Next Generation Hardware Speculation with a Technical Spin [2018]

[phoenix_chipset]

I don't think most people would rush out and buy a 8TF PS5 no matter how you can justify its graphics leap. Customers like numbers and first thing they'll do is compare it to the 1X and 8 vs 6 is a terrible showing of power, it's borderline a carbon copy. 8TF won't even render some of the more demanding current gen games at 4k/30fps much less next gen. I really hope Sony don't release anything less than a 12TF base PS5 regardless of its graphics leap, it's marketing suicide otherwise.

Sounds hard to believe. What games are there that vega 56 can't run at 4k 30? Off the top of my head I can only think of quantum break. Or FF15 but neither one are the best of ports let's be honest.

Besides, personally I see how gorgeous ratchet and clank is on the pro, or the detail in shadow of the colossus at 1080p60 and that's at 4tf... yeah, 8tf won't lead to insufficiently great looking games. You act as if the Xbox one X isn't already an impressive graphical display.

We can have 12tfs+ consoles, just expect to pay either a hell of a lot up front, or have no ssd and be prepared for some 4-5 minute load screens on games like gta and stuttering while asset streaming.

 

[ultragpu]

Sounds hard to believe. What games are there that vega 56 can't run at 4k 30? Off the top of my head I can only think of quantum break. Or FF15 but neither one are the best of ports let's be honest.

Besides, personally I see how gorgeous ratchet and clank is on the pro, or the detail in shadow of the colossus at 1080p60 and that's at 4tf... yeah, 8tf won't lead to insufficiently great looking games. You act as if the Xbox one X isn't already an impressive graphical display.

We can have 12tfs+ consoles, just expect to pay either a hell of a lot up front, or have no ssd and be prepared for some 4-5 minute load screens on games like gta and stuttering while asset streaming.

The 1X is hardly impressive, the combination of ultra settings or Very High settings with native 4k rendering is just way too much for it, it can't even do CBR 4k 100% of time. A 8TF console coding for next gen graphics natively will indeed yield much better graphics but will be at the expense of native resolution. So if we're talking solely about a console aiming for native 4k/30 next gen graphics it's vastly underpowered period, it's pathetic in that sense.
I'm all for $499, $599 or higher cost brutes or just wait for a few more years to see what happens.

 

[RDGoodla]

We can have 12tfs+ consoles, just expect to pay either a hell of a lot up front, or have no ssd and be prepared for some 4-5 minute load screens on games like gta and stuttering while asset streaming.

If next gen consoles use 7nm process, we can calculate the difference in power consumption. TSMC 10nm finfet consumes 35% less power than 16nm finfet, and 7nm is 40% less than 10nm.That means 7nm transistor consumes 65% x 60% = 39% power of a 16nm transistor.

Therefore we can expect next gen consoles with 2.5X performance of 16nm x1x, which means 15TF. If PS5 choose 7nm Nvidia GPU then PS5 may have 30% more performance which equivalent to a 20TF AMD GPU.

 

[RDGoodla]

Price for one thing. Performance per dollar is more important for a console than performance per watt.

Since Nvidia GPU often has 30% more performance per watt than AMD GPU, does AMD still have significant advantage considering performance per dollar?

 

[Grall]

We can have 12tfs+ consoles, just expect to pay either a hell of a lot up front, or have no ssd and be prepared for some 4-5 minute load screens on games like gta and stuttering while asset streaming.

Yes, the loading will get really problematic with so much RAM (it's already slow as it is.) What about striped harddrives for next gen? SSD is probably a bit too much to ask yet methinks. Maybe they could add a M.2 socket where users could install one optionally... Well, I can dream can't I? (After all, Nintendo consoles all up through Gamecube had expansion slots underneath them... )

Since Nvidia GPU often has 30% more performance per watt than AMD GPU, does AMD still have significant advantage considering performance per dollar?

Nvidia is greedy (read: expensive.) Also, their architecture is different, so console development tools and programmers would all have to be re-geared. Also, NV has no x86 processor cores, so more re-gearing necessary there, possibly quite some unpredictability initially where performance is concerned and so on.

Also, both Sony and MS would have to have the desire to switch to NV, or there'd be two different console architectures with vastly different basic characteristics on the market, which big-game developers won't like; it would add a lot of additional costs on top of an already very expensive game development process.

The console which sticks with x86 (IE, AMD) would (speculation alert!) probably see the most development effort, since porting to PC would be comparatively less painful.

 

[Shifty Geezer]

Since Nvidia GPU often has 30% more performance per watt than AMD GPU, does AMD still have significant advantage considering performance per dollar?

Probably. nVidia is selling to other far more lucrative sectors like super computing and AI cars. AlStrong posted a slide somewhere recently that showed gaming as a $100 billion market and automobile as a $10 trillion. Console money is a rounding error in that space. nVidia is going to sell its silicon to the markets that generate the most revenue and pay the highest premiums. Unless they do it for charity or PR, they won't offer the consoles an economic part, meaning whatever extra is needed to be spent to cool an AMD solution is still probably cheaper.

As others say, you also have to think about CPU. Is it more cost effective to go with an nVidia and an ARM core, or to go nVidia and Intel CPU (ha ha ha), or to go Intel SoC with AMD integrated graphics, or to go AMD all-in-one? The latter has the most invested in the console space and good working relationships.

 

[phoenix_chipset]

Last I checked the high end sata drives from crucial and Samsung cost about $240 for a TB, now consider the price almost 2 years from now with the bulk buying discount. Plus they wouldn't use the most expensive drives like that anyway. Imo skipping the hybrid solutions and just going full SSD will be easily doable with a touch of sacrifice elsewhere.

 

[AlNom]

AlStrong posted a slide somewhere recently

Couldn't be!

 

[Shifty Geezer]

Pretty sure it was your evil alter-ego posting through a time-displaced connection from the past. If that wasn't the case (seems unlikely), maybe I typed your old name by mistake?

 

[Deleted member 13524]

I think releasing two tiers with 2 performance targets at launch would be a mistake for consoles and therefore won't happen.

First because all consoles suffer from a reduced catalog at launch and Sony/MS definitely don't want devs spending additional time on two performance targets that would delay their release schedules.
Second because splitting consumer base at launch isn't very smart, as fewer gamers will want to get the "poor version" and might just wait until the faster version is cheaper - which in turn will hamper the adoption rate.
Third because the main point of the mid-gens is to keep the hardware updated throughout the typical ~6-7 year cycle for keeping PC migration in check and to have a faster reaction to the competitor's offerings. So Sony will probably still want to release a mid-gen in ~2022, and a high-end version launched in 2019 would make this mid-gen harder to develop.

I don't think most people would rush out and buy a 8TF PS5 no matter how you can justify its graphics leap. Customers like numbers and first thing they'll do is compare it to the 1X and 8 vs 6 is a terrible showing of power, it's borderline a carbon copy.

What if Sony sells the PS5 as a 18 TFLOP FP16 console? Is that better?
10 times more compute power than the PS4!!*

Regardless, should Sony be that worried about the XboneX adoption?
Sure, processing power probably played a relevant role back in 2013/2014. But in 2016-2018 I think Sony's merit stands on the software.

In a time where all the big publishers are trying really hard to force feed us with games-as-a-service in every genre, Sony is right now the publisher launching the largest amount of AAA single-player games with no strings attached.
During a 3 year period (2016-2018) we'll have The Last Guardian, Uncharted 4, Gravity Rush 2, Horizon Zero Dawn, Knack 2, GT Sport, Detroit Become Human, God of War 4, Dreams and Insomniac's Spider-Man. And even for 2019 we'll get TLOU 2 and Death Stranding.
I'm only talking about AAA games with a full-fledged single player experience. No "half-game plus the rest on DLCs" and no microtransactions. I'm not even counting with remasters, PSVR games, PlayLink and the smaller PSN games.
Simply no one else is pumping out 3 major AAA single-player full-experience games so consistently.

 

[Anarchist4000]

Yes, the loading will get really problematic with so much RAM (it's already slow as it is.) What about striped harddrives for next gen? SSD is probably a bit too much to ask yet methinks. Maybe they could add a M.2 socket where users could install one optionally... Well, I can dream can't I? (After all, Nintendo consoles all up through Gamecube had expansion slots underneath them... )

Maybe not M.2, but flash storage is getting rather fast and affordable. A top of the line SSD/NVMe drive wouldn't be a requirement. Would definitely be more compact and durable. Going to a cartridge style approach, with all active games plugged in, might be practical. Could also work as expansion with a dozen or so devices hot plugging for added storage.

The console which sticks with x86 (IE, AMD) would (speculation alert!) probably see the most development effort, since porting to PC would be comparatively less painful.

With the AIO designs, there could be a strong argument for adopting one out right or even virtualizing the platform. That may be cheaper than rolling a new hardware platform every couple years and Hades Canyon for example would even be smaller than current offerings. Ignoring price for the time being, but scaling the concept would help.

 

[Grall]

Going to a cartridge style approach, with all active games plugged in, might be practical.

You're forgetting the games market is quickly heading towards downloaded software rather than physical. With increasing numbers of games stores going bankrupt there won't be a place to buy cartridge games for many people, and this trend will only continue going forwards. You could buy physical games online, but why do that and then wait a couple days for delivery (more if there's a weekend or holiday in the way) when you can just download the game and have it playable in perhaps an hour or even less, depending on connection.

Physical games also take up shelf space, and lives are already cluttered enough for many people.

 

[Shifty Geezer]

I read the 'cartridge style approach' as meaning flash storage provided as blank cartridges you could plug in and download/install to. That'd be pretty cool, having storage with a few games on you can swap out with another 'stick', whatever format it takes. Rather than buying a large external HDD and having it sitting at the end of a lead, or a tiny SD card that can only fit one game, go the middle route and have maybe 256 GBs flash in a Game Cart that you can have a few games on, and have multiple of these in a drawer. Then have a classy paper label you can write on with a Biro for the games contents, which you can cross out when you change them. Sweet!

 

[Anarchist4000]

That's roughly what I was going for. Even the tiny SD cards are up to 256GB now and lining the side of the box with USB ports or SD card readers wouldn't be difficult. Storage will get cheaper over time and users could add storage easily enough if required. Not unlike Steam having library folders that could be removable.

 

[east of eastside]

Probably. nVidia is selling to other far more lucrative sectors like super computing and AI cars. AlStrong posted a slide somewhere recently that showed gaming as a $100 billion market and automobile as a $10 trillion. Console money is a rounding error in that space. nVidia is going to sell its silicon to the markets that generate the most revenue and pay the highest premiums. Unless they do it for charity or PR, they won't offer the consoles an economic part, meaning whatever extra is needed to be spent to cool an AMD solution is still probably cheaper.

As others say, you also have to think about CPU. Is it more cost effective to go with an nVidia and an ARM core, or to go nVidia and Intel CPU (ha ha ha), or to go Intel SoC with AMD integrated graphics, or to go AMD all-in-one? The latter has the most invested in the console space and good working relationships.

NVIDIA *is* in a position of significant financial strength that it could bid for a next-gen console contract as a PR win, and to undermine AMD's semi-custom business.

Furthermore, consider the following other larger strategic business factors that would support an NVIDIA+ARM+Sony PS5 console tie-up--

ARM is owned by Japanese mega-conglomerate Softbank. Softbank has a $4 billion dollar investment stake in NVIDIA through its Vision fund-(https://www.bloomberg.com/news/arti...-take-4-billion-stake-in-u-s-chipmaker-nvidia).

Softbank operates ARM with more financial resources focusing on growth over profits than when it was a publicly listed company. Quote directly from ARM CEO, Simon Segars-

He said the Japanese company, led by Masayoshi Son, has provided Arm with more “raw resources” than it had as a listed business and has left the entity to operate independently.

https://www.ft.com/content/268e495e-1c98-11e8-aaca-4574d7dabfb6

Further, same article, Segars, explains Softbank CEO, Masayoshi Son, encourages 'synergies' between Softbank businesses, i.e., NVIDIA+ARM.

“Masa is keen on synergies between SoftBank entities — Arm and Sprint [the US telecoms company] for example —

So, the larger strategic case for an NVIDIA/ARM/Sony tie-up is certainly there. NVIDIA could go for the PR win while Softbank could broker the sweetheart deal with Sony for PS5, acting entirely consistent with how it operates its businesses. The subsidy for the break-even deal could come from Softbank and not NVIDIA and it would be a win-win for Softbank.

Other factors..

Japanese game industry consolidation around ARM/NVIDIA, two Japanese tied chips. Both PS5 and Switch would share ARM/NVIDIA architecture in common, a boost of efficiency for Japanese game industry. PS5p portable spin-off also would be facilitated by ARM/NVIDIA hardware.

It is also a Japanese cultural norm to pick the Japanese supplier over foreign whenever possible, i.e., Nintendo wanting NEC in GameCube and SEGA choosing NEC over 3DFX in Dreamcast.

Further, I see Sony going for a 'clean-break' tech-wise and Apple like 'silo' strategy w/ PS5. There is a mistaken assumption repeated that Sony wants to further align with PC when, in reality, because of PS4 strength and focus on 'exclusives', Sony's more likely strategic move is an Apple like break-away to its own silo. "You want the best games on the best tech, you play in our eco-system".

 

[BRiT]

I could see Sony going more for ARM+Nvidia on PS5 because they're not as invested in backwards compatibility as Microsoft is. I don't think it's likely at all, but if anyone was going to go that way it would be Sony before MS.

 

[Deleted member 13524]

NVIDIA *is* in a position of significant financial strength that it could bid for a next-gen console contract as a PR win, and to undermine AMD's semi-custom business.

Furthermore, consider the following other larger strategic business factors that would support an NVIDIA+ARM+Sony PS5 console tie-up--

ARM is owned by Japanese mega-conglomerate Softbank. Softbank has a $4 billion dollar investment stake in NVIDIA through its Vision fund-(https://www.bloomberg.com/news/arti...-take-4-billion-stake-in-u-s-chipmaker-nvidia).

Softbank operates ARM with more financial resources focusing on growth over profits than when it was a publicly listed company. Quote directly from ARM CEO, Simon Segars-

https://www.ft.com/content/268e495e-1c98-11e8-aaca-4574d7dabfb6

Further, same article, Segars, explains Softbank CEO, Masayoshi Son, encourages 'synergies' between Softbank businesses, i.e., NVIDIA+ARM.



So, the larger strategic case for an NVIDIA/ARM/Sony tie-up is certainly there. NVIDIA could go for the PR win while Softbank could broker the sweetheart deal with Sony for PS5, acting entirely consistent with how it operates its businesses. The subsidy for the break-even deal could come from Softbank and not NVIDIA and it would be a win-win for Softbank.

Other factors..

Japanese game industry consolidation around ARM/NVIDIA, two Japanese tied chips. Both PS5 and Switch would share ARM/NVIDIA architecture in common, a boost of efficiency for Japanese game industry. PS5p portable spin-off also would be facilitated by ARM/NVIDIA hardware.

It is also a Japanese cultural norm to pick the Japanese supplier over foreign whenever possible, i.e., Nintendo wanting NEC in GameCube and SEGA choosing NEC over 3DFX in Dreamcast.

Further, I see Sony going for a 'clean-break' tech-wise and Apple like 'silo' strategy w/ PS5. There is a mistaken assumption repeated that Sony wants to further align with PC when, in reality, because of PS4 strength and focus on 'exclusives', Sony's more likely strategic move is an Apple like break-away to its own silo. "You want the best games on the best tech, you play in our eco-system".

When was the last time Sony used japanese tech?
Better yet, when was the last time Sony used japanese tech for the sake of it being japanese?

I'd get your post if you were talking about Nintendo, but not Sony.

 

[ultragpu]

What if Sony sells the PS5 as a 18 TFLOP FP16 console? Is that better?
10 times more compute power than the PS4!!*

Regardless, should Sony be that worried about the XboneX adoption?
Sure, processing power probably played a relevant role back in 2013/2014. But in 2016-2018 I think Sony's merit stands on the software.

In a time where all the big publishers are trying really hard to force feed us with games-as-a-service in every genre, Sony is right now the publisher launching the largest amount of AAA single-player games with no strings attached.
During a 3 year period (2016-2018) we'll have The Last Guardian, Uncharted 4, Gravity Rush 2, Horizon Zero Dawn, Knack 2, GT Sport, Detroit Become Human, God of War 4, Dreams and Insomniac's Spider-Man. And even for 2019 we'll get TLOU 2 and Death Stranding.
I'm only talking about AAA games with a full-fledged single player experience. No "half-game plus the rest on DLCs" and no microtransactions. I'm not even counting with remasters, PSVR games, PlayLink and the smaller PSN games.
Simply no one else is pumping out 3 major AAA single-player full-experience games so consistently.

But then the competition could always counter with 12 TF FP16 as well, there's no use for them to be deceptive about it especially in the days of internet, tech sites, YT and social media.
Sony really don't have much to worry about right now, not even the best looking X1 enhanced titles could beat most of their first party titles available and upcoming. Like you said Sony thrives on AAA budget no nonsense SP games and that's truly a huge driving point to their success but one should never rest on their laurels at any stage. The stage is always changing, evolving and to be left behind or nerfed in power is the last thing you want in such a competitive industry. Sony should play their cards right, rushing out with a weak console is not beneficial to anyone.

 

[MrFox]

Well they could mention both numbers correctly like we're now seeing with GPUs.

Cerny didn't even mention FP16 nor ID Buffer when he unveiled the Pro and only talked about that at GDC because it's not simple to explain and would have been misinterpreted. He's never been known to over hype.

Over hyping is generally known as a dangerous strategy which can backfire. Like the whole cell processor hype (kutaragi), or the cloud giving the power of 3 consoles to each player (spencer), or esram is magical and nobody can see the full 1080p on a 60" TV (panello). Because later on, these execs might end up on the other side of the question, and they can't erase the internet.

If they use some form of tensor cores next gen, it's not going to be easy giving FP numbers without creating ridiculous hype.

 

[AlNom]

"can't TOP this!"