Do you think it would be a mistake for MS and/or Sony to launch in 2012?

[Dominik D]

I think this hasn't been discussed before:

http://news.softpedia.com/news/TSMC-Sees-Increased-Demand-for-the-28nm-Process-Node-236571.shtml

It states that TSMC's current wafer output is around 20k/mo and that this should increase to around 100k/mo in 2012 (300mm wafers) with 28nm-oriented Fab 15. Customers at this point are AMD, Nvidia, Altera, Qualcomm and Xilinx with Broadcom, LSI Logic and STMicroelectronics rumored.

So for, say, 200mm^2 per chip their peek expected capacity in 2012 would be 306 dies per wafer so 30,6 mil chips per month assuming 100% yield (which won't happen, obviously). I think it's fairly reasonable to expect TSMC to produce chips for 2012 console if their plans and estimations are doable.

Oh, I have a question too. What is a "typical" yield in, say, first and fifth year of production? Is it more like for example 30% to 90% or is it something better (or worse)?

//edit
NVM, got my Q answered: http://bnrg.cs.berkeley.edu/~randy/Courses/CS252.S96/Lecture05.pdf

 

[fearsomepirate]

Kinect sports (3million) and Dance Central (2.5 million) say hi.

Note my use of the present tense, "neither are moving software." Sure, it moved a bit of software back when it launched, but so did the Dreamcast. For the "fastest-selling entertainment product in history" or whatever MS is billing it as these days, I sure haven't seen a lot of Kinect games in the NPD top ten for a long time.

 

[dobwal]

Note my use of the present tense, "neither are moving software." Sure, it moved a bit of software back when it launched, but so did the Dreamcast. For the "fastest-selling entertainment product in history" or whatever MS is billing it as these days, I sure haven't seen a lot of Kinect games in the NPD top ten for a long time.

Hard to know if its moving hardware or not because unless you keep up with ongoing sales of Kinect games released last year, you would have no idea especially since all notable releases have been centered around the holiday seasons.

Its should be telling that Kinect centric games releases planned for the last 3 months of this year (somewhere in the neighborhood of 30-40 games) far out number PS Move centric titles.

 

[wco81]

Has Kinect progressed beyond simple sports and dance games?

Will the next Halo or COD game support Kinect? I thought one of the problems was that any gameplay mechanic which required precision timing wouldn't work well because there's no way to replicate button presses, since it doesn't really track finger movements?

 

[TheChefO]

Didn't the first 90nm gpu based desktop cards release during the same quarter as the launch of the 360 with the X1300 Pro releasing in Oct 05 and the 360 releasing in Nov 05? Didn't Nvidia release its first 90nm gpu based cards in Jan 06?

28nm would have to be delayed past q4 2012 to delay the next Xbox launch unless there is something tremendously wrong with yields even after a delay into q4 2012 to make launching MS's next console too costly for holiday 2012.

Indeed.

xb360 launched right along side the first 90nm GPUs.

Hardware-wise, it seems if MS/Sony wanted to, they could launch early next year as the 28nm node isn't a limiting factor if 2005 90nm is any indication.

Late 2012 is very comfortable for 28nm.

 

[TheChefO]

I think this hasn't been discussed before:

http://news.softpedia.com/news/TSMC-Sees-Increased-Demand-for-the-28nm-Process-Node-236571.shtml

It states that TSMC's current wafer output is around 20k/mo and that this should increase to around 100k/mo in 2012 (300mm wafers) with 28nm-oriented Fab 15. Customers at this point are AMD, Nvidia, Altera, Qualcomm and Xilinx with Broadcom, LSI Logic and STMicroelectronics rumored.

So for, say, 200mm^2 per chip their peek expected capacity in 2012 would be 306 dies per wafer so 30,6 mil chips per month assuming 100% yield (which won't happen, obviously). I think it's fairly reasonable to expect TSMC to produce chips for 2012 console if their plans and estimations are doable.

Oh, I have a question too. What is a "typical" yield in, say, first and fifth year of production? Is it more like for example 30% to 90% or is it something better (or worse)?

//edit
NVM, got my Q answered: http://bnrg.cs.berkeley.edu/~randy/Courses/CS252.S96/Lecture05.pdf

To be fair, the article doesn't state when TSMC will be up to the 100k/mo rate, but that is their full capacity.

Interesting to know their current rate though being 20k/mo.

Yet another indicator pointing to an "all clear" for 2012 28nm.

 

[AlNom]

To be fair, the article doesn't state when TSMC will be up to the 100k/mo rate, but that is their full capacity.

Interesting to know their current rate though being 20k/mo.

Yet another indicator pointing to an "all clear" for 2012 28nm.

You do realize they have to stock up partway through the year, not at the end of the year. This will not be a cheap endeavour either way you try to justify it.

Don't you think they should have mentioned Nintendo on the clients list?

 

[Dominik D]

@TCO: Not really "all clean". If numbers from the PDF I linked are any indication then 2012 launch is a very tight proposition. Not impossible, but risky and extremely costly. There are tons of other factors and perhaps paying upfront and starting early is the winning strategy, dunno. But if TSMC can't get to 100k wafers/mo by, say, April, there's no next-gen in 2012.

//edit
I should write: no next-gen in 28nm.

 

[TheChefO]

You do realize they have to stock up partway through the year, not at the end of the year.

Ok ... and what about this report leads you to believe MS/Sony could not be stocking up and manufacturing their next-gen chips right now?

Nothing that I can see.

Expensive? Sure. I never said otherwise.

BTW - When xb360 launched at 90nm, how long were Nvida and AMD 90nm GPUs on the shelf?

 

[Dominik D]

Minus two.

 

[BRiT]

And in significantly lesser quantities.

 

[AlphaWolf]

Note my use of the present tense, "neither are moving software." Sure, it moved a bit of software back when it launched, but so did the Dreamcast. For the "fastest-selling entertainment product in history" or whatever MS is billing it as these days, I sure haven't seen a lot of Kinect games in the NPD top ten for a long time.

Most of those sales were this year. Kinect sports was just off the top 10 for several months at the beginning of this year. Kinect doesn't really have any big AAA titles there's little reason to expect they would chart with big numbers, that doesn't mean there aren't lots of them moving in reasonable quantities earning money for MS.

 

[Joe DeFuria]

I see lots of posts that basically indicate that whether or not 2012 is too early depends on the state of component (chip) availability.
I don’t see it that way at all.

I hate to break it most of you, but we are actually starting to reach diminishing returns as far as graphics goes. Note I said starting and diminishing. Not that there are no returns to be had of course!

What I'm saying is that with each 5 year period, the "difference" between graphics from one gen to the next is or course less and less noticable. This is particularly true with consoles that have a resolution "ceiling". I believe we have reached the point with the current generation that the NEXT generation is going to be a tough sell to the masses based soley on graphics fidelity. Certainly with only 5 years between generations, and the LAST generation being produced with pretty much bleeding edge hardware for the time.

It's no surprise to me that we're going on year 7 with X-Station consoles with no definitive talk of upgrade in sight. I'm thinking that if "improved graphics" is to be the MAIN selling point of the next gen X-Stations, we're going to need at least 10 years between generations. That being said...It's pretty clear the players in the industry understand this based on how consoles have evolved over the past couple generations.

So the goal will be to really push some other differentiators other than graphics to wean folks off the current gen and into the next gen. Either expand upon existing concepts, or introduce new ones.

Nintendo: as we know, they will continue to do their own thing: Make hardware that is itself near profitable or actually profitable to sell at launch, new control paradigms, rely on key franchises. They have not historically differnetiated their generations by expanding away from games, but on the way we interact with them.

Microsoft has some obvious avenues to explore: Kinnect 2.0, leverage Windows 8 ecosystem and continued x-box Live expansion. Heck, just putting a blu-ray drive in the new X-box will be a nice little differentiation from their current gen.

Sony: I'm not sure what Sony's direction with PS4 is going to be beyond "better hardware" and "something with Home". But I have to believe they have something planned.

So whether or not 2012 is "too early" is to me not dependent on availability of graphics or CPU chips manufactured on a certain process. A couple more process shrinks will be required BEYOND 2012, IMO, for graphics / performance to make a difference enough to be a MAIN differentiator from consoles of the current generation to next.
The question is, are all the pieces in place / ready for the OTHER (non graphics factors) to be marketed? For Microsoft, will a clear Kinnect hardware upgrade ready for 2012? Will Windows 8 devices have acceptance? 2012 will be to early for that...but I think Fall 2013 could be the right time for MS. If things go "right" for MS, 2012 will be the year that Windows phones (Nokia / mango) start to really get some decent traction, and by the end of 2012 and into 2013 (again, if things go "right") Win 8 at least for Tablets and Phones will be seen by the public as a viable competitive platform.

This would set the stage for fall 2013 for MS, IMO.

Sony? I really have no idea...we've seen Sony pseudo-embrace "openess" with PS3 (Linux)...maybe they will embrace a google / android ecosystem? Leverage Vita in some novel way? Do something special with Home that we haven't even thought of? Whatever Sony is going to market beyond "better graphics," they better have it ready for a console launch. Since I don't know what they are cooking up, I can't say if 2012 is too early for Sony...I just know that if it's just "PS3 with better graphics" then 2012 is too early.

 

[TheChefO]

I see lots of posts that basically indicate that whether or not 2012 is too early depends on the state of component (chip) availability.
I don’t see it that way at all.

I hate to break it most of you, but we are actually starting to reach diminishing returns as far as graphics goes. Note I said starting and diminishing. Not that there are no returns to be had of course!

What I'm saying is that with each 5 year period, the "difference" between graphics from one gen to the next is or course less and less noticable. This is particularly true with consoles that have a resolution "ceiling". I believe we have reached the point with the current generation that the NEXT generation is going to be a tough sell to the masses based soley on graphics fidelity. Certainly with only 5 years between generations, and the LAST generation being produced with pretty much bleeding edge hardware for the time.

It's no surprise to me that we're going on year 7 with X-Station consoles with no definitive talk of upgrade in sight. I'm thinking that if "improved graphics" is to be the MAIN selling point of the next gen X-Stations, we're going to need at least 10 years between generations. That being said...It's pretty clear the players in the industry understand this based on how consoles have evolved over the past couple generations.

So the goal will be to really push some other differentiators other than graphics to wean folks off the current gen and into the next gen. Either expand upon existing concepts, or introduce new ones.

Nintendo: as we know, they will continue to do their own thing: Make hardware that is itself near profitable or actually profitable to sell at launch, new control paradigms, rely on key franchises. They have not historically differnetiated their generations by expanding away from games, but on the way we interact with them.

Microsoft has some obvious avenues to explore: Kinnect 2.0, leverage Windows 8 ecosystem and continued x-box Live expansion. Heck, just putting a blu-ray drive in the new X-box will be a nice little differentiation from their current gen.

Sony: I'm not sure what Sony's direction with PS4 is going to be beyond "better hardware" and "something with Home". But I have to believe they have something planned.

So whether or not 2012 is "too early" is to me not dependent on availability of graphics or CPU chips manufactured on a certain process. A couple more process shrinks will be required BEYOND 2012, IMO, for graphics / performance to make a difference enough to be a MAIN differentiator from consoles of the current generation to next.
The question is, are all the pieces in place / ready for the OTHER (non graphics factors) to be marketed? For Microsoft, will a clear Kinnect hardware upgrade ready for 2012? Will Windows 8 devices have acceptance? 2012 will be to early for that...but I think Fall 2013 could be the right time for MS. If things go "right" for MS, 2012 will be the year that Windows phones (Nokia / mango) start to really get some decent traction, and by the end of 2012 and into 2013 (again, if things go "right") Win 8 at least for Tablets and Phones will be seen by the public as a viable competitive platform.

This would set the stage for fall 2013 for MS, IMO.

Sony? I really have no idea...we've seen Sony pseudo-embrace "openess" with PS3 (Linux)...maybe they will embrace a google / android ecosystem? Leverage Vita in some novel way? Do something special with Home that we haven't even thought of? Whatever Sony is going to market beyond "better graphics," they better have it ready for a console launch. Since I don't know what they are cooking up, I can't say if 2012 is too early for Sony...I just know that if it's just "PS3 with better graphics" then 2012 is too early.

Good points.

However,
I think the differentiating factors that have occurred this gen are also leading to consoles which have life beyond gaming. Especially life beyond hardcore gaming.

I think this is where you'll see the stretching of the console cycle as the bottom end casual experience is extended via a cheap xb360 and a cheap ps3.

The casuals and family are just being introduced to this generation.

But this is a completely different market.

One that, like you said, doesn't appreciate subtlety (10x!) in graphic advancement.

They will be fine with xb360/ps3 with kinect2/move-more and software that takes advantage along with app integration and all the rest of the goodies that i expect to come in the following years.

None of those extras "require" a new console. xb360/ps3 fit that role just fine.

However, the hardcore gamer, the one that buys a ton of DLC on the console, the one that pushed the xb360 attache rate above 10, the one that can see the difference that a 10x improvement will bring to the gpu and will spend the $400 to get the launch unit ...

This demographic is starving for a new console.


So, though I agree that diminishing returns will impact how the generation cycle unfolds, I don't think it will translate to next-gen not coming soon.

I think it will translate to overlapping consoles more so than we saw in the past.

 

[upnorthsox]

I hate to break it most of you, but we are actually starting to reach diminishing returns as far as graphics goes. Note I said starting and diminishing. Not that there are no returns to be had of course!

You mean like how a 6870 has 23.3x ALU and 1.8x clock over Xenos yet only 6.2x gflops? It would seem that throwing trannies at the problem is fully into diminishing returns.

 

[Shifty Geezer]

Are we still counting performance in GFlops instead of actual achieved work?

 

[Joe DeFuria]

So, though I agree that diminishing returns will impact how the generation cycle unfolds, I don't think it will translate to next-gen not coming soon.

I think it will translate to overlapping consoles more so than we saw in the past.

I can see that as well.

If the intention is to stretch the "overlap" between generations though, this would put a higher emphasis on backwards compatibility with the new consoles, imo.

I don't think MS and Sony would be able to go with the approaches they went this gen, essentially "lip service" to backwards compatibility.

Having a fully backwards compatible console carries its own set of risks though (technical implementation and others), but I think it would be needed if the intention would be to really extend / embrace a two tier marketing model where the current generation would be entry level and the next gen would be high-end for the forseeable future.

 

[upnorthsox]

Are we still counting performance in GFlops instead of actual achieved work?

Ok, putting aside the fact that ALU's are just little flop processors. What metric would you want to use to measure whether a 23.3x increase in the ALU count is returning anything close to that in performance? What's achieved work? Pretty pictures? What's pretty?

 

[Shifty Geezer]

Ok, putting aside the fact that ALU's are just little flop processors. What metric would you want to use to measure whether a 23.3x increase in the ALU count is returning anything close to that in performance? What's achieved work? Pretty pictures? What's pretty?

If all that mattered was increasing the flops, they could just put in more of the older shader units. Exactly the same transistor budget could be spent on taking Xenos and multiplying up its shader units. There's a reason AMD haven't done this - not all flops are created equal. More transistors need to be spent on programmability, for example, that enable features otherwise impossible on older, simpler execution units. Try getting Xenos*23 to render some new DX11 geometry shader effects...

If you want to measure real-world performance, you take a game running on one GPU and then run it on the new one and see what improvements you get in resolution/framerate. You then take a game written for a later, more functionally advanced GPU and see what speed it runs on the older one. You might see 6x the improvement in framerate of Skyrim (someone can actually give a reasonably accurate figure on some game comparisons if they wanted to do the research) on a 6870 versus Xenos, but you'll also see a far bigger advantage running stuff Xenos can't do.

 

[TheChefO]

If all that mattered was increasing the flops, they could just put in more of the older shader units. Exactly the same transistor budget could be spent on taking Xenos and multiplying up its shader units. There's a reason AMD haven't done this - not all flops are created equal. More transistors need to be spent on programmability, for example, that enable features otherwise impossible on older, simpler execution units. Try getting Xenos*23 to render some new DX11 geometry shader effects...

If you want to measure real-world performance, you take a game running on one GPU and then run it on the new one and see what improvements you get in resolution/framerate. You then take a game written for a later, more functionally advanced GPU and see what speed it runs on the older one. You might see 6x the improvement in framerate of Skyrim (someone can actually give a reasonably accurate figure on some game comparisons if they wanted to do the research) on a 6870 versus Xenos, but you'll also see a far bigger advantage running stuff Xenos can't do.

Agreed, there are features in dx11 that would be near impossible for prior gen hardware, but much of this will be a non-issue as gpgpu takes hold and gpus become more programmable.

I'm pretty sure neither next gen machine will have to worry about a non-dx11 compatible gpu and most likely both will have GCN (or comperable) gpus.

As for the other, well I don't have a good feeling about it.

I assume they will be in talks shortly about either acquisition, or exclusive partner rights.

They will be quite a commodity for whoever picks them up ...