019 though.

However, this gen games at 1080p will have longer staying power IMO.

 

[boredofcanada]

Can't Sony and MS just make a bomb ass 600 dollar console and be done with new versions for at least a decade? We all own PS4's and XB1's. Trade in will reduce the cost when it comes time to upgrade. Why the rat race? What happened to ten year cycles?

 

[Shin]

I think that time bewteen consoles will get longer since it takes longer to make games now. It's also more expensive, so I don't see Sony releasing a PS4 in 2019. There are games that have been in development since 2015, and still not done (God of War).

That has already been established I believe, probably goes back to PS3 era of game creation due to exotic architecture.
Though despite moving to x86 which should save time, gamers demand more of their games so we're still seeing 3 years minimum.
I'm of opinion some games will get a enhanced edition no matter when PS5 releases, be it 2019 or 2020, think RDR2, DS, TLoU2, GTAVI, HZD2 etc.

Can't Sony and MS just make a bomb ass 600 dollar console and be done with new versions for at least a decade? We all own PS4's and XB1's. Trade in will reduce the cost when it comes time to upgrade. Why the rat race? What happened to ten year cycles?

Would have been great, a true no compromise machine but people will say buy a PC and all that dumb shit.

 

[ArkhamFantasy]

Can't Sony and MS just make a bomb ass 600 dollar console and be done with new versions for at least a decade? We all own PS4's and XB1's. Trade in will reduce the cost when it comes time to upgrade. Why the rat race? What happened to ten year cycles?

Look at the reaction to Scorpio's $499 price point, most people dont want to spend that much money on a console. Price is king.

 

[Shin]

Look at the reaction to Scorpio's $499 price point, most people dont want to spend that much money on a console. Price is king.

It's unfortunate, though given that it's part of the same eco-system having 2 SKU's isn't a bad thing.
If they make back the R&D spent then that should be fine, those that really want the absolute best will buy it.

 

[jdstorm]

Can't Sony and MS just make a bomb ass 600 dollar console and be done with new versions for at least a decade? We all own PS4's and XB1's. Trade in will reduce the cost when it comes time to upgrade. Why the rat race? What happened to ten year cycles?

No because tech doesn't work like that. Ten year cycles were always a stupid idea. Could you imagine being locked into using a iphone 1 right now? Thats essentialy what you are asking for. The Iphone 1 had 128MB of EDRAM and 4GB of flash memory in its cheapest form. The Iphone 7 has 3GB of Ram and can have 128GB or even up to 256GB of memory.

Edit: if you want better hardware the real solution if you are in the US is typically to write to your local polititions and campaign for minimum and median wages to be increased. Giving more people more disposable income is the only way you will see an increased price be tollerated by the market.

 

[RedSwirl]

Not really, just sony stuck with Cell last gen so impossible to do BC. However, PR people like to talk gen no gen rubbish. Both next gen X86's will play current gen, we will just get a whole lot of marketing blurb differences.

Tech-wise it makes the most sense for Sony to do something very similar to what Microsoft is planning, but interviews indicate Sony still believes very much in hard generational transitions. I think PS5 is going to be just a significantly upgraded PS4 under the hood with no requirement to keep supporting PS4. If that happens though the biggest question I have is how Sony will market it.

Will Sony just market it as the next big thing without really mentioning that it's basically a souped-up PS4 Pro with a new CPU, maintaining a "wall" between PS4 and PS5 purely for PR purposes? Will there be PS4 games that can run in a "PS5 mode" to replace cross-gen games and if so, how will that be marketed? Right now I imagine boxes being labeled "PS4 and PS5" the way Microsoft is doing now with BC Xbox 360 games. Or maybe "Enhanced on PS5" or something.

Eventually though all this would be working towards a point where there are no longer PS4 or PS5 or PS6 games but just "PlayStation games," with labels indicating the minimum PlayStation hardware needed to run. I see Microsoft very likely doing this eventually at some point -- you just see "Xbox" on the box and then maybe in fine print it says "Requires Xbox One or Later" or "Requires Xbox One X or Later."

The closest precedent we have for this right now is the Game Boy line. In fact the Xbox One is almost following the same pattern: original release, smaller model, enhanced model that plays the same games, etc. The only difference is the Game Boy Color had a lot of exclusive games while also enhancing non-exclusive games, so maybe that's how PS5 will work.

 

[ArkhamFantasy]

It's unfortunate, though given that it's part of the same eco-system having 2 SKU's isn't a bad thing.
If they make back the R&D spent then that should be fine, those that really want the absolute best will buy it.

I love the Scorpio concept, if you're going to make a premium console then make a premium console. The pro is an affordable premium console, which kind of contradicts itself and isnt really cheap or powerful. I like my pro but i would have happily spent an extra 100$ to make it more like the Scorpio.

I think people just went a little crazy with their Scorpio expectations, they literally thought it was going to change the console war back in MS's favor, thats insane. That's as good of a price as we could have asked for considering the parts inside of it.

 

[avaya]

So take some assumptions....

PS4's Liverpool APU was 28nm TSMC, die size 348mm2
20 CU / 8 x Jaguar
1CU + Cache = 5.6mm2
1 Jaguar + Cache = 6.5mm2

Total size of CUs = 112mm2
Total size of Jaguars = 52mm2

All from chipworks. Those are target die sizes given we want to aim for initial cost of USD90/chip to match PS4 initial pricing.

New CPU

1 Ryzen CCX block (4-scores + 8MB L3) is 44mm2 at 14nm LPP Samsung.

If we move from 14nm to 7nm/GloFlo we can scale down to 0.42x the size (I use 6T-SRAM cell area as a proxy here, it's 0.0269um2 on 7nm/GloFlo vs. 0.064um2 on 14nm/Samsung). Implies we can best case shrink the Ryzen CCX to c.18mm2 at 7nm.

So the next PlayStation has a very good chance of having 2 x Ryzen CCX which is 8 Cores/ 16 Threads and 16MB L3 Cache (best case size of 36mm2 vs. 8 Jaguars at 52mm2 in Liverpool). Caveats apply here (you can't use SRAM cell size as a scaling factor you idiot etc. etc.)

New GPU

On the 28nm TSMC process 1CU = 5.6mm2 including caches. 6T-STRAM cell area on that process is 0.1270um2, at 7nm you could get a chip that is 0.21x smaller.

So how many CUs can we fit in the PS4's original 112mm2 area for 20 CUs? 5.6mm2 x 0.21 = 1.186mm2 new CU area at 7nm, implies you could fit 94CUs on the die.

Liverpool had 10% redundancy so 94 x 90% = 85CUs. This translates into....

85 x 64 Stream Processors x 2 flops/cycle x 800MHz = 8.7TFLOP at Liverpool base clock.

If clock is taken up to 1.1GHz those 85CUs would generate 12TFLOP (2x Scorpio).

We will be lucky to get more than 12TFLOP on the GPU side, CPU side of things is more secure but Sony could skimp and go to 1 Ryzen CCX if it felt it could use more CUs. Such a sacrifice would add another few TFLOPs.

 

[Shin]

So how many CUs can we fit in the PS4's original 112mm2 area for 20 CUs? 5.6mm2 x 0.21 = 1.186mm2 new CU area at 7nm, implies you could fit 94CUs on the die.

It's funny that you bring this up because Navi is rumored to have 128NCU, personally I don't think it will have that many.
But what if it does happen though, then a cut down in consoles would be in the ballpark you mentioned.
I just had another look at AMD's roadmap and Zen 2 is on 7nm already, that's 1H 2018 (not sure about core size).
As for memory frequency I'd expect it to be higher than 1.1Ghz, perhaps closer to 1.2?

 

[avaya]

It's funny that you bring this up because Navi is rumored to have 128NCU, personally I don't think it will have that many.
But what if it does happen though, then a cut down in consoles would be in the ballpark you mentioned.

My assumptions also assume there is not much of a size difference between the old GCN CUs vs. the NCUs, which may not be the case. Also what process is Navi supposed to debut on btw?

 

[Theonik]

^ That's close to what I expect, though Sony might be able to bump up clocks slightly. Who knows. I think they might be able to get to 1.5GHz on the GPU side.

I think they'd go for 16 or 24GB GDDR6 memory in that as well, probably the later, should be a good match with that kind of GPU. Of course this all relies on us getting to 7nm FF+ for APU size dies by 2020. It is likely but process is slowing and phones are soaking up a lot of the production.

 

[Shin]

My assumptions also assume there is not much of a size difference between the old GCN CUs vs. the NCUs, which may not be the case. Also what process is Navi supposed to debut on btw?

7nm, though it was already pushed back 1 year from 2018 to 2019, so maybe it will launch with 7nm EUV (at least it's available at GloFo if AMD finds it necessary).
Great post you wrote there btw, I like reading that kind of stuff.

^ That's close to what I expect, though Sony might be able to bump up clocks slightly. Who knows. I think they might be able to get to 1.5GHz on the GPU side.

I think they'd go for 16 or 24GB GDDR6 memory in that as well, probably the later, should be a good match with that kind of GPU. Of course this all relies on us getting to 7nm FF+ for APU size dies by 2020. It is likely but process is slowing and phones are soaking up a lot of the production.

Actually it makes sense to run them at lower speeds because that thing is going to run hot as hell, it's also 2x+ XBOX's CU count.
24Gb would be nice and leaves a lot of headroom, all developers could seamlessly stream data without much loading (if at all).
Game will be created for 4K I think, whatever PC GPU's are using ATM shouldn't be the baseline the amount, VRAM required might be increased.

 

[avaya]

7nm, though it was already pushed back 1 year from 2018 to 2019, so maybe it will launch with 7nm EUV (at least it's available at GloFo if AMD finds it necessary).
Great post you wrote there btw, I like reading that kind of stuff.

Cheers. I would have imagined they would debut it on the first 7nm process available which would mean volume in Q4 2019 on current GloFlo roadmap in this case.

^ That's close to what I expect, though Sony might be able to bump up clocks slightly. Who knows. I think they might be able to get to 1.5GHz on the GPU side.

I think they'd go for 16 or 24GB GDDR6 memory in that as well, probably the later, should be a good match with that kind of GPU. Of course this all relies on us getting to 7nm FF+ for APU size dies by 2020. It is likely but process is slowing and phones are soaking up a lot of the production.

The memory I am pretty certain will not be stacked on die any variant of the type because they are so far from console form factor prime time. GDDR6 is a solid bet, I remember it may have been you that mentioned it previous but what is the reason for the 16 or 24GB increment? Is it based on the density available at the time? Also what is the bus size expected in each scenario and the best case peak bandwidth? 768GB/s?

Regarding GPU clock... I went for 1.1GHz for base clock on GPU though to reach a 12TFLOP target and at your 1.5GHz assumption those 85CUs would be 16TFLOP (expecting PC parts with double the die size at 30TFLOP).

I am hopeful for Q4 2019 debut, but I can easily see it slipping into H1 2020.

 

[Theonik]

Cheers. I would have imagined they would debut it on the first 7nm process available which would mean volume in Q4 2019 on current GloFlo roadmap in this case.



The memory I am pretty certain will not be stacked on die any variant of the type because they are so far from console form factor prime time. GDDR6 is a solid bet, I remember it may have been you that mentioned it previous but what is the reason for the 16 or 24GB increment? Is it based on the density available at the time? Also what is the bus size expected in each scenario and the best case peak bandwidth? 768GB/s?

Regarding GPU clock... I went for 1.1GHz for base clock on GPU though to reach a 12TFLOP target and at your 1.5GHz assumption those 85CUs would be 16TFLOP (expecting PC parts with double the die size at 30TFLOP).

I am hopeful for Q4 2019 debut, but I can easily see it slipping into H1 2020.

GDDR6 modules are going to be available early next year in 8 and 16Gb densities. Not every manufacturer is making the former chips though. 8 chips with 32 bit interfaces gives a 256bit bus, same as the PS4, for a capacity of 16GB and 512GB/s in bandwidth. (at their 16Gbps clocks) Alternatively 12 chips in that density as found on the XBO X yields a 384-bit bus and 768GB/s of bandwidth, with a total capacity of 24GB.

In theory they can do an odd arrangement as well if they want or they can get lower clocked modules for cheaper if those come to pass. It's how Sony was able to double from 4 to 8GB in the last minute with PS4. (their suppliers got slower chips with 2x the density in the last minute, bumping from 4 to 8Gb) Micron is launching at 12-14Gbps. The JEDEC standard for GDDR6 allows up to 32Gb modules which means you could see double that amount as well but I doubt Sony gets 32Gb modules at an affordable price by 2019/20 or at all. (JEDEC GDDR5 allows up to 16Gb modules but we never saw those either)

E:
The theoretical peak before Sony needs to do something very clever and/or break the bank, bearing in mind the size of the APU restricts the width of the bus, is 16 chips of 32Gb densities for 64GB of ram on a 512bit bus for 1TB/s of bandwidth. Not likely to happen though.
/E

Clockwise, AMD seems to be able to get Polaris to 1.4GHz on the RX580 chips, the XBO X takes advantage of those improvements to hit these clocks. I expect AMD might be able to get to 1.5 for an APU in 2019/20 though if it's made on 7nm it might be that they process is not really mature enough to push very high clocks on it and they might opt for 1.2-1.3GHz. We'll see I guess.

 

[jdstorm]

I know nothing about the specifics of how this stuff works.

In saying that 16-24GB still feels really low for a console that will have to exist in 2025.

Especially since 4K will become more of a thing.

With a likely 300% in CPU power allowing for a greater amount of objects on screen that wouldn't that jump essentially represent a need for 3x the memory assuming it was mostly used on more objects/NPCs and not smarter AI?

Then beyond that wouldn't the jump from 1080p to 4K use increased texture sizes. So wouldn't recreating a 2016 level game with 3x the objects and 4K textures take

8GB x 4 x the difference between 1080p textures/shaders and 4K textures/Shaders?

 

[Shin]

I am hopeful for Q4 2019 debut, but I can easily see it slipping into H1 2020.

Seems that TSMC already have risk production planned of 5nm in 2019, given that Sony/AMD used them for PS4/Pro it might be where they manufacturer PS5 SoC.
It should be cheaper also because it's a Taiwanese company, where else GloFlo is owned by a company from UAE and most of their fabs are located in the West.
One year later for mass production puts us at 2020, Zen 3 and Navi+ will be ready long by then (I'm guessing AMD will use GloFlo for the PC/desktop market).

 

[Matt]

I know nothing about the specifics of how this stuff works.

In saying that 16-24GB still feels really low for a console that will have to exist in 2025.

Especially since 4K will become more of a thing.

With a likely 300% in CPU power allowing for a greater amount of objects on screen that wouldn't that jump essentially represent a need for 3x the memory assuming it was mostly used on more objects/NPCs and not smarter AI?

Then beyond that wouldn't the jump from 1080p to 4K use increased texture sizes. So wouldn't recreating a 2016 level game with 3x the objects and 4K textures take

8GB x 4 x the difference between 1080p textures/shaders and 4K textures/Shaders?

It's fine.

 

[Rosstimus]

So Pachter thinks game consoles are headed toward an iPhone style iteration cycle huh? I'm not so sure personally. I think we'll see a "real" PS5, but it won't get released til 2022 or 2023 ish. After last gen I think that long console cycles are here to stay.

 

[kyser73]

Tech-wise it makes the most sense for Sony to do something very similar to what Microsoft is planning, but interviews indicate Sony still believes very much in hard generational transitions. I think PS5 is going to be just a significantly upgraded PS4 under the hood with no requirement to keep supporting PS4. If that happens though the biggest question I have is how Sony will market it.

Will Sony just market it as the next big thing without really mentioning that it's basically a souped-up PS4 Pro with a new CPU, maintaining a "wall" between PS4 and PS5 purely for PR purposes? Will there be PS4 games that can run in a "PS5 mode" to replace cross-gen games and if so, how will that be marketed? Right now I imagine boxes being labeled "PS4 and PS5" the way Microsoft is doing now with BC Xbox 360 games. Or maybe "Enhanced on PS5" or something.

Eventually though all this would be working towards a point where there are no longer PS4 or PS5 or PS6 games but just "PlayStation games," with labels indicating the minimum PlayStation hardware needed to run. I see Microsoft very likely doing this eventually at some point -- you just see "Xbox" on the box and then maybe in fine print it says "Requires Xbox One or Later" or "Requires Xbox One X or Later."

The closest precedent we have for this right now is the Game Boy line. In fact the Xbox One is almost following the same pattern: original release, smaller model, enhanced model that plays the same games, etc. The only difference is the Game Boy Color had a lot of exclusive games while also enhancing non-exclusive games, so maybe that's how PS5 will work.

You're overthinking this.

The PS5 can be marketed like any other new-gen console. If Sony have demonstrated one thing over 4 generations it's that they've got a strong and numerous core audience who will buy their kit.

 

[Shin]

Still curious how they'll increase the amount of memory without widening the bus.
I've yet to find a GPU or console where the amount of memory x 32-bit isn't the equation.
PS4 is 256-bit (8x32), XBOX is (12x32) 386-bit, 16Gb would result in 512-bit, something is off with this.
So I wonder how they are going to increase the amount of memory while staying at 384-bit, people keep saying that 512-bit isn't doable and bla bla but no proof to back it up.
Assuming they go for 384-bit as well then isn't it going to be either 12Gb or 24Gb (clam-shell) or am I seeing things wrong if so prove how?

 

[mrklaw]

So take some assumptions....

PS4's Liverpool APU was 28nm TSMC, die size 348mm2
20 CU / 8 x Jaguar
1CU + Cache = 5.6mm2
1 Jaguar + Cache = 6.5mm2

Total size of CUs = 112mm2
Total size of Jaguars = 52mm2

All from chipworks. Those are target die sizes given we want to aim for initial cost of USD90/chip to match PS4 initial pricing.

New CPU

1 Ryzen CCX block (4-scores + 8MB L3) is 44mm2 at 14nm LPP Samsung.

If we move from 14nm to 7nm/GloFlo we can scale down to 0.42x the size (I use 6T-SRAM cell area as a proxy here, it's 0.0269um2 on 7nm/GloFlo vs. 0.064um2 on 14nm/Samsung). Implies we can best case shrink the Ryzen CCX to c.18mm2 at 7nm.

So the next PlayStation has a very good chance of having 2 x Ryzen CCX which is 8 Cores/ 16 Threads and 16MB L3 Cache (best case size of 36mm2 vs. 8 Jaguars at 52mm2 in Liverpool). Caveats apply here (you can't use SRAM cell size as a scaling factor you idiot etc. etc.)

New GPU

On the 28nm TSMC process 1CU = 5.6mm2 including caches. 6T-STRAM cell area on that process is 0.1270um2, at 7nm you could get a chip that is 0.21x smaller.

So how many CUs can we fit in the PS4's original 112mm2 area for 20 CUs? 5.6mm2 x 0.21 = 1.186mm2 new CU area at 7nm, implies you could fit 94CUs on the die.

Liverpool had 10% redundancy so 94 x 90% = 85CUs. This translates into....

85 x 64 Stream Processors x 2 flops/cycle x 800MHz = 8.7TFLOP at Liverpool base clock.

If clock is taken up to 1.1GHz those 85CUs would generate 12TFLOP (2x Scorpio).

We will be lucky to get more than 12TFLOP on the GPU side, CPU side of things is more secure but Sony could skimp and go to 1 Ryzen CCX if it felt it could use more CUs. Such a sacrifice would add another few TFLOPs.

Nice. Couple of things. Active area on PS4 APU was only about half the die partly to layout spacing but partly for housekeeping circuits. This should remain fairly static and not need to increase next gen, so would that result in a net gain of active dir space so you could target more than e.g. 152mm2 for the GPU?

Also any info yet on likely navi CU size? Presumably there will be performance improvements at the same size so that TFLOP number might go up by a few percent

 

[THE:MILKMAN]

I was looking at a Orbis roadmap from an old GAF thread and wonder if it gives a resonable pointer as to where PS5 is at in development currently?

2011 June/July – PC with Win7 64-bit and a "jailbroken" ATI r9 graphics card
2011 Sept/Oct – PC with Win7 64-bit and a "jailbroken" ATI r10 graphics card
2012 Q1 – disclosure to more developers
2012 E3 – potential unveiling window start
2012 Jul – devkits for engineers writing OS
2012 Q3 – first true hardware prototype devkits
2013 E3 – potential unveiling window end
2013 Q4 – launch

Now I'm not sure the first two 2011 steps would be required this time round but think its possible PS5 is at the 2012- Disclosure to more developers step (like Matt).

To back up the above here is one of the earlier articles I can find about Orbis that has interesting info when compared to the above VGLeaks roadmap:

Our main source tell us that "select developers" have been receiving dev kits for the new console since the beginning of this year. Revised and improved versions of these kits were sent out around GDC, while more finalised beta units will be shipped to developers towards the end of 2012.

http://kotaku.com/5896996/the-next-playstation-is-called-orbis-sources-say-here-are-the-details

I can't see a reasonable reason why developers would need any info <July 2017 if it was to launch late 2019 or later? Combine that with all the articles about PS5 and all the Sony people freely talking too and it seems we aren't very far off at all.

 

[notaskwid]

I think they should try to get TLOU2 out for PS4 since people has already bought a PS4 to play it. Same thing for other announced and unreleased games that might be out in 2019, like Shenmue 3 and FF7 Remake. Otherwise they're just going to piss people off and the negative word of mouth will hurt the sales.
Unless they go for full forward and backward compatibility, then they can do whatever they want. That would probably be the best approach if the gens gets chopped up into smaller bumps. You jump in whenever you want at whatever quality you can afford.

Just like Versus XIII, Agent, The Last Guardian, Nioh, etc pissed people off the PS4.
Not saying those games won't be released for the PS4 though, I'm pretty sure that at the very least TLOU 2 will.

 

[Shin]

I was looking at a Orbis roadmap from an old GAF thread and wonder if it gives a resonable pointer as to where PS5 is at in development currently?

Things started rolling last year April with the move of a key engineer back to the hardware department.
I've kept in mind that SCE & SNEI merged into SIE in April 2016, but he also went back to his previous role - hardware engineer.
Cerny said that development on PS4 started in 2008, you can find it in a Euro gamer article, everything matches Masayasu Ito LinkedIn profile.

 

[EASTC0AST_GH05T]

Are Pacthers predictions usually spot on? I only ask because this news is starting to follow me around the web outside of GAF.