Predict: The Next Generation Console Tech

[east of eastside]

PPC A2 is maybe or probably tuned for lower frequencies, as an Atom or E350 is, so it might not be able to reach > 3GHz.

Yeah, I noticed that..Okay, is there a more appropriate PPC core then?

I guess there is a PPCA2 version that runs at 2.3ghz @ 45nm. Perhaps a 28nm might cross 3ghz?

 

[Brimstone]

If the CPU performance is no longer key in getting overall performance up, then the choice of which cpu becomes less important. This being the case, the investments by all involved in Cell would lead me to believe this would be the best architecture to choose.

I think it's a safe bet Sony is going with a 4k Blu-Ray format with the PS4. The H.265 codec will require processing muscle. A CELL v2.0 CPU with plenty of SPE seems likely.

 

[RedVi]

So you're saying they learnt nothing from this gen? 4k BD drives will delay the console and make it much more expensive while still only allowing for the same cost to be spent on chips (or less), Cell v2.0 will be pretty useless once again for multi-platform developed titles. All I can say if they go for Cell again is that they beef up the PPU's and only double (or something) the amount of SPE's. Say 4 PPU's with OoOE and more cache and 12-16 SPE's. MS will likely go with 4-6 of those same PPU's, if sony were to think one or two was enough, they'd end up at a severe disadvantage.

 

[Brad Grenz]

4K BD drives are just regular BD drives with software support for a new codec (h.265) and/or additional readable layers. In neither case can the cost be compared to the first mass-produced blue laser diodes the PS3 was using... And Cell 2.0 would be useful in multiplatform games for the same things Cell has been this gen: physics, sound, post processing effects including AA. The codebase is already there. So in fact, what we're talking about is a PS4 with all the advantages the PS3 had, and none of the disadvantages.

 

[upnorthsox]

Monster indeed!

That would be rather large and potentially cut down on the GPU budget in a significant way (either that or bring the BOM up a good chunk leading to either less profit down the road or higher MSRP).

If having a 4 PPE 6 SPE Cell would prove problematic, I'd think it would be easier for them to literally do a pair of Cell twins.

Having said that, I can't think of a reason they couldn't do a 4PPE 6SPE Cell.

1 PPE is attached to the 6 SPE's just as it is now in Cell, and the trio of PPE's sits next to the 1st PPE. Have a large shared cache for the 4 PPE's and it's a wrap.

I could be wrong on the above, but from what I've seen in other designs it seems quite possible.

I don't think so, but it depends on L3 edram. A trio of mini-Cells at 32nm would be a little bigger than the current 45nm cell and running in the sub 40W range. That'd be with the standard 1MB per thread of L3 (12MB in total) for the cores. They could always cut that out and go a little smaller or go the opposite and add 1MB per spe for a total of 24MB of L3 edram. It'd depend if they wanted to emphasis the spe's or not, but a cpu slightly larger than the current one is not some kind of budget buster.

 

[TheChefO]

Aside from the 6 SPE for compatibility, how many SPE's do the 4 PPE's include? Or are the VMX units taking their place? Just want to understand this concept.. And does this still fit under the cost, die size, and thermal requirements for a GPU dominant console?

Aside from the 6? None. The 6 SPE's would primarily be there for BC sake, while also providing for a number crunching workforce going forward.

The VMX units in the PPE's would be for additional processing power and also for those that have an issue for one reason or another in tapping SPE's.

Essentially allowing for easy porting between xb720 and ps4.

The size of such a chip would be roughly double what a 28nm PS3 Cell would be (roughly 120mm2).
The TDP would be well manageable, roughly the same powerdraw as a current 40nm Cell in PS3.


In conclusion, such a design would be exactly tuned to what would be required for a GPU-centric design.

 

[TheChefO]

So in fact, what we're talking about is a PS4 with all the advantages the PS3 had, and none of the disadvantages.

This is exactly where I'm looking at this from as well.

Even if we took the exact same Cell as in PS3 and used it in PS4, but bumped the GPU up, there would be significant upswing as the Cell is no longer being forced as a GPU assistant, and as we've said, the code libraries have been built for utilizing Cell.

The familiarity of Devs with Cell is there now. They have the codebases to sample from. They have the best usecases to work with. And with Cell not being tied down to graphics work, it can be free to work on other aspects of the game worlds.

It's truly a win-win.

 

[TheChefO]

I don't think so, but it depends on L3 edram. A trio of mini-Cells at 32nm would be a little bigger than the current 45nm cell and running in the sub 40W range. That'd be with the standard 1MB per thread of L3 (12MB in total) for the cores. They could always cut that out and go a little smaller or go the opposite and add 1MB per spe for a total of 24MB of L3 edram. It'd depend if they wanted to emphasis the spe's or not, but a cpu slightly larger than the current one is not some kind of budget buster.

You're right, it wouldn't be that large.

4 PPE, 12 SPE would be in the neighborhood of ~160mm2.

There would have to be a bit of code rewriting though for existing software to work in this model of 4spe/1ppe.

 

[Xenus]

Not really from what I gather most code running on Cell runs on a job dispatch system. So it really wouldn't care so much how the SPE's are configured just if one is free or not. There maybe some things that need redone but on a whole it doesn't seem like that much.

 

[Shifty Geezer]

And Cell 2.0 would be useful in multiplatform games for the same things Cell has been this gen: physics, sound, post processing effects including AA. The codebase is already there. So in fact, what we're talking about is a PS4 with all the advantages the PS3 had, and none of the disadvantages.

I wouldn't look at it like that. The majority of devs didn't get on with Cell this gen, and haven't mastered it nor become fluent with it. Code portability is next to non-existent in real terms from title to title, so you'll have to rewrite code for whatever processor is in PS4. And as a lot of the code will likely be shifted to GPGPU work (even if PS4 had Cell, cross-system development would mean GPGPU focus for XB3 and PC codebases), what you save will likely be negligable.

I could see SPEs being included for BC and maybe for some standard libraries. eg. A Sony engineered audio library to handle all audio seems like it'd be very valuable given bkilian's revealing post on audio consumption. Otherwise the silicon would go unused by most developers, and add cost and complexity to the system design. Sony would be better off with a hardware paltform that'll be easy for them to support their software layers on for the grand Unified Software Experience. Rather than rewriting browsers for every single device, they should have a common hardware aspect (bunch of compiler flags, rather than completely different code bases) that makes the same code very portable and maintainable. Otherwise, keeping all that digital content active across all devices will be a PITA.

The only exception I can see to that is if they create a true Cell 2 that'll see wider adoption than Cell did, which is extremely unlikely. Broadcom is making tiny processors to power media that serve the major functions of Cell's intended power consumers, hence Cell never made it into TVs or STBs or anywhere else. That'd be the same for Cell 2 unless it's very different in some way I can't imagine. Without a true Cell 2, putting 6-8 SPEs on the PS4 CPU would be a difficult choice to make as there's no obvious favour for either side.

 

[Xenus]

That is what I hate about what the console domain is becoming vs what it used to be. You shouldn't design consoles to be similar for the sake of portability you should design consoles for the best performance given certain factors. If they end up similar so be it but don't design similar consoles for the sake similarity.

 

[Teasy]

Unfortunately its becoming a necessity with the cost of game development increasing so much.

 

[Shifty Geezer]

But that era is over I believe. Back in the day, the differences in your hardware made for different experiences. There was reason for them. Now that graphics have consolidated around the DX shader model format, there's nothing to be gained by doing things differently. Unless there's a breakaway alternative like a raytracing console or a voxel console, with custom hardware to enable stuff not doable on other devices, and providing a marketing edge to make the alternative technology workable, there's no real point to it. The future is the selection of games and experiences. It's not as interesting as it used to be, but it's not necessarily bad either. I mean, PCs really are dull compared to the old 8-bit and 16-bit computers + consoles. They're just generic parts mixed-and-matched under the same OS. But what I can do on my PC now is incredible! It's so much more useable as a creative tool for life that it doesn't need to be a hobby in itself. The nerd in me would like to see bizarro hardware the likes of PS2 that would never be seen outside a console but which enabled some amazing things, but the whole-other-human-being in me will just be happy to see good games and smart software, and hope simplified hardware will bring stability and serviceability to the whole experience.

 

[MrFox]

For sony, 4K comes for free.

The old Cell can already decode two 1080p streams at 40Gbps total. They only need twice that performance to do 4k in H264 (or quad-HD). The drive costs practically nothing today, a firmware update can make the drive read 128GB BDXL discs which is plenty for a full length film at over 100 Gbps (they could spec to 144Gbps mux, for a 4x drive requirement). Any GPU can display quad-HD. HDMI can output quad-HD too. It all comes for free, they'd be stupid not to implement it.

But that's an unimportant feature. I want them to make a CPU/GPGPU combination with the memory bandwidth necessary to have effortless global illumination, and traced soft-shadows. Anything less and they might hold on to the PS3/360 because doubling the CPU and GPU is exactly what Nintendo did with the Wii. I can't see the difference between gamecube and Wii. 2x isn't enough of a jump to start a new generation.

 

[east of eastside]

I don't think backwards compatibility is worth the investment to support it. Conversely, Sony is forgoing the opportunity to sell Uncharted Collection in 1080p for PS4. Loss of compatibility with PSN titles would be more of a concern to me.

 

[Xenus]

Indeed the nerd in me cries with these latest years. At least the last gen Cell and Xenos were interesting. This gen all the predictions are far less so. I miss the days when exotic tech was used not because it was exotic but it was a legitimately better option at least in some cases for the budget at the time. Hopefully Ray tracing will be there next gen or the gen after if only for the sake of differentiation. Building a computer for a set price range is just so boring technically. I do that all the time when I build my own rig.

I want the original Cell B.E. I want Larrabee. I want XDR2. I want something that makes them more interesting then a custom PC built for a set price range in a box.

 

[steveOrino]

Yeah, literally everything is "off the shelf" now in terms of hardware. No more crazy custom stuff like the PS2 E.E. The economics just do not make sense anymore. Not a bad thing just a sign of the times.

The only thing that I wonder about is how platform makers are going to differentiate (beyond choice of gimmicky motion control) when everything is more or less the same (ATI GPU/Power based CPU). Yeah there will be some variation in performance but everything will be designed around the lowest common denominator anyway.

 

[TheChefO]

Sony is forgoing the opportunity to sell Uncharted Collection in 1080p for PS4.

Sony is already showing a good bit of maturity and customer-centric thinking with their Vita/ps3 cross compatible games.

You don't need to buy the same game over for each device.

This is the way forward.

Milking old franchises for the "HD experience" has already happened. I'm 100% sure the vast majority of attempts to resell every ps3/xb360 game as a 1080p version will fail 9 times out of ten.

A true remake where higher res textures, new models, new shaders, etc are employed, sure, that could work. But that would work regardless of the ability to plug the old game in or not.

Did the compatibility of Halo1 really steal sales away from Halo Anniversary edition?

I don't think so.

Loss of compatibility with PSN titles would be more of a concern to me.

Indeed.

This is the main concern.

 

[upnorthsox]

I was thinking that Sony might forgo backwards compatibility (seeing as how it's switching from Nvidia to AMD, anyways?) and might go for something simpler and more efficient in the CPU department - like 4 or 6 PowerA2 cores, or other suitable Power core solution, and stay away from Cell entirely if there is significant cost, die and thermal budget gains to be gained by doing so?

Sony may forego BC but it won't be because of switching from NV to AMD and it certainly won't be to go with a cluster of A2 cores. It'll be because there's no compelling business case for it.

Things may not work this way in the real world, but I wouldn't doubt AMD would not want to have anything to do with Cell on several levels.. If this chip is an SoC then adding Cell is going to complicate it significantly. For one, how would manufacturing of a Cell based SoC work? Global Foundries has never touched the Cell processor.. Future die shrinks will require more complicated redesigns as well, owing to the complexity of Cell. Lastly, this might be my own interpretation, but I think as a matter of pride AMD wouldn't want to combine its tech with Cell, it's really contrary to its vision of heterogeneous and GPGPU computing.

Why would AMD care? Did they drop or refuse MS when they wanted to combine Xenon and Xenos on a SOC? GloFo is a non-issue, it'll be IBM and/or Toshiba as the fabs for PS4. From a fab standpoint, Cell is no harder than any other probable design and less complicated than most. I'm not sure how a heterogeneous cpu like Cell runs contrary to a vision of heterogeneous computing and if AMD would let their pride get in the way (of anything) then they wont get the contract.

In any case, it's my feeling that Cell is done in consoles if anyone wants to give that some thought and expand upon it..

At this time last year I was all but certain that Cell was done but the latest changes at Sony have brought a glimmer of hope. This is because its appearant demise has nothing to do with performance and everything to do with politics.

 

[east of eastside]

Why would AMD care? Did they drop or refuse MS when they wanted to combine Xenon and Xenos on a SOC? GloFo is a non-issue, it'll be IBM and/or Toshiba as the fabs for PS4. From a fab standpoint, Cell is no harder than any other probable design and less complicated than most. I'm not sure how a heterogeneous cpu like Cell runs contrary to a vision of heterogeneous computing and if AMD would let their pride get in the way (of anything) then they wont get the contract.

Doesn't it run smack in the face of GPGPU computing?

The entire SoC can be fabbed at GloFo if its more standard PPC cores in a simpler configuration coupled with an AMD gpu, perhaps, whereas something more custom (Cell) might have to be done at IBM's Fishkill. I could be wrong about this point, I'm just looking at what would maximize manufacturing efficiency - (Having the entire chip and all its cores fabbed at one location). I'm drawing GloFlo into this from something called Common Platform - a manufacturing alliance recently announced between GloFo and IBM.

http://www.commonplatform.com/

At this time last year I was all but certain that Cell was done but the latest changes at Sony have brought a glimmer of hope. This is because its appearant demise has nothing to do with performance and everything to do with politics.

Can you elaborate on this a bit more, please. I'm just not quite certain what you might be referring to..

From my viewpoint, I interpret Vita as a total abandonment of the Kutargi design philosophy...more standardized parts, ease of programming, and efficiency as the design priorities.