Predict: The Next Generation Console Tech

[Rangers]

What I think is a far more interesting question, perhaps deserving it's own topic, is not what initial process the next gen consoles are on, but how the heck they cost reduce from there?

Lets say in 2011 one or more "next gen" consoles comes out at..what? 45nm? 32nm? Either way you've got precious few if any process shrinks left to reduce cost..

I think this could contribute to one or more manufactorers not shooting for the cutting edge technology...because I'm not sure what the cost reduction picture is from then out on silicon.

 

[Acert93]

Based on Sony's own definition of a 10 year lifecycle, the release of the PS4 in 2010-2012 is very likely. Especially if the release pattern of their other "10 year lifecycle" videogame products are considered, as well as general market forces and competition.

 

[liolio]

it's very likely that if Ms doesn't manage to convince Intel, all the actors (ms, sony,nintendo) will start at 45 as it will cheap and widely available.

This will also have an impact on silicon bugdet, transition from 32nm 22nm could be difficult.

 

[hoho]

I mean, if the jump from PS2 to PS3 was massive (from 266 megahertz...............to 3.2 gigahertz), then what more with the PS4 by year 2012. We should be able to reach 10 Ghz teritorry by then................while still maintaing max power consumption at not more than 300 watts.

33MHz in PS1 to 266MHz in PS2 is 8x jump in clock speed, from 266MHz to 3.2GHz it is 12x jump. Now assuming similar scaling we should have a 18x clock speed increase with PS4 giving us nearly 60GHz. Too bad real world isn't as good as predictions based on past events indicate

 

[Dominik D]

Based on Sony's own definition of a 10 year lifecycle, the release of the PS4 in 2010-2012 is very likely.

Do I have to remind you some other definitions they have already contradicted? They are still hoping for PS3 to profit. Announcing that PS3 will be butchered in 2 years would kill PS3 support immediately. You have to sell "10 years lifecycle" BS and do your work at the same time. IMO waiting till 2012 is not a profitable plan for Sony.

 

[Betanumerical]

Do I have to remind you some other definitions they have already contradicted? They are still hoping for PS3 to profit. Announcing that PS3 will be butchered in 2 years would kill PS3 support immediately. You have to sell "10 years lifecycle" BS and do your work at the same time. IMO waiting till 2012 is not a profitable plan for Sony.

Why not?, people still buy the PS2 don't they? and they are making a shite load of profit off that. I can see the same happening with the PS3.

 

[archangelmorph]

Why not?, people still buy the PS2 don't they? and they are making a shite load of profit off that. I can see the same happening with the PS3.

Exactly..

Killing off the PS3 in 2 years time is NOT profitable especially considering the significant losses they'll make in the initial 2-3 year roll out of a newer platform..

After losing billions on the PS3 so far I highly doubt they'd prolong any further losses by jumping ship just when the PS3 is starting to reach it's apex with respect to profitability..

Also bear in mind the platform (PS3) provides Sony with a greater scope for revenue growth (retail software, peripherals & hardware, digital downloads & possibly in the future; user-generated content..) than any of the previous platforms ever did therefore it's alot more likely that the PS3 will remain as a viable, financially successful platform for a very long time..

Don't forget, MS killed off the xbox because they couldn't break even due to over-investment.. Even if they attempted to sustain the platform for another 5 years they still wouldn't have made anything off it.. Sony on the other hand have the benefit of a tried and tested, stable business model which allows them to project profitability in an adaquate time frame so they really have no incentive to rush into the next generation for the sake of beating the competition..

And if PS3 can manage to pull itself out of the current situation and out-perform the 360 in lifetime sales & market share, they wouldn't ever have to consider MS a threat to their business model again (after starting out the gate with absolutely nothing in their favour this time around..)

 

[liolio]

An interesting new :http://www.beyond3d.com/content/news/534

It seems that intel and ms don't agree on the future of 3D rendering (i know there's more into this than computer graphic I'm reading the related thread on this forum).

MS and Intel could settle over some change in the future of DirectX I can't see Ms and Intel fighting each others...

May a settlement between intel and Ms over how the cpu could be included in the future rendering pipeline of directX will be part of intel MS negociation for the next Xbox.

If Ms fail to convince Intel I think that IBM would be able to offer something close to larrabbee to MS

EDIT trying to be clearer

What I mean is the same as newell about multi core said If intel takes the larrabee route (not as GPU but as a cpu with outstanding number crunching alities and able to add something to the rendering pipeline) even Ms will have to take in account this decision.
Ms will settle over Intel and the gpu actors to keep everybody as happy as it can (and running XP/Vista + directX )

 

[liolio]

Exactly..

Don't forget, MS killed off the xbox because they couldn't break even due to over-investment.. Even if they attempted to sustain the platform for another 5 years they still wouldn't have made anything off it..

I think nobody forget that Ms killed the xbox, but they killed it because they were losing too much money on hardware due to how bad the contracts MS signed with intel nvidia were (+ the cost of HDD).

 

[Megadrive1988]

The RAM in all likelyhood will be GDDR-5.


The GPU will be the main focus just like Xenos with EDRAM to help take the pressure off of a single narrow bus.



I'll guess the CPU will have 6 Power PC out-of-order cores.

fixed with the widest external memory bus now being 512-bit ( R600 / Radeon 2900), I can't see how a next generation console's external bus will be anything less than 256-bit, I'd be hoping for 512-bit bit PLUS EDRAM for really super ultra high bandwidth they'll probably be only 256-bit though.

 

[AlNom]

fixed with the widest external memory bus now being 512-bit ( R600 / Radeon 2900), I can't see how a next generation console's external bus will be anything less than 256-bit, I'd be hoping for 512-bit bit PLUS EDRAM for really super ultra high bandwidth they'll probably be only 256-bit though.

Why not? GDDR4 already provides double the bandwidth of GDDR3 at the same clock speed and bus width. And by 2012, we'll probably see GDDR5 being mass produced.

 

[hoho]

Why not? GDDR4 already provides double the bandwidth of GDDR3 at the same clock speed and bus width.

Am I missing something or you are wrong? I always thought that GDDR4 only offered the possibility of higher clock speeds. At same frequency and bus width GDDR3/4 should be equal. I'm not sure what GDDR5 brings.

 

[Betanumerical]

Am I missing something or you are wrong? I always thought that GDDR4 only offered the possibility of higher clock speeds. At same frequency and bus width GDDR3/4 should be equal. I'm not sure what GDDR5 brings.

Wiki seems to think differently, then again it isn't always the most accurate source.

GDDR4 SDRAM introduced DBI (Data Bus Inversion) and Multi-Preamble to reduce data transmission delay. Prefetch was increased from 4 to 8 bits. The maximum amount of memory banks for GDDR4 has been increased to 8. To achieve the same bandwidth as GDDR3 SDRAM, the GDDR4 core runs at half the speed of a GDDR3 core of the same raw bandwidth. Core voltage was decreased to 1.5 V.

Hynix Semiconductor has introduced the industry's first 1 Gib GDDR5 memory. It supports a bandwidth of 20 Gib/s on a 32-bit bus, which enables memory configurations of 1 GiB at 160 Gib/s with only 8 circuits on a 256-bit bus.[5]

 

[liolio]

[wet dreams]
http://arstechnica.com/news.ars/pos...s-core-to-core-optical-dream-in-progress.html

if this can be available by 2010/11 it could allow for a chip mixing cpu and gpu and provide very high bandwidth between them!
Intel is also working on something close.

along with the new rambus announcement....
[/wet dreams]

 

[hoho]

To achieve the same bandwidth as GDDR3 SDRAM, the GDDR4 core runs at half the speed of a GDDR3 core of the same raw bandwidth

DDR2 also needs only half the core clock of DDR1 to achieve same bandwidth. Of course DDR2 has IO bus running at twice the clock speed of DDR1. Could GDDR5 vs GDDR3/4 be similar?

http://www.xbitlabs.com/articles/memory/display/ddr2-ddr_2.html

 

[Arun]

GDDR4's internal clock is half GDDR3's for a given level of bandwidth, and memory burst size is doubled. AFAIK, the same happened again between GDDR5 and GDDR4.

Why not? GDDR4 already provides double the bandwidth of GDDR3 at the same clock speed and bus width. And by 2012, we'll probably see GDDR5 being mass produced.

Surely you must mean 2008, and not 2012? For a next-generation console, GDDR6 or a XDR equivalent seems more likely to me.

As for my next-gen predictions, I expect Microsoft to go with AMD, partially for political reasons (as opposed to Intel and NVIDIA). Sony will go with NVIDIA, with a slightly customized DX11 design that is only capable of pixel shading. Intel has no chance with Nintendo, so that's either NV or AMD again. I also predict that every single upcoming console will debut as a single-chip solution with an integrated CPU-GPU. All of them will be manufactured at TSMC on a 32nm process.

 

[liolio]

I agree with the one chip thingy, but th esame way intel put to core II duo on one chip (hoping for optical interconnections).
But I think that next gen will be out around 2010/11
So for me a 45nmprocess is more likely.

 

[kyetech]

I think Intel has a good chance of a look in with MS this time around due to their Fab / process advantage (and reliability of continued process shrinking going forward). Plus their GPU / CPU tech post 2009 I think politics aside, Intel would be a smart choice for a higher end console producer this time around. And we know sony wont go Intel. Intel will surely want a piece of the expanding living room market with their new GPU / CPU graphics ideas.

Interesting that you think it will be an intergrated single chip. So how much die space do you think they will throw at it initially?

400mm^ @32nm would be around 4.5 billion transistors?

Maybe thats too high,

I guess more like 300 - 350mm^ would be more like it? around 3.3+ billion transistors @ 32nm

Seems like a crazy amount now, but in 2011 itll be just above mainstream performance right?

 

[kyetech]

I agree with the one chip thingy, but th esame way intel put to core II duo on one chip (hoping for optical interconnections).
But I think that next gen will be out around 2010/11
So for me a 45nmprocess is more likely.

Definitely dont think 2010, I think they will aim for eeking out life of this gen as much as poss and having 32nm at their disposal, I get the feeling no one is in a rush this time around to cut the cycle short.

I think 2011 summer / fall. or even early 2012.

Thats my prediction.

WII2 or WII HD whatever will be a different story / different business model. and is not really included. Hell they could launch every three years if they wanted.

 

[AlNom]

Surely you must mean 2008, and not 2012? For a next-generation console, GDDR6 or a XDR equivalent seems more likely to me.

I was trying to be conservative with something that should be in high supply and relatively cheap.