Next Gen GPU architecture GCN, RDNA, Navi 10/20 (PS5 Navi Hybrid, Xbox Navi Pure) *spawn*

Status
Not open for further replies.
That's a bit of a stretch. Mid doesn't need be exact. We have 'mid-life' not at the mid-point of our lives, not knowing how long we're actually going to live. Mid gen refresh could comfortably be 3 years in and 4 or 5 years out and still be a mid-gen device rather than a 3/8ths generation update. ;)

True enough but the reason Sony gave for PS4 Pro was to mitigate the bleed of PlayStations owners to PC. Clearly Sony felt three years was the right time to prevent that so why is it okey to slip to four/five/six+ years from PS4 Pro?

Hardware wise. PS5 is and has always been a new generation in the traditional sense for Sony. It's only recently that they confirmed BC; perhaps they were waiting on emulators and whatnot to be solid enough before finally committing to it, and were always ready to launch without if beneficial?

You make it sound like Sony announced PS5 some time back and only just confirmed PS4 backwards compatibility but they happened at the same time in the same Wired article.

I disagree with the rest. In the traditional and literal sense, a new Sony console brought a radically different CPU, GPU, broader architecture and platform ecosystem and any form of backwards compatibility was the inclusion of some previous generation hardware. PS5 is the polar opposite of that in that it will have the modern evolution of the same CPU, GPU and broader architecture, with some solid state tech thrown in for good measure.

I am not expecting Sony to throw the baby out with the bathwater by building a complete new devkit, SDK and ecosystem platform because they will want to give developers the opportunity to continue to target PS4 owners, with the option of better experiences on PS5 just as there are for PS4 Pro.

In an alternate timeline, 2013 PS4 > 2018 4Pro > 2020 PS5, 4Pro fits just as it does in the current predicted timeline of 2013 PS4 > 2016 4Pro > 2020 PS5. that PS5 also being possible in 2019, or 2021. A later 4Pro would have had different sales of course, with more people being likely to wait until PS5 then upgrade.

In that alternative timeline Sony would have lost more PS4 owners to PC. Presumably three years was the sweetspot for them to maintain continued customer interest and investment in the platform. I know it's Sony but I'm pretty sure they didn't just throw at dart at a calendar. Well, mostly sure.
 
I am not expecting Sony to throw the baby out with the bathwater by building a complete new devkit, SDK and ecosystem platform because they will want to give developers the opportunity to continue to target PS4 owners, with the option of better experiences on PS5 just as there are for PS4 Pro.

Sure, but my point with the "reset" was about the appearance of 4Pro and now the impending generation where a majority of users are still on the base SKU, so presumably most are simply waiting for a bigger leap than the tic-toc update (always been ticked about Intel's definition but maybe I'm forgetting some explanatory talk about why the tic isn't the first iteration followed by toc, then starting on a new gen with the next tic and awerjawoifadfl kaf :V ) even if it's conceivable that there can still be a lot of downports owing to the more stable architecture. Folks are perhaps not a little conditioned already by the phone cycles, sticking to a similar pattern of purchasing with "every-other" HW update - be it always the new numbered generation (the bigger shift) or always the update SKU ("S").
 
Why would Sony release PS4 Pro (2016) just three years after PS4 (2013). If PS4 Pro was a "mid-generation" system, wouldn't that have made 2019 the logical launch year (six years after PS4), for the next console?
.
the ps4pro was made to fill the 4K market, it coincided with that market need and thats was why it came out when it did.
With the ps4 going so strong l felt 2019 was not the right year.
Further more the tech used wouldn’t be ready in 2019 at the price point suitable for the console space .
After all this gen they’re using up to date and cutting edge tech, so for all to fall into place in 2019 was a big ask.
They’re my reasons.
 
Slow down about the RT thing. On paper you can do RT with Vega/Pascal. So they can say "yes we have RT", not talking about performances, and be right. I don't see RT dedicaced hardware (nVidia style) in a 2020 console... It would take too much space.
Now, in a console and with a semi custom gpu, I've no doubt they can tweak vega/navi to be more efficient at it, but I don't believe it will be usable in most games... Like the 4k hype about the ps4pro. It's possible, but native 4k is out of reach for most games due to performances...
 
Slow down about the RT thing. On paper you can do RT with Vega/Pascal. So they can say "yes we have RT",
Declaring you have generic RT without intersection acceleration doesn’t seem... viable performance wise. You may as well announce that current generation supports RT.

Edit: to note, powerVR ray tracing was designed for mobile performance. Low wattage, and very minimal silicon budget. Very ideal for consoles if they can ramp up the performance to match the available compute.
 
Last edited:
Slow down about the RT thing.
We have solid word that Sony wanted RT as part of their hardware. however it's implemented, it must be fast enough to be viable for at least Sony's first parties to use effectively (those first parties being who Sony would run the potential tech by to see if it's worth adding or not).

Edit: I'll double-down on my response by linking this again...

sony-playstation-5-2-650x355.jpg


Listing 'SSD', we know Sony haven't just slapped an SSD in there, but have developed and patented techs to make an optimal solution. The addition of RT isn't just lip-service to hardware capable of running RT in theory but never to be used in practice as it isn't fast enough - it's one of the key tenets of the platform that Sony have felt need to mention already.
 
Last edited:
And just to go over a couple of other things I noted from the Wired article.

1, The initial and immediate response to the revealed specs was $500 minimum, this will be $1000! No way this is $399 etc. This prompted a very swift response on Twitter by Peter Rubin to post an exchange about price he felt the need to leave out of the original article!?

2, He talked specifically and only about Death Stranding being a two-platform release while noting Mark gave a smile and pregnant pause when asked about it. Since which we got a surprise to many November release date for the game.
 
I’m not going to say I’m team2019 or team2020. But I will acknowledge that the once incoherent set of rumours have come together to produce what appears to be a viable roadmap for Sony to have taken to launch 2019 if that was/is their intent. I’m not sure if they got caught up in delays or what not, but if they decided to go with the 2019 roadmap/design, at the very least I would expect it to release earlier than XB2
 
I’m not going to say I’m team2019 or team2020. But I will acknowledge that the once incoherent set of rumours have come together to produce what appears to be a viable roadmap for Sony to have taken to launch 2019 if that was/is their intent. I’m not sure if they got caught up in delays or what not, but if they decided to go with the 2019 roadmap/design, at the very least I would expect it to release earlier than XB2

I agree. That is where I'm at right now. A mod on Era did say last year 2019 was the plan but never did confirm the rumours of a delay.
 
We have solid word that Sony wanted RT as part of their hardware. however it's implemented, it must be fast enough to be viable for at least Sony's first parties to use effectively ...

I hope you're right, but I don't see how they could have faster (or as fast) RT tech than nVidia, with a reasonable budget silicon / price. Or they have some magical patents out there...

We'll see...
 
Why not? Why couldn't they, for example, license PVR's tech? Why can't AMD have some similar, suitable solution? I don't see anything magical about NV's solution that precludes anyone but them being able to accelerate raytracing.
 
Sure, but my point with the "reset" was about the appearance of 4Pro and now the impending generation where a majority of users are still on the base SKU, so presumably most are simply waiting for a bigger leap than the tic-toc update (always been ticked about Intel's definition but maybe I'm forgetting some explanatory talk about why the tic isn't the first iteration followed by toc, then starting on a new gen with the next tic and awerjawoifadfl kaf :V ) even if it's conceivable that there can still be a lot of downports owing to the more stable architecture.

Understood, and it'll be interesting to see how the PS5 transition goes. If Sony are keen to support PS (I assume they are, it's their cash cow and will continue to be) and the lines between PS4/Pro/PS5 enhanced PS4 titles are blurred, will there be a meaningful reset at all? Maybe that will happen only when bona fide PS5-only titles ship in number. Annually iterating platforms like PC and mobile don't reset, they just roll on until support for older software/hardware fades out organically.

Folks are perhaps not a little conditioned already by the phone cycles, sticking to a similar pattern of purchasing with "every-other" HW update - be it always the new numbered generation (the bigger shift) or always the update SKU ("S").

I don't know what to extent the tech industry's tick-tock cycles drive purchasing decisions. Intel's clock is long broken and with mobile it's often about the length of your contract. I remember when the norm in the UK was 12 month cell contracts, then it was 24 months and 36 month contracts are gaining traction. 12 or 24 months can be iterative or truly significant with genuinely new and interesting hardware but customers are still limited by their contracts. Consoles remain free from this, for the most part - who knows where services like Game Pass will go!
 
I hope you're right, but I don't see how they could have faster (or as fast) RT tech than nVidia, with a reasonable budget silicon / price. Or they have some magical patents out there...

We'll see...
As I understand it, what nvidia innovated for RT acceleration in their implementation was the use of ML /tensor for denoising. I believe the other items have been completed by other companies before nvidia like PVR solution.
 
Annually iterating platforms like PC and mobile don't reset, they just roll on until support for older software/hardware fades out organically.
Key points being that they annually iterate, and as a result abstract the hardware to some disadvantage. Apple gets away with it by using closed hardware and same core architecture - they haven't changed CPU ISA four times in four products. PC gets away with it by swallowing up the overheads with brute-forced power, and relentless driver updates. Android only gets away with it half-heartedly - it's a crap-shoot whether your handset is compatible with a game or not. Highly abstracted, low performance engines like Unity take care of that while leaving half your performance on the table.

Consoles still occupy a unique position and I don't think their audience wants that to go away. In a world where everything is updated all the time, consoles offer a stability and uniformity that brings with it a sense of excitement. If you buy a TV, you know it'll be outdated in two years. Same with your phone. Same with your any-other-gadget. When you buy a new console, you know it's going to be a huge improvement on what you had before and you know you won't have to worry about it being replaced in a few years.

A swap to an generationless platform may work as a business, but it will also mean the end to hardware excitement. Just as phone updates are passé now, with phone companies cramming in all sorts of irrelevant gimmicks to try and make them exciting, there'll be nothing new and worthwhile to excite people into upgrading. It'll just be a slow, gradual transition, which for the hardware companies means no longer being able to capitalise on the trustworthy hardware cycle and its opportunities for starting afresh, drawing new crowds, generating buzz, etc.
 
As I understand it, what nvidia innovated for RT acceleration in their implementation was the use of ML /tensor for denoising. I believe the other items have been completed by other companies before nvidia like PVR solution.
And yet that hasn't been used in any gaming implementations yet, has it?
 
And yet that hasn't been used in any gaming implementations yet, has it?
This part I’m unsure of, at first I thought this way too but I am no longer sure. @Dictator might have a better idea here. There could be things happening at a driver level I’m unsure of. I recall seeing bar graphs about how tensor flops were being included in the calculations on toms hardware. I dunno but hopefully this is cleared up soon.

But thinking out loud there may exist specific RTX extensions that devs can call on for their denoising as opposed to building and offering their own denoiser function.
 
Last edited:
Key points being that they annually iterate, and as a result abstract the hardware to some disadvantage.
There isn't a whole lot between the low metal APIs of GNM, Metal and DirectX and consoles are moving into that area where, if you want today's games to run on tomorrow's consoles, you need to fatten up the abstraction. Microsoft do this with VMs and low level APIs, we don't quite know how Sony will run PS4 games on PS5 but I strongly believe that PS5 will, from a developers point of view, just look like PS4 Pro+ Extreme Turbo, with some extra APIs for PS5-specific hardware.

Consoles still occupy a unique position and I don't think their audience wants that to go away.

Part of that audience yes, but not that segment that are buying PS4 Pro and Xbox One. Mid gen consoles have demonstrated you can accommodate everybody.

In a world where everything is updated all the time, consoles offer a stability and uniformity that brings with it a sense of excitement. If you buy a TV, you know it'll be outdated in two years. Same with your phone. Same with your any-other-gadget. When you buy a new console, you know it's going to be a huge improvement on what you had before and you know you won't have to worry about it being replaced in a few years.

Equally, most console owners know that with march of technology and an unchangeable hardware platform, you carry the risk that later software performance may go down the drown. Many end-of-generation PS2 and 360/PS3 games were too technically ambitious and fell back to low resolutions and/or poor frame rates.

A swap to an generationless platform may work as a business, but it will also mean the end to hardware excitement.Just as phone updates are passé now, with phone companies cramming in all sorts of irrelevant gimmicks to try and make them exciting, there'll be nothing new and worthwhile to excite people into upgrading.

Phones are a poor analogy as the emphasis on new phones are mostly utilitarian. You want your phone to feel good, run fast, last all day and and have decent features and these are all things smart phones have done for many years. Cramming in more cores and RAM (or as you say, irrelevant gimmicks) really isn't doing much for your average user. Consoles are driven by better gaming experiences which are directly benefited by more cores and more RAM and gaming technologies continue to race forward.

It'll just be a slow, gradual transition, which for the hardware companies means no longer being able to capitalise on the trustworthy hardware cycle and its opportunities for starting afresh, drawing new crowds, generating buzz, etc.

There was plenty of excitement over PS4 Pro and Xbox One X and even little bumps like Xbox One S and even the All Digital Xbox One. New products will create interest even if you're not looking at anything revolutionary. People who bought launch machines and wanted a better experience upgraded. People who didn't have launch machines and bought into this generation after the mid-ten consoles launched had a decent choice of cost/performance profiles, just like every other product sector out there.
 
There was plenty of excitement over PS4 Pro and Xbox One X...
They were first-time ever mid-gen refreshes - of course there was going to be excitement! In terms of sales, they haven't set the world alight. When annual refreshes are common-place, excitement will drop off a cliff. It's just not possible to be as excited for an annual PlayStation - the time between is far less and the level of advancement will be far smaller. It's like having a small birthday or Christmas every month instead of one large one a year; it just can't hold the same value or interest or command the same levels of enthusiasm.
...and even little bumps like Xbox One S and even the All Digital Xbox One.
And PS2 slim. But nothing on the level of a whole new console with whole new experiences.
New products will create interest
But not in quantity. "Oh, the next PlayStation is out. What's different? Oh, it's faster. It also has a new tech that no games will use for the next three years until the install base is large enough to warrant its utilisation."

It's that latter point that really sets consoles apart; actually kick-starting adoption of a GPU tech because devs are guaranteed an audience large enough to make it worthwhile.

...just like every other product sector out there.
No other product sector out there runs alongside next-gen content that takes 3+ years to make. Bring in raytracing to a generic platform and it's three years before those RT-designed games come out. Or never, in the case of PVR. Introduce RT to a console, know you're bringing it, get your first parties creating games to use it and bam, day one, a whole new world of software. That's what makes consoles exciting. Everything they've ever done has pretty much been doable on PC before hand, but it's the consoles that make it because they offer these complete shifts, hardware and software.
 
Status
Not open for further replies.
Back
Top