*spin-off* Length of the generation & HW... things

[iroboto]

Spencer on twitter said DX12 wont be a massive change

Man, not going to lie, this is pretty disappointing to me. Don't care that x1 is weak, but we are going into a 5+ year console cycle with technology that was released and designed in 2009; granted not much of it was used because of MS locking APIs/features to OS releases, but still sigh. 11 brought a lot of awesome, was hoping more on that front.

 

[Kaotik]

Man, not going to lie, this is pretty disappointing to me. Don't care that x1 is weak, but we are going into a 5+ year console cycle with technology that was released and designed in 2009; granted not much of it was used because of MS locking APIs/features to OS releases, but still sigh. 11 brought a lot of awesome, was hoping more on that front.

I don't think this console gen will be 5+ year one, the hardware is just too weak relative to rest of the industry (the new consoles have never been this weak in comparison before)

 

[iroboto]

I don't think this console gen will be 5+ year one, the hardware is just too weak relative to rest of the industry (the new consoles have never been this weak in comparison before)

But it can't be too short either. The cross gen games will be coming out all the way into 2015; it would be an extremely short cycle and with hardware ie Morpheus still coming out that cycle would be extremely short if it wasn't at least 7 years IMO.

 

[Kaotik]

But it can't be too short either. The cross gen games will be coming out all the way into 2015; it would be an extremely short cycle and with hardware ie Morpheus still coming out that cycle would be extremely short if it wasn't at least 7 years IMO.

Nothing would prevent Morpheus and the like being compatible with the next consoles too

 

[iroboto]

Nothing would prevent Morpheus and the like being compatible with the next consoles too

Yes. I need to think out of the box more often.

 

[blakjedi]

Considering its all PC hardware essentially (with the exception of the ESRAM)

BC should be a given in the future...hell forward compatibility should be a given pretty much too...

 

[iroboto]

Considering its all PC hardware essentially (with the exception of the ESRAM)

BC should be a given in the future...hell forward compatibility should be a given pretty much too...

Lol I hope they make the kinect 2.0 today forward compatible if it is a short life cycle then lol.

Is it wrong for me to be under the impression that this gen is moving towards forward+ rendering? It seems like it fits the bill for both consoles.

 

[Deleted member 11852]

I don't think this console gen will be 5+ year one, the hardware is just too weak relative to rest of the industry (the new consoles have never been this weak in comparison before)

I doubt it'll be less than five years. Before they make a single cent of profit (in real terms) they have to recoup every dollar spent on R&D for this generation. First party games often as not don't make much profit, if any profit at all.

How capable the consoles are in technical terms, nor how they compare to the PC, really aren't a factor. And if you look at the hardware distribution of Steam users, it's clear that in many ways the consoles compare well to the greater majority of PCs. E.g. less than 1 in 5 people have more than 2Gb VRAM, 1 out of 4 people are gaming on a 1366x768 display (a typical laptop resolution). Half the folks out there only have a dual core CPU (Core2Duo, i3, i5 etc). More than half of people have 4Gb of less RAM. Not exactly great hardware for gaming based on the minimum/recommended specs appearing on more recent games.

 

[zupallinere]

I doubt it'll be less than five years. Before they make a single cent of profit (in real terms) they have to recoup every dollar spent on R&D for this generation. First party games often as not don't make much profit, if any profit at all.

.

Well if the next generation of consoles is built upon the work of this generation then you could think of the next Gen being substantially paid for r and d wise. Big if of course

 

[Deleted member 11852]

Well if the next generation of consoles is built upon the work of this generation then you could think of the next Gen being substantially paid for r and d wise. Big if of course

Yup, very big if. But if any PlayStation architecture looks to make the leap into the following generation and stand any chance of architectural backwards compatibility, I'd say an 80x86 processor and AMD Radeon GPU without EDRAM/ESRAM, is probably it.

However I think that ARM could well be a serious contender next time.

 

[AlNom]

Yup, very big if. But if any PlayStation architecture looks to make the leap into the following generation and stand any chance of architectural backwards compatibility, I'd say an 80x86 processor and AMD Radeon GPU without EDRAM/ESRAM, is probably it.

However I think that ARM could well be a serious contender next time.

I wonder if a big.LITTLE approach would be feasible with AMD cores instead, (or if it would make any sense to do*). On whatever future node is used, the jaguar cores are going to be elfin-sized, so keep those around for a low power mode and also backward compatibility. The big cores can be whatever new design AMD has.




*most likely pointless

 

[zupallinere]

Yup, very big if. But if any PlayStation architecture looks to make the leap into the following generation and stand any chance of architectural backwards compatibility, I'd say an 80x86 processor and AMD Radeon GPU without EDRAM/ESRAM, is probably it.

However I think that ARM could well be a serious contender next time.

Arm could be the future sure. What would it take to get BC with arm though and if that is not a consideration then would not having BC next gen be more of a problem now than before?

Still an armed future using what was learned this gen could be interesting. If hsa is promising what would that look like on an arm soc or at least huma.

 

[Deleted member 11852]

I wonder if a big.LITTLE approach would be feasible with AMD cores instead, (or if it would make any sense to do). On whatever future node is used, the jaguar cores are going to be elfin-sized, so keep those around for a low power mode and also backward compatibility. The big cores can be whatever new design AMD has.

I must admit, I'm not 100% sold on big.LITTLE. Rather than having a mix of low-power and high-performance cores (and I think they're focused on battery-powered mobile devices), I'd rather a bunch of high-performance cores with better low-power modes. The power profiles of the low-power ARM cores look to be below what would probably be useful in a console - I'm not sure a switched-on console is ever that idle routinely and I think it's more akin to performing the kind of low loads that the embedded ARM core in PlayStation 4 is already doing when the PS4 is in standby.

Of course if you throw AMD cores into the mix well, hmm. I guess it depends on how many low-power cores, whether they could meaningfully contribute to performance and their activity on the bus doesn't hinder more than help. Plus as long as they are fully-fledged cores, i.e. not cut down ones requiring extra effort by the dev to support. Oh and that the transistor count couldn't just be better used for something else.

Other than that, I'm in

 

[3dilettante]

Yup, very big if. But if any PlayStation architecture looks to make the leap into the following generation and stand any chance of architectural backwards compatibility, I'd say an 80x86 processor and AMD Radeon GPU without EDRAM/ESRAM, is probably it.

However I think that ARM could well be a serious contender next time.

ARM the architectural licensing company would not be, or rather a licensee of ARM with a specific implementation would be.
ARM's next round of cores should be more performant, but there still seems to be an inherent lag in time and performance between what the standard cores deliver versus what a custom design can give.

The A57, once it really rolls out, might offer performance somewhere in the range of AMD's Puma cores, which are themselves just a final polish on Jaguar.


That's not enough to emulate another ISA, and in the time frames in question the ARM architectural licensee with the most experience with the current gen's configuration is going to be AMD--should its corporate fortunes be less depressing in five years.

Arm could be the future sure. What would it take to get BC with arm though and if that is not a consideration then would not having BC next gen be more of a problem now than before?

The standard ARM cores do not significantly outpace more targeted and optimized designs in the same time frame.
Because the standard cores and system architectures must serve many disparate customers, a good fraction of this lag and suboptimality is necessary.

While Jaguar doesn't have a raw clock advantage versus the next few generations of ARM like Xenon and Cell had over Jaguar, it becomes uncertain that ARM has any desire to push its standard cores that far beyond what the console chips clock at.
While it is possible a future standard ARM can be significantly more capable, it may not be enough to make it able to emulate 8 Jaguar cores sufficiently.

A custom ARM core from an architectural licensee might have better luck, but then it goes back to AMD being the most likely candidate if its plans work out.
AMD's stated goal is to have a more fungible set of core IP that can create ARM and x86 sibling cores with implementations that push them over a prepackaged core.

There could be a difference in overall power consumption and possibly performance between the ARM and x86 siblings, but it's going to be a secondary effect.
At that point, if BC is a design parameter, why go with an ARM core unless AMD falls apart (not impossible, granted)?

CPUs aside, compatibility with the specific APU bus structure is likely another sticking point.
It's not something many others have chosen to do, and isn't strictly necessary.
If you want to make a compatible successor to an APU, it would help immensely.

 

[Cyan]

I don't think this console gen will be 5+ year one, the hardware is just too weak relative to rest of the industry (the new consoles have never been this weak in comparison before)

5 years at the most for me, 'cos this generation is the least inspired in years. Nintendo WiiU has not proven that it was an amazing idea, like the Wii. Microsoft created an okay all around console but nothing revolutionary. Sony just created a slightly more powerful console than those two, and that's it.

 

[dobwal]

A generation will last however long the market will allow. And performance seems to have little to do with that circumstance.

The PS2 still outsold the Xbox 360 pretty readily month to month in the US for the first year even though the 360 has a pretty good 1st year lineup. And then the Wii and its last gen hardware dominated the next two years. The PS3 and 360 chugged along until late 2013 even though 2 TF gpus start showing up around 2009.

 

[AlNom]

I must admit, I'm not 100% sold on big.LITTLE. Rather than having a mix of low-power and high-performance cores (and I think they're focused on battery-powered mobile devices), I'd rather a bunch of high-performance cores with better low-power modes. The power profiles of the low-power ARM cores look to be below what would probably be useful in a console - I'm not sure a switched-on console is ever that idle routinely and I think it's more akin to performing the kind of low loads that the embedded ARM core in PlayStation 4 is already doing when the PS4 is in standby.

True, I was more thinking of the "easy-BC" path more than anything. The power consumption is just a side-bonus, but even so if that were a path to take, they wouldn't have to include the ARM at all then (one less license).

Also, would it be feasible to have the OS reservation on said "little" cores instead of reserving the newer "big" ones Maybe it's a mess for the memory bus. :s They do already have some funky stuff going on with multiple OS's though.

edit:

Nevermind

 

[Deleted member 11852]

ARM's next round of cores should be more performant, but there still seems to be an inherent lag in time and performance between what the standard cores deliver versus what a custom design can give.

A lot of people are using custom ARM designs offering performance way beyond what consumers are used to seeing in commercial applications like phones and tablets. ARM is becoming more prevalent in closed-box solutions in the aerospace, defence and security industries. Radhardened ARM cores are really driving out the x86 chips and in mission critical packages.

I couldn't imagine for a second that an ARM core in a games console would be anything other than heavily customised. While I commented that PS4 is probably the most likely to be forward compatible with PS5, I don't think this is remotely a priority for Sony. So emulation of x86 wouldn't be a factor.[/QUOTE]

 

[3dilettante]

A lot of people are using custom ARM designs offering performance way beyond what consumers are used to seeing in commercial applications like phones and tablets. ARM is becoming more prevalent in closed-box solutions in the aerospace, defence and security industries. Radhardened ARM cores are really driving out the x86 chips and in mission critical packages.

That goes to the distinction between ARM the platform licensing company and the implementors.
Sony or Microsoft wouldn't go to ARM to design a custom core, they'd look into making or contracting a customized core that uses the ARM ISA.
The number of such implementors is limited, and the number of such implementors with all the additional IP and architectural history that overlaps with what consoles (as we know them) is very limited.

If AMD is successful in what its marketing has promised, there is going to be a higher-performance custom ARM core in the range that a console would like, and it's going to plug into an SOC with the infrastructure with the highest chance of matching Onion and Garlic.
That core is going to have an x86 sibling that would be generally equivalent, and have BC as well as inheriting much of the established tools.

 

[McHuj]

I think it can go either way with how long this generation is. I think it will depend on what the Xbox does and what MS decides to do. If Xbox continues to get outsold significantly and MS wants to remain in the console space, I could see them launch a successor sooner (2018) and that in turn would probably drive Sony to respond in the same time frame. I think if it were up to just Sony they wouldn't mind having a longer gen.

The wild card will be (don't laugh) Nintendo. I predict they'll release a new console in late 2016. The question is will it be X1 power level + a gimmick (likely) or a leap beyond the current gen (not likely)? Does that constitute a new gen? They're kind of in a weird place and I wonder if they are entertaining the possibility of going with x86? But they're more likely to go with ARM all around.

The other two I think they'll stick with x86., Given how the power consumption of a console is dominated by the GPU, I doubt the power savings (if any) of ARM offer a compelling reason to switch.


This thread seems to be veering into "Predict the next-gen consoles" territory.