Predict: The Next Generation Console Tech

[KKRT]

They have been struggling to enter the market for years with no signs of success.

Zune is on the market since 2006. Xbox was released in 2001. Research and support for those two devices are completely from other leagues and they are different in terms of starting new generation.

Profit or not, Microsoft will never leave the game console business. It is part of a big plan for company's future, when almost every home will have a form of electronic online entertainment device

Disagree. They're interested in entertainment market but they dont need to do it through consoles [its the most expensive and risky way].

 

[wco81]

You don't want to say "never" about anything.

MS can subsidize losses in the Entertainment division (whatever it's called) but their stock has been stagnant for a decade because they haven't turned these new ventures into cash cows. So there's at least opportunity costs.

Now with Bach and Allard gone, who knows, there may no longer be a champion for this business. It sounds like Allard left because he got overruled on a strategic direction he advocated.

However, Ballmer did seem to think Natal was a big deal so they seem to have high hopes for it. If it doesn't do well, it's not going to make or break Xbox but they might have to reassess its goals.

Despite the improvement of the X360 over the Xbox, it's still a distant second to the Wii, which is presumably one of the reasons they push this motion controller. So before they launch Xbox Next, they have to assess what it would take to win and what it means to win the console wars.

It may just be that the console isn't the strategic product that they thought it was 10 years ago. Games are a big business and there was talk about the might of the Playstation brand encroaching on PC business -- the hyped PS2 and PS3 were suppose to be capable of replacing the PC for a lot of people. We know how MS reacted to perceived existential threats like Netscape.

But now, it's the smart phone business and search advertising along with cloud services which are the big growth markets, generating mindshare as well as huge profits already. It would not be unreasonable for MS to decide that Windows Phone is more vital to its future than Xbox. And already, Windows Phone is way behind.

It would not be surprising if MS pours resources and energy into that rather than Xbox Next. MS could certainly spend on both and other projects but will it?

 

[Squilliam]

But now, it's the smart phone business and search advertising along with cloud services which are the big growth markets, generating mindshare as well as huge profits already. It would not be unreasonable for MS to decide that Windows Phone is more vital to its future than Xbox. And already, Windows Phone is way behind.

It would not be surprising if MS pours resources and energy into that rather than Xbox Next. MS could certainly spend on both and other projects but will it?

However Xbox Live is an important cloud service starting point. If they didn't have the 360 they would be starting from 0 in respect to many of the cloud services they want to develop. Since they already have a captive market, I can see where the Xbox 360 and Xbox next comes in as an important point of entry/interface with their Live cloud the same as Windows Mobile 7. Im sure Win mobile 7 would be in an even more difficult spot without the Xbox 360, Live, Live Arcade etc.

 

[kyetech]

You don't want to say "never" about anything.

MS can subsidize losses in the Entertainment division (whatever it's called) but their stock has been stagnant for a decade because they haven't turned these new ventures into cash cows. So there's at least opportunity costs.

Now with Bach and Allard gone, who knows, there may no longer be a champion for this business. It sounds like Allard left because he got overruled on a strategic direction he advocated.

However, Ballmer did seem to think Natal was a big deal so they seem to have high hopes for it. If it doesn't do well, it's not going to make or break Xbox but they might have to reassess its goals.

Despite the improvement of the X360 over the Xbox, it's still a distant second to the Wii, which is presumably one of the reasons they push this motion controller. So before they launch Xbox Next, they have to assess what it would take to win and what it means to win the console wars.

It may just be that the console isn't the strategic product that they thought it was 10 years ago. Games are a big business and there was talk about the might of the Playstation brand encroaching on PC business -- the hyped PS2 and PS3 were suppose to be capable of replacing the PC for a lot of people. We know how MS reacted to perceived existential threats like Netscape.

But now, it's the smart phone business and search advertising along with cloud services which are the big growth markets, generating mindshare as well as huge profits already. It would not be unreasonable for MS to decide that Windows Phone is more vital to its future than Xbox. And already, Windows Phone is way behind.

It would not be surprising if MS pours resources and energy into that rather than Xbox Next. MS could certainly spend on both and other projects but will it?

Trouble is though, Microsoft seem very reactive rather than innovative. If they remain like that, then they will never be leaders and all their venture projects cannot hope to access big margins. On the bell curve of product life cycle, if you come in half way through. You are facing a downward slope in growth. Apple on the other hand always seem to be at the beginning of the bell curve with the RIGHT products. 2 things then happen.

1) Brand gets strengthened (perception of market leader) = higher margin.
2) massive growth due to lack of competition.

Then people catch up. then you cant sell with the same margin and the innovation process has to happen again in order to access big margins.

I predict in 10 - 15 years when the apple type product innovation slows down. Apple will be on a level playing field and wont maintain this ridiculous growth / big gross margin system.

 

[liolio]

So, subpar performance, development hassles, and lots of wasted power waiting around for something to do. Yeah, that's a good idea for a console.

Why FP32? because its standard everywhere. It is consistent. Developers are familiar with it, it provides computational commonality with everything else.

Integer bit shifts? Its already in the pipeline, its basically free, and it has its uses.

No vertex processing because obviously you want everything to be slower right?

All the width in the world doesn't mean jack if you are just twiddling your thumbs because you can do anything better.

I've question here, in a recent presentation (slide 46) Crytek states :

We spend too much of computational power per frame!
Precision is mostly redundant
No need to compute colors in 32-bits floating points
Even h/w rasterizers was 12-bits of fixed precision in good old times

So that's may why RolfN think about moving away from FP32, still it's not my question.
Without touching to the hardware and considering a software pipeline as Sweeney describes it, could it be a gain to move from RGB to a "log luv representation" (sorry I don't a better name).
Colors are encoded with 16bits, luminosity on 16bits.
So "decoupling" colors and lightning calculation could that allow for some savings/optimizations?

 

[Buck Naked]

Yes, but I was speaking about year 2013 Even current mid-end Fermis like GF GTX 465 costs less than 300$ in retail and are 4+ times more powerful than RSX.

You're a funny guy man a GTX480 overal is atleast 10 times more powerfull than the RSX. Not strange considering it has over 10 times the transistor count (3200m vs 302m).

 

[TheWretched]

You're a funny guy man a GTX480 overal is atleast 10 times more powerfull than the RSX. Not strange considering it has over 10 times the transistor count (3200m vs 302m).

No it's not...

A car with a 2 liter engine also isn't twice as fast (in general) as a 1 liter engine car, even by the same manufacturer.

 

[ihamoitc2005]

No one needs "accuracy"

I've question here, in a recent presentation (slide 46) Crytek states :

So that's may why RolfN think about moving away from FP32, still it's not my question.

Even current console developers think too much accuracy. Less accuracy and post-process "correction" is better.

 

[AlNom]

R32F shadowmaps

...and I'm gone.

 

[Buck Naked]

No it's not...

A car with a 2 liter engine also isn't twice as fast (in general) as a 1 liter engine car, even by the same manufacturer.

I terms of raw fillrate (Pixel and Texel), memory bandwidth and floating point power the GTX480 isn't 10 times as quick. It's because of it's much more efficient unified architecture and much more advanced feature set a GTX480 is atleast 10 times more powerfull than the RSX. The difference is problably much bigger in reality.

 

[Blazkowicz]

Did you just say Fermi and cheap to produce in the same sentence?

I don't know how cheap it is to produce, but there are brand new Fermi boards selling for 199.89 euros over here. I may get a 128bits ddr3 smaller, cheaper, lower power variant down the road.

 

[Blazkowicz]

Even if Sony goes with simple road in getting 2x Cell/Cell2, 2gb DDR5 and mid-end fermi from 2012 - PS4 will still be at least 6-7 more powerfull than PS3 and it will be probably cost about 300$ to produce - quite cheap for starting machine [and they dont have to worry about loss on every console].
With great SPE libraries like MLAA, depth of field, motion blur, triangle occuling, EDGE, PhyreEngine and many multiplatform engines behind [Frostbite, CryEngine 3, UE 3] they wont have problems with 3rd party developers.

I believe they will rather convert libraries to run on the GPU or CPU+GPU, dump the Cell and use whatever high performance CPU cores they find suitable in terms of power use or political/business considerations.
that or delaying the new gen even further and come up with some "truly unified" many-core solution.

 

[TheWretched]

I terms of raw fillrate (Pixel and Texel), memory bandwidth and floating point power the GTX480 isn't 10 times as quick. It's because of it's much more efficient unified architecture and much more advanced feature set a GTX480 is atleast 10 times more powerfull than the RSX. The difference is problably much bigger in reality.

It simply isn't 10 times as fast in ANY real world application. It won't play a game that runs at 30fps on a RSX equivalent at 300fps (with an equally souped up CPU, that is). It doesn't matter if it is in theory 10000 times as fast, if the games that are run on it don't run that much faster.

 

[Heinrich4]

Sorry if have been discussed before about gpu part, but seeing good posts * and based on what I have observed since the revelation of the R-500/C1/Xenos and Geforce 7800GTX/RSX/NV47(NV47 taped out end 2004) we can believe that if the next gen consoles are released in late 2012 is quite possible** we are today seeing gpus in high end (Fermi/GF100/Geforce 480GTX or Radeon 5870) aproached a lot that will be performances of "PS4 gpu" or "X720 gpu(for 1080P/3D wee need 32/48 ROPs? Both between 2 or 3 billion transistors at 28nm and 80/100 watts?)?

* http://forum.beyond3d.com/showpost.php?p=1451477&postcount=6252
http://forum.beyond3d.com/showpost.php?p=1452472&postcount=6285
http://www.forum-3dcenter.org/vbulletin/showthread.php?p=7935873#post7935873

Date   		Price	    Videocard             Performance improvement

April 2002:	~150€	GeForce 4 MX-440	     
April 2003:	~150€	9600 Non-Pro		    ~100%
April 2004:     ~150€	5700Ultra/9600XT	    ~50%
April 2005:     ~150€	6600GT/9800Pro	            ~100%
April 2006: 	~150€ 	7600GT/X1600XT              ~100%
April 2007: 	~150€ 	X1950Pro 512MiB             ~100%
April 2008: 	~150€ 	8800 GT 512MiB    	    ~125%
April 2009: 	~150€ 	HD4870/GTX 260 896MiB       ~60%
April 2010:     ~150€   HD 5770 1024MiB 	    ~0%

** if Sony and Microsoft does not folow "wii way of life" in next gen...

 

[menmau]

Sorry if have been discussed before about gpu part, but seeing good posts * and based on what I have observed since the revelation of the R-500/C1/Xenos and Geforce 7800GTX/RSX/NV47(NV47 taped out end 2004) we can believe that if the next gen consoles are released in late 2012 is quite possible** we are today seeing gpus in high end (Fermi/GF100/Geforce 480GTX or Radeon 5870) aproached a lot that will be performances of "PS4 gpu" or "X720 gpu(for 1080P/3D wee need 32/48 ROPs? Both between 2 or 3 billion transistors at 28nm and 80/100 watts?)?

* http://forum.beyond3d.com/showpost.php?p=1451477&postcount=6252
http://forum.beyond3d.com/showpost.php?p=1452472&postcount=6285
http://www.forum-3dcenter.org/vbulletin/showthread.php?p=7935873#post7935873

Date           Price        Videocard             Performance improvement

April 2002:    ~150€    GeForce 4 MX-440         
April 2003:    ~150€    9600 Non-Pro            ~100%
April 2004:     ~150€    5700Ultra/9600XT        ~50%
April 2005:     ~150€    6600GT/9800Pro                ~100%
April 2006:     ~150€     7600GT/X1600XT              ~100%
April 2007:     ~150€     X1950Pro 512MiB             ~100%
April 2008:     ~150€     8800 GT 512MiB            ~125%
April 2009:     ~150€     HD4870/GTX 260 896MiB       ~60%
April 2010:     ~150€   HD 5770 1024MiB         ~0%

** if Sony and Microsoft does not folow "wii way of life" in next gen...

Heinrich4, I have to disagree with that table, I simply find it quite misleading.
For example HD4870 is an high end card, comparing HD5770 which is a mid end, that does not even has one full year?
Also I have never seen HD4870 selling for around 150€, in early 2009.
Well this are just some points, and would require a separate discussion.
That table is just a trying to prove some way of thinking. Not the reality.

Nevertheless I agree with your analysis, as the number pointed by you are approximated to what will be possible in the coming years.

 

[Heinrich4]

Heinrich4, I have to disagree with that table, I simply find it quite misleading.
For example HD4870 is an high end card, comparing HD5770 which is a mid end, that does not even has one full year?
Also I have never seen HD4870 selling for around 150€, in early 2009.
Well this are just some points, and would require a separate discussion.
That table is just a trying to prove some way of thinking. Not the reality.

Nevertheless I agree with your analysis, as the number pointed by you are approximated to what will be possible in the coming years.

Im fully agree with you but sometimes(sorry again...new of topic...) i have impression the gpus released since 2006/2007(coming Unified shaders arch) does not quite impressive like we see in 1999/2000 with geforce 256 SDR and great improvement end 2002 with Radeon 9800. Im wrong ? Is there any huge improvement coming with new models like TBDR in powerVr series 6(used in next gpu ps4?) or even new render paradign (some Ray tracing "way of life" or at least fully Deferred Render card)?

About gpu consoles maybe 22nm is enough mature or evolved for 2012/2013 days who nows.

 

[corduroygt]

I don't know how cheap it is to produce, but there are brand new Fermi boards selling for 199.89 euros over here. I may get a 128bits ddr3 smaller, cheaper, lower power variant down the road.

The selling price of a product has nothing to do with the production cost though

Oh, by the way, I came upon this nice article by John Hable from ND on CPU/GPU convergence:
http://filmicgames.com/archives/401

 

[menmau]

Im fully agree with you but sometimes(sorry again...new of topic...) i have impression the gpus released since 2006/2007(coming Unified shaders arch) does not quite impressive like we see in 1999/2000 with geforce 256 SDR and great improvement end 2002 with Radeon 9800. Im wrong ? Is there any huge improvement coming with new models like TBDR in powerVr series 6(used in next gpu ps4?) or even new render paradign (some Ray tracing "way of life" or at least fully Deferred Render card)?

About gpu consoles maybe 22nm is enough mature or evolved for 2012/2013 days who nows.

Yes that is right, I found it interesting that nowadays we are not seeing impressive achievements that we saw before with Geforce 256 or with ATi 9800.
I believe the word that can fits the best this situation is evolution, instead of innovation.
So that is my real question, of what will be the next generation all about, numbers aside, will it be an evolution or an innovation?

 

[TheChefO]

So that is my real question, of what will be the next generation all about, numbers aside, will it be an evolution or an innovation?

Business realities of the industry at the moment dictate a safe approach to design and development in hardware and software.

In fact, they may be attempting to push this gen until an on-live type system can replace console development and skip a "true nextgen" systems all together.

If they had their way, I'm sure they'd love to be there today.

 

[aaronspink]

Oh, by the way, I came upon this nice article by John Hable from ND on CPU/GPU convergence:
http://filmicgames.com/archives/401

I think that post can actually make you stupid. Its hard to get so much so wrong in so short a number of words.