Predict: The Next Generation Console Tech

[Shifty Geezer]

While on the flip-side, the MS tools would then be available in principle to PS4 developers. But at that point, you have to wonder why have two consoles if they're basically the same thing?!

I wonder if next gen, Sony could be convinced to partner with MS, solving all these issues caused by competition that have messed them about so much? Much as I'd hate that, I think they'd both be much better off financially with one joint system and only to Nintendo to compete with. If there were only XB360 or PS3 this gen, MS or Sony's be doing very well assuming 100% of the HD gaming market.

 

[eastmen]

The question is maybe, why would they want to? They've got an acknowledged advantage in terms of development tools that would surely need a massive amount of re-engineering to work with Cell.

Not only that but they may choose to go with a smaller cpu again this gen putting more trasnistors or money towards a bigger gpu or more ram.

I'm not to sure if a huge chip is worth while .

The waternoose is about a 160m transistors. Doubling that would be 320m tranistors and would give you most likely more than cell type power. Trippling that would give you 480m transistors.

The cell is about 250m trasnistors. Doubling that to a 2x16 would be 500m trasnistors. I wouldn't what would be the better chip at that point ?

DOubling or trippling the waternoose would give you BC out of the box providing your gpu isn't totaly diffrent

 

[Sound_Card]

Hence my reason why their is no one but MS that not let MS use Cell. Though, my point is not exactly to suggest they will. I think MS is being very tight lipped on everything of course, but based on that old report of ex sun engineers working with a MS chip design team, I have reason to believe they are building a in house CPU to be fabbed out some place else. Interesting how things seem to be playing out from my point of view. Sony seems to be building emphasis on the CPU (Cell) and branching off that. While MS I would assume will be placing emphasis on the GPU. My take on MS's next CPU is that it's going to be designed to run cool, and lower power. Lot's of focus will be placed on the GPU, and DX11(or better).

 

[fellix]

Cell (the Power Element) and Xenon are common architectures, minus the few gazillions of VMX registers in the last one (compare).

Anyway, I think MS should stick to its current overall design and extrapolate it to what the technology for the next console will offer, price/performance wise.
If the eDRAM is to survive the test of the time, it should be sized in a way a featured back-buffer (720p*2xAA, or "naked" 1080p) to fit in there without tiling. Depth/Stencil tests per pixel would also go up.
System memory, as if to stay in UMA fashion, would benefit from what GDDR5+ can offer at the time, on the same 128-bit interface, and given the full differential signaling is implemented by then, the bandwidth should go up few-fold.
The GFX part seem to be the easiest one - double almost everything in there, or acquire a new and improved design (more DX11-ish, Compute Shaders and so on).
And finally, the CPU realm... do we need more threads in there?

 

[RudeCurve]

Sun's Rock chip architecture is Open Source. I'd like to see MS's next Xbox use a modified version of it. Of course reality says MS will likely use the current X360 CPU configuration and add more cores. Why fix something when it isn't broke?

 

[3dcgi]

Sun's Rock chip architecture is Open Source. I'd like to see MS's next Xbox use a modified version of it. Of course reality says MS will likely use the current X360 CPU configuration and add more cores. Why fix something when it isn't broke?

Why Rock? I like the architecture for what it's meant for, but open Source doesn't mean free so there must be a technical reason for using it.

 

[Sound_Card]

I was hoping they would adopt suns proximity concept.

 

[3dilettante]

Rock's been delayed for at least a year, though.

There's been nary a peep about the chip of late, and Sun promised some very different hardware implementations of scout threading and some form of transactional memory.

New means unverified, which might explain the delay, but one of the drawbacks to hardware transactional memory support is that it is rather complex thing to get right.
Sun's Rock drama might not be appealing to a console maker, though there is time to iron out issues (that aren't showstoppers Sun hasn't disclosed) by the time next-gen parts are available.

Rock's overall design favors more computationally heavy server loads than Niagra, but that's still far from consoles.
I'm not sure how some features, like a shared L1 instruction cache between 4 cores, matches up with the typical game.


Sun's proximity connect endeavors haven't born fruit as of yet, and I haven't heard if they are still working on it.

Things like thermal cycling and reliably keeping the contacts positioned for long periods would would have to be reliably engineered, and it doesn't appear that those problems have been solved.

 

[ban25]

I suppose if Sun's still around circa 2012, we'll find out.

 

[Squilliam]

Rock's been delayed for at least a year, though.

There's been nary a peep about the chip of late, and Sun promised some very different hardware implementations of scout threading and some form of transactional memory.

New means unverified, which might explain the delay, but one of the drawbacks to hardware transactional memory support is that it is rather complex thing to get right.
Sun's Rock drama might not be appealing to a console maker, though there is time to iron out issues (that aren't showstoppers Sun hasn't disclosed) by the time next-gen parts are available.

Rock's overall design favors more computationally heavy server loads than Niagra, but that's still far from consoles.
I'm not sure how some features, like a shared L1 instruction cache between 4 cores, matches up with the typical game.


Sun's proximity connect endeavors haven't born fruit as of yet, and I haven't heard if they are still working on it.

Things like thermal cycling and reliably keeping the contacts positioned for long periods would would have to be reliably engineered, and it doesn't appear that those problems have been solved.

So is this the likely direction Microsoft is heading? This is the first time I've really heard of Sun as much of a CPU designer. What kind of architecture is likely to pop out the other end from using this technology? Im sorry I just don't understand this direction as its the first I've really heard of it aside from hearing a while back that they were looking to design their own CPU.

 

[RudeCurve]

Sun's been designing server CPUs for decades in the form of the SPARC processors.

http://en.wikipedia.org/wiki/SPARC

 

[3dilettante]

So is this the likely direction Microsoft is heading?

I haven't seen any evidence of it. I was replying to some earlier speculation concerning Sun, and that the drama around Rock and Sun's viability as a CPU designer would work against it.

This is the first time I've really heard of Sun as much of a CPU designer. What kind of architecture is likely to pop out the other end from using this technology? Im sorry I just don't understand this direction as its the first I've really heard of it aside from hearing a while back that they were looking to design their own CPU.

Sun's CPU designs haven't pulled in the headlines all the time, but SPARC has been around for quite a while. It was one of the RISC processor lines, along with POWER and MIPS (ARM possibly in the future), that x86 has competed against.

http://en.wikipedia.org/wiki/SPARC

x86 has effectively defeated these processor families in the CPU market for anything above embedded and below mid-level servers.

Sun's primary niche has been shrinking and not in the direction of consoles.

 

[geeQ]

I think with PlayStation 3 Sony shifted direction from there previous console strategy... One that allowed them to create a 'platform' that could be used as a basis to evolve shared technologies and design philosophy to each new generation. Allowing Sony to strengthen there market share by creating a common hardware/software standard, offer true backwards compatibility/synergy between generations and ultimately save Sony money long term on development costs.

As many have said already I think PlayStation 4 will look very similar to PS3, at least from a hardware/software point of view. So we're talking:

Cell2 or similar based cpu,
Nvidia derived GPU
Blu Ray
2 pools of ram - 1 for cpu & 1 for gpu
Storage device (HD, SSD)
HDMI output, wireless, etc
Similar operating system/software infrastructure

The common 'platform' strategy also has other advantages, including providing developers with a common design/instruction set over each subsequent generation of PlayStation. I guess long term this could lead to lower development costs all round even though each PlayStation generation will feature more cores/ram etc.

Personally I would expect Sony to debut PlayStation 4 for around the £300 mark. I don't think the Wii will affect this strategy at all. Sony will still play the high technology game next round, but of course there expenses will be far less now that many of the technologies inside PS4 are already established one way or another inside PS3. That being said there will be no more 'reinventing the wheel' so to speak.
The one area that Sony will definitely be looking into changing though is of course the whole idea of how we interact with our games and this of course is purely because of Nintendo's success with the Wii.

 

[ban25]

Sony has said the same things in the past regarding technology investments. In the PS2 era, they showed presentations with "EE2" and "GS2" monikers. If they aren't happy with the roadmap for CELL, then they are likely to go another route. Given that SCE has new management in place, this is all the more likely.

As I've said in the past, the console market has typically been most successful with a defacto monopoly. In this sense, it might be wise for Sony to partner with Microsoft on the next generation of consoles, as the current three-way split is not very beneficial for anyone (except Nintendo).

 

[eastmen]

I think with PlayStation 3 Sony shifted direction from there previous console strategy... One that allowed them to create a 'platform' that could be used as a basis to evolve shared technologies and design philosophy to each new generation. Allowing Sony to strengthen there market share by creating a common hardware/software standard, offer true backwards compatibility/synergy between generations and ultimately save Sony money long term on development costs.

As many have said already I think PlayStation 4 will look very similar to PS3, at least from a hardware/software point of view. So we're talking:

Cell2 or similar based cpu,
Nvidia derived GPU
Blu Ray
2 pools of ram - 1 for cpu & 1 for gpu
Storage device (HD, SSD)
HDMI output, wireless, etc
Similar operating system/software infrastructure

The common 'platform' strategy also has other advantages, including providing developers with a common design/instruction set over each subsequent generation of PlayStation. I guess long term this could lead to lower development costs all round even though each PlayStation generation will feature more cores/ram etc.

Personally I would expect Sony to debut PlayStation 4 for around the £300 mark. I don't think the Wii will affect this strategy at all. Sony will still play the high technology game next round, but of course there expenses will be far less now that many of the technologies inside PS4 are already established one way or another inside PS3. That being said there will be no more 'reinventing the wheel' so to speak.
The one area that Sony will definitely be looking into changing though is of course the whole idea of how we interact with our games and this of course is purely because of Nintendo's success with the Wii.

I see them putting less efforts into the cpu next gen. I bet it will just be a 2x18 cell chip with some improvements into each of the cores perhaps moer local store for the ppus .

That would put them at around 500m tranistors for cell 2 on a 32nm process mabye?

They would def go big on the gpu and I think they would switch to simply one large pool of ram.

Going with two pools will add to the cost

 

[pc999]

So is this the likely direction Microsoft is heading?

None can tell, probably not even MS.

A the present time there is to many variables, many of the m not even tech related. To point a few, economics, overall strategy (more Wii, XB360 or more PS3 like) the degree of sucess of some implementations/interfaces and their current R&D status, the deals they can make, will many people want small and cute or big and powerfull, what devs want (a new paradigm or use what they already know)... This in a ever evolving context (eg AMD going fabless, Intel making GPUs...)

There are few from a tech POV too, eg, does a 9 core Xenon+ a good choice? That already stars to seem like a Larrabe... What kind of GPGPU performance can you extract from a DX11 GPU and is it better than a CPU for things like physics animation...? If yes then why invest in a CPU like Xenon.

Probably there is safer bets than others (Nintendo and MS with ATI, Sony using a Cell legacy CPU...).

Personally I wouldnt be suprissed if Nintendo (low/mid-end) and/or MS (mid-high-end) presented us a Fusion based console, as they should be great performers in comparition to today consoles, easy to dev, "cheap and low power" and are naturally costum. Plus AMD is working on a few things Z-RAM and accelarators that could really be helpfull (voice recg and such)...

Anyway this isnt really a prediction just a possibly (safer?) bet, as Fusion could end a disaster, they could prefer have moreflexibility etc...

 

[Squilliam]

Thanks for your input on Sun and the SPARC architecture pc999 and 3dilettante.

Instead of looking at the CPU/GPU architecture I think perhaps it would be interesting if we can look at the memory and storage architecture, which is just as important as the other sub-systems.

Is it better to use a SSD and less ram so they can implement an Arcade type SKU, or use the current architecture and just implement more RAM to use as a cache?

 

[fehu]

I think tha fusion is overstated
when it will launch, and for some time, will simply be a lowend cpu glued to a lowend gpu, mainly for lowend notebook
how can be good on a next-console?

Maybe the wii2... but i'm sure that outside is full of custom better solutions

 

[one]

AMD showcased their implementation of server-side game processing as "Fusion Render Cloud."
http://www.bit-tech.net/news/2009/01/09/amd-demos-fusion-render-cloud/1

 

[Squilliam]

I don't believe the concept or idea of a much simpler architecture with the GPU and CPU implemented in the same die or on a multi-chip module is going to go away any time soon. It makes too much sense from a design and cost scaling perspective that it has to be considered as a strong contender for being the architecture of choice in the next generation.