Predict: The Next Generation Console Tech

[liolio]

Actually, I came here to query... What would an all Intel console look like? Or maybe intel+ATI GPU. Would it have a Xeon? Nehalem? Larrabee? Or for AMD's side, you have Fusion and whatever r800/900/infinity. A strictly P/C based console!

Given the console low power requirement and if Intel manage to offer MS a deal they're not scared off.

I would dream of @ 32nm :

Two chips including both atom cores and larrabee (2?)cores say in a 1/4 1/5 proportion maybe a L3 cache would help perf (at the cost of less logic)
Real fast communication between the two chips.
Nice bandwith to UMA RAM
Maybe a tiny Intel SSD
A tiny HDD(by the system would come tiny could be huge by actual standard) real slow and cheap.

 

[Heinrich4]

My guess as i said before for ps4 is CBE with 32 SPUs and 4 PUs + 16 SPUs + 4 PU Vizualizer and 4 Pixel Engines (4 core gpus only with pixel shaders array?) with 64MB eDRAM (maybe enough for 1080P) in one die/chip like patent in 2002 in USA with 8GB RAM.

X"720" maybe some advance in powerpc archtecture with 8 or more cores with gpu shader unified integrated with UMA 8GBs at least.

 

[Gubbi]

Above all else, focus will be on cost next gen, IMO.

If your cost is too high compared to the competition you either bleed to death subsidizing sales or lag behind in sales with detrimental effect on publisher/studio support and long term revenue.

Sony and MS will aim for a $400 launch price, Nintendo for $250-300.

Silicon development looks shaky beyond 32nm using either double patterning immersion or EUV lithography for 22nm. The end of silicon integration scalability is likely to result in more conservative ICs since they can't be reduced in size as rapidly (or at all), so I think we'll see around 200-300mm^2 in total for a MPU+GPU combo. I'm aware that the end of silicon scaling will alter the economics of foundries since tools will be amortized over much longer timeframes and cost therefore will fall significantly.

The pricepoint of $400 will also dictate how much RAM will be in there. I don't think a HDD will be standard. With Flash ram pricing in freefall we'll probably see a system with 8-16GB Flash ram as standard and a HDD as an add-on. With different SKUs of course, similar to the 40/80GB PS3 and Arcade/Pro/Elite 360.

Cheers

 

[Carl B]

I agree with the sentiments of a reduction in initial die size vs the present gen; without the same level of scaling assurance present going forward from their launches, the economics change in terms of how much can be absorbed on hardware loss.

RAM will depend on whether the memory industry is still in its present squalid state or whether it can reclaim pricing power, and I think on storage, I'm in agreement that the base consoles will probably ship with a nominal solid-state capacity, with the majority of capacity available through add-on/drop-in drives.

All of the above obviously assumes no landscape altering breakthroughs for the semi industry in the near future.

 

[archangelmorph]

What like MetaRAM..?

 

[AlNom]

Metaphysical RAM

 

[Megadrive1988]

My guess as i said before for ps4 is CBE with 32 SPUs and 4 PUs + 16 SPUs + 4 PU Vizualizer and 4 Pixel Engines (4 core gpus only with pixel shaders array?) with 64MB eDRAM (maybe enough for 1080P) in one die/chip like patent in 2002 in USA with 8GB RAM.

 

[Heinrich4]

http://www.imagepup.com/up/Vdzs_1205346738_CELLplusVS.jpg[/img]

Thanx Megadrive1988 for pic


( and im dreaming with this configuration for console since 2002/2003)

 

[Mart]

8 GB would be a dream, but by 2011/2012 that won't even be "normal" for PCs.

We're talking 4 years from now right? And in less than a year, we'll have Intel's Nehalem, which'll bring 3 channels of DDR3. Now I don't think a real enthousiast will stick only a measly 3GB set of DIMMs in that, that would be a step back when a lot of people now allready use 4GB. Instead, people with money will bite the bullet and go with 3*2GB, 6 GB total. And that's at the start of 2009. So you're saying three years later, people will still be stuck with less than that in ordinary PC's?

Btw, does anyone know if there's a direct relation between Moore's law and pc memory? The law would suggest memory doubles about every 18 months, but is that true in reality?

I can't seem to edit my post, so sorry about the double post. Just wanted to add that the 8GB we're talking about here isn't just system ram, it's also video ram, just as 360's 512MB is both.

ModNote: Editing privilege appears after a certain amount of days registered and after you post more. -AlS

 

[Shifty Geezer]

We're talking 4 years from now right? And in less than a year, we'll have Intel's Nehalem, which'll bring 3 channels of DDR3. Now I don't think a real enthousiast will stick only a measly 3GB set of DIMMs in that, that would be a step back when a lot of people now allready use 4GB. Instead, people with money will bite the bullet and go with 3*2GB, 6 GB total.

Why? I have a 1 GB system when the RAM cost me over £100 and it handles everything. My friends have recently bought 2 GBs only because it's so cheap. What are people going to be doing on their PCs that'll warrant >4GBs? Professionals will want more as fidelity of productions increases. People will buy more RAM when it comes bundled in their PC. If the cost drops dramatically perhaps it'll become commonplace. But otherwise, why is anyone going to start putting huge amounts of RAM in their PC when most of it will sit unused most of the time? That's a waste of money!

 

[Heinrich4]

In my opinion 8GB for next gen it make sense...

Some reference in playstation universe:
psone/psx -> 2MB RAM in 1994
ps2 -> 32 MB RDRAM in 2000 =~16 * more tham previus gen
ps3 -> 512 MB XDR/DDR3 in 2006 =~16 * more tham previus gen
"ps4" in 2011/2012 -> 8GB

 

[Shifty Geezer]

My guess would be 4GB. I'm not sure more RAM would have any benefit. Perhaps there'd be a big lump of fast Static RAM in there as a data cache. Resolution will be 1080p max. 4xAA on that is plenty enough. Asset resolution will not explode as improved visuals will come from improved processing. eg. Prebaked lightmaps can be forgotten as realtime calculated AO will take its place. Smarter vertex rendering and realtime displacement will eliminate the need for stupid resolution models. 4GB with crazy amounts of BW will serve gamers far better than an excess of RAM IMO and be more economical for the manufacturers.

Also for historical precedent, the consoles have always been short on RAM compared to the PC space. I doubt PC's will go far beyond 4GBs for previously mentioned reasons, and so precedent would suggest consoles stuck at 2 GB.

 

[Farid]

The memory question is probably the most interesting one in this debate, so far.

Because while it would be hard to argue against the need for processing power. For reasons others than economical and production feasability ones, that is. It wouldn't be talking nonsense to say that the RAM amount needed in these future consoles wouldn't have to be the "highest possible."

With more parallel computational power and I/O bandwidth available (from fast BRD drives, plus the addition of caching from HDD/solid state memory), you'll have access to a ton of bitmap graphics assets (the worst offenders usually). And if you consider the fact that this new processing power would allow you to rely on more complex shader operations, you might end up not needing much higher resolution maps, or more of them, compared to current gen for certain effects.

Personally, I do not see the need for hand created textures (normal and diffuse) stored in bitmap form go down next-gen. It has been claimed that this now current-gen would be all about procedurally generated content, including in-surface shaders (A.K.A. textures), yet it's been anedotical more than anything else. And I still don't see it happening next-gen neither.

The best, easiest and cheapest way to get realistic/good looking surfaces will still be about using painted (from scratch or touched up pictures) textures for diffuse maps. And for that reason, we can't go for a too low amount, as far as RAM is concerned, in our speculations.

 

[D3B]

Here is a interview with John Carmack.

No matter who does what, the next generation is going to be really good as rasterization, that is a foregone conclusion. Intel is spending lots of effort to make sure Larrabee is a competitive rasterizer. And it’s going to be ball park competitive, we’ll see how things work out, but a factor of 2 plus or minus is most likely. But everything is going to be a good rasterizer. We should have enough general purpose computational ability to also be able to do some of these other novel architectures and while everybody thinks it’s going to be great I have to reiterate that nobody has actually shown exactly how it’s going to be great. I have my ideas and I’m sure other people have their ideas but it’s completely possible that the next generation of high end graphics is just going to be rasterizing like we do today with a little more flexibility and 10x the speed.

http://www.pcper.com/article.php?aid=532

 

[Shifty Geezer]

Personally, I do not see the need for hand created textures (normal and diffuse) stored in bitmap form go down next-gen. It has been claimed that this now current-gen would be all about procedurally generated content, including in-surface shaders (A.K.A. textures), yet it's been anedotical more than anything else. And I still don't see it happening next-gen neither.

The question is why? As you say, and I've noticed, we aren't seeing fabulous procedural content, 2 years into XB360's life with it's tessellation engine and direct Synthesis bus. These design decisions seem to be being ignored by developers. Are they just limited? Or is there not enough spare CPU power to add synthesis to already demanding games? Or are the theories too academic and not practical enough to make it into real applications?

Working smarter rather than harder, there should be lots of scope for shifting requirements from storage to processing. Dirt maps and procedural detailing could add interest to lower-resolution textures. It ain't happening though! The lack of progress in the now does raise concerns about how developers will approach next-gen. If they are still reliant on ever bigger textures and vector datasets, something will have gone very wrong with the industry IMO. Indeed, even looking at small download titles we're seeing no experimentation here where it could be used. Is there just no interest in forging a new technology?

 

[pjbliverpool]

We're talking 4 years from now right? And in less than a year, we'll have Intel's Nehalem, which'll bring 3 channels of DDR3. Now I don't think a real enthousiast will stick only a measly 3GB set of DIMMs in that, that would be a step back when a lot of people now allready use 4GB. Instead, people with money will bite the bullet and go with 3*2GB, 6 GB total. And that's at the start of 2009. So you're saying three years later, people will still be stuck with less than that in ordinary PC's?

Thats a gooid point. I had assumed my next upgrade would take me to 4GB, maybe 8GB at a push. But with the triple memory lanes of Nehalem (which will almost certainly be my next CPU - 8 core version) it seems that 6GB is pretty much a dead cert for me.

I guess thats going to be some time in 2009.

 

[Mike11]

Above all else, focus will be on cost next gen, IMO.

If your cost is too high compared to the competition you either bleed to death subsidizing sales or lag behind in sales with detrimental effect on publisher/studio support and long term revenue.

Sony and MS will aim for a $400 launch price, Nintendo for $250-300.

Silicon development looks shaky beyond 32nm using either double patterning immersion or EUV lithography for 22nm. The end of silicon integration scalability is likely to result in more conservative ICs since they can't be reduced in size as rapidly (or at all), so I think we'll see around 200-300mm^2 in total for a MPU+GPU combo. I'm aware that the end of silicon scaling will alter the economics of foundries since tools will be amortized over much longer timeframes and cost therefore will fall significantly.

I don't know about the others (IBM&Co.), but at least Intel is on track for 22 nm in 2011 according to the newest roadmaps. I guess the rest is 1-2 years behind like today. But beyond 22 nm is the big unknown (EUV etc.).

We're talking 4 years from now right? And in less than a year, we'll have Intel's Nehalem, which'll bring 3 channels of DDR3. Now I don't think a real enthousiast will stick only a measly 3GB set of DIMMs in that, that would be a step back when a lot of people now allready use 4GB. Instead, people with money will bite the bullet and go with 3*2GB, 6 GB total. And that's at the start of 2009. So you're saying three years later, people will still be stuck with less than that in ordinary PC's?

Nehalem will only bring 3 channels for the Xtreme plattform. And since it won't be socket compatible to the mainstream plattform like today, i suspect the market share will not be that big. By 2011/2012 64-bit Windows will still have a much smaller installed user base than 32-bit, so for most Desktop PCs and Notebooks 4 GB will be the limit. Also console RAM was never bigger than a quarter of mainstream PC RAM. So even if a new 64-bit only Windows comes out in 2010 (unlikely) and wins incredibly fast market shares so that by 2011/2012 there are more 64-bit systems out there than 32-bit, more than 4 GB unified RAM for the consoles is unlikely imho.

 

[kyetech]

Considering that both consoles will go for a 5 years life-cycle minimun , this will give us a xbox3 at Q4-2010 and a PS4 at Q4-2011.

For next xbox:
To say that a G200 from Q4-2008 will have the same graphical capabilites with xbox3 at Q4-2010, is like saying that a FX 5950 is more or less equal to Xenos
SLI or not, they will not be in the same league*.

For PS4:
Q4-2008 G200
Q4-2009 G300
Q4-2010 G400
Q4-2011 G500 (A derivative of G500 as PS4 GPU*)

( *Without even mentioning the high possibilities for dual GPUs )

Carmack thinks a three year projection is not that screwy. For targeting hardware using top end rigs. cool.

Whether it makes sense for gaming to have these thousand dollar graphics cards is quite debatable but it’s really good for developers; to be able to target something high end that’s going to come out three years from now by being able to pay more money today for 2x more power.

http://www.pcper.com/article.php?aid=532

 

[DeadlyNinja]

Hey guys, let say Nintendo makes a super duper ultra machine next generation. Now, do you REALLY want to see a Mario that looks like this?

Here's a link to an HD version of the picture, because, it's not next gen unless it's HD.

http://img.photobucket.com/albums/v207/goodbean/boards/realmario.jpg

 

[AlNom]

That's no longer the same art direction even... Mario in HD would be more akin to a Pixar film. High Def != Real