Next Generation Hardware Speculation with a Technical Spin [2018]

[ultragpu]

No way in hell Sony's first party will be ready their native PS5 game for a 2019 launch, at least not from the big boys "GG, ND, SMS, SP". They will be crossgen titles sure, but them alone don't have enough incentive for people to upgrade. We need a game built ground up from the 1st party to really showcase what the system can do and I don't trust third party to be the ones to pull it off. If HZD2 is ready for launch at 2019 then it's a win, anything less is a bust, nor launching a <10tf weaksauce.

 

[iroboto]

No way in hell Sony's first party will be ready their native PS5 game for a 2019 launch, at least not from the big boys "GG, ND, SMS, SP". They will be crossgen titles sure, but them alone don't have enough incentive for people to upgrade. We need a game built ground up from the 1st party to really showcase what the system can do and I don't trust third party to be the ones to pull it off. If HZD2 is ready for launch at 2019 then it's a win, anything less is a bust, nor launching a <10tf weaksauce.

there might be some unnamed projects that are specifically built for it. I don't think it would be entirely abnormal. Xbox launched with Ryse son of Rome and that's largely considered still quite a beautiful game and it's a launch title that in no way could be done on the older 360s. While I do agree that you're unlikely to see sequels to existing IPs as being a launch title, I don't think that rules out new IPs necessarily.

I also don't see less than 10tf as being weak sauce. Not sure why'd you'd look at it that way.

 

[msia2k75]

there might be some unnamed projects that are specifically built for it. I don't think it would be entirely abnormal. Xbox launched with Ryse son of Rome and that's largely considered still quite a beautiful game and it's a launch title that in no way could be done on the older 360s. While I do agree that you're unlikely to see sequels to existing IPs as being a launch title, I don't think that rules out new IPs necessarily.

I also don't see less than 10tf as being weak sauce. Not sure why'd you'd look at it that way.

Many are simply too focused on the tf thing imo...

 

[anexanhume]

GDDR6 parts support a slightly reduced clock at 1.25v instead of 1.35v, so 11gbps with four chips would be sort of low power. Lpddr5 would be better.

But there's no way the ps4 SoC would have a usable wattage on 7nm, not for a portable console. It would need to be about 4 times less power than the 14nm slim?

I think it would still be 15-20W too high at 7nm, yup.

 

[phoenix_chipset]

No way in hell Sony's first party will be ready their native PS5 game for a 2019 launch, at least not from the big boys "GG, ND, SMS, SP". They will be crossgen titles sure, but them alone don't have enough incentive for people to upgrade. We need a game built ground up from the 1st party to really showcase what the system can do and I don't trust third party to be the ones to pull it off. If HZD2 is ready for launch at 2019 then it's a win, anything less is a bust, nor launching a <10tf weaksauce.

I mean killzone 3 launched in February 2011 and they were ready for ps4's launch. And... HZD launched in February last year Not convinced any first party (or insomniac) would ever make a cross gen game I mean when has that ever happened?

Naughty dog seemingly has 2 teams or at least they did on ps3, uncharted 3 was not done by the UC2 team. Not sure I care about what they make anymore, but from guerrilla a killzone in the vein of 2 on ps5 would be amazing. SP and SMS yeah they won't make launch.

I think anything over 8tf will be absolutely fine and the real generational leap will come from a vastly better cpu and ssd. Not saying I would be shocked to see 10tf but with something like temporal injection even 8 would produce an incredible looking game.

 

[Deleted member 11852]

Oh god yes, it's so stupid that we get the ps5 in 2019 esp. since how slow tech is advancing. To me the whole point of the pro is to stretch the gen out! Graphically there's still gas in the tank.

Sony said what the point of the Pro was: the reduce the bleed of owners switching to PC mid-gen as the tech in consoles ages. Like Gemüsepizza posted above, and just like the end of the 360/PS3 generation, we're seeing more and more ambitious games struggling to deliver a solid 30fps on the base 2013 consoles.

This is unlikely going to get better.

 

[London Geezer]

Sony said what the point of the Pro was: the reduce the bleed of owners switching to PC mid-gen as the tech in consoles ages. Like Gemüsepizza posted above, and just like the end of the 360/PS3 generation, we're seeing more and more ambitious games struggling to deliver a solid 30fps on the base 2013 consoles.

This is unlikely going to get better.

I mean, I want next gen as much as the next guy, but I'm looking at it from the perspective of us already having the X, at around 6 TFlop, producing great visuals but basically the same as a Pro, more or less.

And now we're talking about a 'next gen' in a year from now, maybe at 10TFlops or less. That's just not 'next gen' to me.

Think of the last game I completed - Spiderman. Now think how much better that would be on a PS5? Sure, it would be native 4K (we probably wouldn't see much of a difference there). The biggest 'glitch' with the game is the reflections, which can only really be fixed with ray-traced reflections, and we've discussed this to death - not happening on next gen consoles.

So we end up with a 'next gen' console whose main purpose is the same purpose of the current Pro (make games look better on 4K, but they're the same games, with broken reflections and improvements that are barely perceptible).

Sure, TFlops aren't everything, but they do give an idea of what we could expect.

Or maybe Debby Downer is back in the building and I should just wait and STFU.

 

[ultragpu]

there might be some unnamed projects that are specifically built for it. I don't think it would be entirely abnormal. Xbox launched with Ryse son of Rome and that's largely considered still quite a beautiful game and it's a launch title that in no way could be done on the older 360s. While I do agree that you're unlikely to see sequels to existing IPs as being a launch title, I don't think that rules out new IPs necessarily.

I also don't see less than 10tf as being weak sauce. Not sure why'd you'd look at it that way.

I admit Crytek could perhaps surprise us with something judging by Ryse's graphics. 10tf is weak if devs are intending to mainstream 4k native, at that point you'd hardly be seeing any significant visual leap over current gen. Temporal Injection or CBR are definitely useful but I doubt company would choose that faux K route when current midgen refresh are doing that already. I guess it kinda depends on how the PR spins it.

I mean killzone 3 launched in February 2011 and they were ready for ps4's launch. And... HZD launched in February last year Not convinced any first party (or insomniac) would ever make a cross gen game I mean when has that ever happened?

Naughty dog seemingly has 2 teams or at least they did on ps3, uncharted 3 was not done by the UC2 team. Not sure I care about what they make anymore, but from guerrilla a killzone in the vein of 2 on ps5 would be amazing. SP and SMS yeah they won't make launch.

I think anything over 8tf will be absolutely fine and the real generational leap will come from a vastly better cpu and ssd. Not saying I would be shocked to see 10tf but with something like temporal injection even 8 would produce an incredible looking game.

Would also be my dream!
Also what I mean was first party would have been busy making PS4 games thus if BC is confirmed then they become crossgen titles, not as in they're making crossgen titles by supporting legacy consoles.
Again 8tf is only good for high end PS4 level graphics at native 4k, even if you CBR it you still wouldn't have enough juice for more complex sims and lighting running altogether.

 

[Jay]

Temporal Injection or CBR are definitely useful but I doubt company would choose that faux K route when current midgen refresh are doing that already. I guess it kinda depends on how the PR spins it.

You've said it yourself, if going native didn't show any difference then they would go with cbr, temporal injection, dynamic res, all the above.
Those methods aren't exotic like they used to be, and are considered like any other part of the decision process when developing a game.
Those methods will also continue to improve.
Be sold as graphic / 4k mode.
Few people actually care if it's native as long as it looks good.

 

[iroboto]

So we end up with a 'next gen' console whose main purpose is the same purpose of the current Pro (make games look better on 4K, but they're the same games, with broken reflections and improvements that are barely perceptible).

Sure, TFlops aren't everything, but they do give an idea of what we could expect.

Or maybe Debby Downer is back in the building and I should just wait and STFU.

There in lies the issue with mid-gen refreshes. When we stop treating them like mid-gen refreshes, and expect the next gen console to be able to double that as well.
For everyone with a base model, they should, feel the jump. Anyone with a mid-gen refresh isn't going to relish this jump as much. Perhaps it's best to wait for the mid-gen refresh of next gen

 

[ultragpu]

You've said it yourself, if going native didn't show any difference then they would go with cbr, temporal injection, dynamic res, all the above.
Those methods aren't exotic like they used to be, and are considered like any other part of the decision process when developing a game.
Those methods will also continue to improve.
Be sold as graphic / 4k mode.
Few people actually care if it's native as long as it looks good.

I'm actually Team re-projection myself, you get more bang for the buck for sure. However, more power is always merrier and let's not give them any ideas to stop pushing the boundaries because we have a nifty little technique. A 12-14tf console doing 4k cbr is gonna truly blow minds so we really should be bargaining for more power anyway within an acceptable price range rather than be content so easily don't you think?

 

[Deleted member 13524]

But there's no way the ps4 SoC would have a usable wattage on 7nm, not for a portable console. It would need to be about 4 times less power than the 14nm slim?

As I mentioned before, I don't think the PS4 Slim is a fair starting point:

- Its SoC was probably developed with only cost savings in mind and not power reducion. TSMC claimed a 70% reduction in power consumption from 28HPM to 16FF+ at ISO performance but the actual difference at the wall between the Slim and Original barely reaches 45% at best. It seems they focused mostly on transistor density for making a smaller and cheaper SoC. There aren't any changes since the PS4 Slim doesn't even have updated HDMI 2.0 / HDCP2.2 outputs and H265 video decoders like the XBoneS, so it's probably the closest there is of an optical shrink.

- Measurements at the wall don't count with power loss from an 120/230V AC -> 12V DC rectifier with a typical efficiency of 80%. Battery powered devices don't need this.

- Spinning optical drive, spinning SATA HDD, etc.Compared to eMMC the difference is probably massive.

- Voltage regulators made for ~100W TDP waste a lot more energy than regulators for ~20W TDP.

- "Unlimited" radio power for Bluetooth and WiFi

 

[Megadrive1988]

I'm actually Team re-projection myself, you get more bang for the buck for sure. However, more power is always merrier and let's not give them any ideas to stop pushing the boundaries because we have a nifty little technique. A 12-14tf console doing 4k cbr is gonna truly blow minds so we really should be bargaining for more power anyway within an acceptable price range rather than be content so easily don't you think?

I completely agree with this. Yes you are right.

 

[turkey]

Could you use CB as a form of dynamic resolution, so you always hit '4k' but you may have to do more or less reconstruction depending on the load?

 

[VitaminB6]

I'm actually Team re-projection myself, you get more bang for the buck for sure. However, more power is always merrier and let's not give them any ideas to stop pushing the boundaries because we have a nifty little technique. A 12-14tf console doing 4k cbr is gonna truly blow minds so we really should be bargaining for more power anyway within an acceptable price range rather than be content so easily don't you think?

I agree with this as well. I'm sure with time this technique will only improve. and it's already producing great results. Also what happens when 8k TV's become affordable? I think they just recently became available although extremely expensive. Won't display resolutions eventually hit a point of diminishing returns unless you want to have a 500" TV? I've not seen an 8k TV in person but I can't imagine there's a huge mind blowing difference when compared to 4k on an average size display but I could be wrong.

 

[temesgen]

there might be some unnamed projects that are specifically built for it. I don't think it would be entirely abnormal. Xbox launched with Ryse son of Rome and that's largely considered still quite a beautiful game and it's a launch title that in no way could be done on the older 360s. While I do agree that you're unlikely to see sequels to existing IPs as being a launch title, I don't think that rules out new IPs necessarily.

I also don't see less than 10tf as being weak sauce. Not sure why'd you'd look at it that way.

I need to understand how those 10tf will be used first, simply 4k isn't good enough IMO. I get that developers can use system resources in the manner which seems best to them but I do think the overall philosophy that goes into the tools and hardware design matter.

 

[iroboto]

I need to understand how those 10tf will be used first, simply 4k isn't good enough IMO. I get that developers can use system resources in the manner which seems best to them but I do think the overall philosophy that goes into the tools and hardware design matter.

A lot of the horsepower when we moved from last gen to this gen was the move from 720p to 1080p.
At 250 GF vs 1.34TF, Xbox One only has 5.36x more power than the 360. When we don't factor in architectural/feature differences.
But just looking at resolutions 900p and 1080p, 1.6x and 2.25x more pixels over 720p respectively, the remaining horsepower has been there driving pixel quality.

So 10TF being 5.2x more than 1.8 TF for PS4... yea 6ish TF to get to 4K.
With 4 extra TF to amp up the quality, not counting any new features/architectural difference and any new algorithms.

To me this looks pretty decent. You get a strong CPU, and a UHD blu ray drive to support 100GB discs.

The issue is whether or not there are enough 4K owners out there to really make full use of the resolution at launch - that's a lot of power being used to downscale resolution.

 

[phoenix_chipset]

A lot of the horsepower when we moved from last gen to this gen was the move from 720p to 1080p.
At 250 GF vs 1.34TF, Xbox One only has 5.36x more power than the 360. When we don't factor in architectural/feature differences.
But just looking at resolutions 900p and 1080p, 1.6x and 2.25x more pixels over 720p respectively, the remaining horsepower has been there driving pixel quality.

So 10TF being 5.2x more than 1.8 TF for PS4... yea 6ish TF to get to 4K.
With 4 extra TF to amp up the quality, not counting any new features/architectural difference and any new algorithms.

To me this looks pretty decent. You get a strong CPU, and a UHD blu ray drive to support 100GB discs.

The issue is whether or not there are enough 4K owners out there to really make full use of the resolution at launch - that's a lot of power being used to downscale resolution.

360 is 204 gflops in practice and 240 on paper. So it's really like a 6x jump to xbone and 9 to ps4.

So it's safe to say we won't even see a ps4 > 360 jump on the graphics side so I view any graphics jump we get as just icing on top of good cpu's and some kind of ssd.

I said 8 before but honestly that's just the minimum so i'm not disappointed. Most likely we'll see 10-12 tf, I just don't count on it.

10tf to 12 with 1440-1600p temporal injection would look incredible no one should be disappointed

 

[Deleted member 11852]

I mean, I want next gen as much as the next guy, but I'm looking at it from the perspective of us already having the X, at around 6 TFlop, producing great visuals but basically the same as a Pro, more or less.

I don't wholly disagree but equally I am not responsible for writing code for consoles which do seem ridiculously CPU-limited which makes some console games all the more impressive. But we've had a very candid insights into how Sony R&D a new console with Mark Cerny's 'The Road of PS4' presentation so we know this the process is pretty much an unstoppable train once started and that process has definitely started.

You can change anything but not without a cost and Sony are still rebuilding from numerous disastrous decisions across several businesses a decade ago which has left the company a shadow of its former self. They really can't afford it.

 

[Metal_Spirit]

So 10TF being 5.2x more than 1.8 TF for PS4... yea 6ish TF to get to 4K.
With 4 extra TF to amp up the quality, not counting any new features/architectural difference and any new algorithms.

To me this looks pretty decent.

It might well be decent. But unless some ground breaking new tech is present, it would be a small jump than the one we had when passing from last generation to the current.
720p to 1080p is a 2,25x increase in resolution. 1080p to 4K is a 4x increase in resolution.
If we divide the Gflops/Tflops we had/have by the 720p/1080p, we can get the gain of flops per pixel we had on the generation jump.
To get the exact same jump from the current generation to a next one at 4K we would need more than 26 Tflops (just do the math). This will not be achieved for next generation. It would be too expensive.
But we can get by with a lot less. Using reconstruction tecniques like the PS4 Pro does, we can reduce the Tflops needs to about 13 Tflops.
With these 13 Tflops we could have all current gen games at native 4K, 60 fps (Due also to Zen), and a new generation, with exact the same gains over the current one that we had on the current over the previous, at checkerboard rendering, or other similar tech. Once again this is just math.
10 Tflops would be a gain alright. It would be suficient. But owners of mid generation consoles will shure be disapointed by it. The jump on those cases would be small on the GPU side.
According to Tweaktown, Navi 10 will be able to reach up to 30 Tflops. This seems to mean that a cheaper and with lower clocks version of it can easily get to about 13.
Without any ground breaking new tech that allows performance gains, thats about the performance I'm betting on.