Predict: The Next Generation Console Tech

[AlNom]

"The TSMC 28nm process has surpassed the previous generation's production ramps and product yield at the same point in time due to closer and earlier collaboration with customers"

How long time has passed since first 28nm things rolled out from their product lines? Surprassing previous process in first few months isn't anything remarkable.

Believing in such vague PR was the first mistake.

Actually the wafer numbers are real, the production is real, the chip sizes are real (based on gpu+ cpu sizes this gen at launch) the wafer cost estimates are in the ballpark.

No, you're assuming some number in between the current and full production. That's not real because you're just guessing for the sake of arguing.

Even chip sizes aren't real, they are just some number you're assuming is going to be the same. You have no idea what the power and thermal characteristics are compared to the last gen. There is simply no information to indicate what target chip sizes are going to be.

You're just stacking ball park guesses on top of one another.

The point I'm getting at is that it isn't impossible, isn't unreasonable, and isn't unthinkable that 2012 on 28nm is a very real possibility.

28nm isn't impossible. What I find highly improbable though is for Nintendo to suck up the costs of a new node instead of the significantly more robust and cheaper previous node.

 

[TheChefO]

You have no idea what the power and thermal characteristics are compared to the last gen. There is simply no information to indicate what target chip sizes are going to be.

Indeed.

And such is the topic of discussion in a prediction thread.

As for Nintendo using 28nm or not, I don't think most people in this thread care either way as it's likely to be an also-ran.

I'm more interested in where Sony and MS take next-gen.

Neither of which will be roadblocked out of 2012 by the 28nm process node.

 

[AlNom]

And such is the topic of discussion in a prediction thread.

Pulling numbers out of the air to support your belief? At least work with ranges. Your cherry picked arguments are not productive.

Neither of which will be roadblocked out of 2012 by the 28nm process node.

Which is wholly irrelevant because they aren't launching in 2012. It's a line of arguments and predictions that leads nowhere.

 

[TheChefO]

You're just stacking ball park guesses on top of one another..

There's a difference between a guess and an estimate.

Guess - to form an opinion from little or no evidence

Estimate - to determine roughly the size, extent, or nature of

 

[TheChefO]

...At least work with ranges.

....

Let's narrow it down to a range:

20k is current production at TSMC.

Let's assume they can't ramp production at all ... 20k/mo wafers

4m/mo 300mm^2 chips @ 50% yield (pretty reasonable I'd say) = 2m/mo capacity - draw from this what you will on MS requirements.

100k/mo wafers = 10m/mo 300mm^2 chips @ 50% yield is their fab capacity.

So 2-10m units per month is a reasonable production range to expect in 2012.



As for costs:

http://semiaccurate.com/2011/09/08/exclusive-tsmc-raises-prices-on-amd-and-nvidia/


So my target of $7,500/wafer should be rather generous and the upper limit of what to expect.

As to whether or not 2012 will or will not happen for MS/Sony... How is it that you can categorically state that they will or will not?

My purpose for these posts is to point out what is indeed possible within a given budget and timeframe.

Now unless you have something which proves otherwise, I'm not sure what your point is.

 

[willardjuice]

His point is you have no idea what you're talking about.

 

[McHuj]

Don't post much here, but I would just like to throw in my two cents. I believe that the next xbox will be an 32nm SOC with 32nm eDram:
http://public.dhe.ibm.com/common/ssi/ecm/en/tgd03018usen/TGD03018USEN.PDF

I think that in order to hit 3+ GHz clocks for the CPU, you need to stay at a high performance process like SOI. The manufacturing, power, and overall cost benefits will drive the decision to keep design as a SOC. I'll be pretty surprised if there's a 28nm GPU.

For late 2012, I think 32nm will be more than ready. By mid next year, Global Foundries should have shipped millions of chips on 32nm: Llano, Trinity, Bulldozer, and other APU's. I think that's plenty of time to mature the process to reasonable yields.

If you look at the A8-3850 APU, currently developed on 32nm, it's 228 mm^2 and 1.45B transistors. TPD 100W and a RETAIL price of $135, I think that's a good barometer for expectations. Obviously the chip will be much more customized in terms of architecture and physical design layout as well as on a much more mature process. But, llanoe is shipping today at a reasonable consumer price, in reasonable volumes so I don't think 32nm is out of the question for next gen in 2012.

My prediction will be 1.5B transistors for the SOC (roughly 500M for the CPU, 1B for the GPU) and another 500M for the eDram, all under 150W TPD. In terms of transistor count, that would be around ~4X increase (I don't think that's improbable). With the fine system optimization in the architecture, bandwidth, and clocking, I bet real performance would probably be closer to 6x-8x over the current generation.

Would that be good enough? If it shipped at $299 and MS didn't lose money on it?

 

[Shifty Geezer]

There's a difference between a guess and an estimate.

Guess - to form an opinion from little or no evidence

Estimate - to determine roughly the size, extent, or nature of

Let's assume they can't ramp production at all ... 20k/mo wafers

4m/mo 300mm^2 chips @ 50% yield (pretty reasonable I'd say) = 2m/mo capacity - draw from this what you will on MS requirements.

And that 50% is a guess as there's no evidence. As others point out, a 60% figure quote without any explanation what's being made is completely useless. So yes, you are guessing, as when you do rely on evidence it's pretty much guessed figures itself.

 

[fehu]

Globalfoundries delayed the costruction of a new fab because they don't have enought customer request to justfy it, so i doubt that this will happen

 

[Proelite]

I think 2012 is out of the question as far as launch dates go. This gives a studio a year's time to finish up whatever they're working on for next generation. Considering that there hasn't been any leaks of games in development for next generation, and that multiple studios have stated that they're in the dark when it comes to next generation, 2012 seems highly unlikely.

 

[TheChefO]

And that 50% is a guess as there's no evidence. As others point out, a 60% figure quote without any explanation what's being made is completely useless. So yes, you are guessing, as when you do rely on evidence it's pretty much guessed figures itself.

Guess is without evidence.

I've shown evidence.

60% yields 1 year after 40nm

28nm ramping better than 40nm

Is it unreasonable to assume 50% yields when TSMC was stating 60% for 40nm and all signs are pointing to 28nm going better than 40nm?

Do you have other evidence suggesting my numbers are off, or are you just busting my balls?

 

[hoho]

For late 2012, I think 32nm will be more than ready. By mid next year, Global Foundries should have shipped millions of chips on 32nm: Llano, Trinity, Bulldozer, and other APU's. I think that's plenty of time to mature the process to reasonable yields.

Didn't AMD scrap their 32nm fusion cpu's that they were supposed to produce in GF and instead redid the design for GF as they supposedly had somewhat better working production line?

 

[TheChefO]

His point is you have no idea what you're talking about.

ok ... point out where I'm wrong.

 

[hoho]

I've shown evidence.

You've showed absolutely nothing that would be of any use when it comes to yields.

60% yields 1 year after 40nm

I'll ask again:
What chip size?
What kind of redunancy?
Is it binned?

28nm ramping better than 40nm

For how many months after first chips came from the fab?

Is it unreasonable to assume 50% yields when TSMC was stating 60% for 40nm and all signs are pointing to 28nm going better than 40nm?

Yes, very much so. Especially for console chip that is extremely suspectible due to not being binnable.

Do you have other evidence suggesting my numbers are off, or are you just busting my balls?

Our main point is your data is far from reliable.

 

[Deleted member 13524]

Guess is without evidence.

I've shown evidence.

60% yields 1 year after 40nm

28nm ramping better than 40nm

Is it unreasonable to assume 50% yields when TSMC was stating 60% for 40nm and all signs are pointing to 28nm going better than 40nm?

Do you have other evidence suggesting my numbers are off, or are you just busting my balls?

Nearly 4000 posts and you still don't know that, here in B3D, guesses with less than 92% probability of happening are severely punished with an agressive session of balls busting??!




j/k

 

[TheChefO]

I think 2012 is out of the question as far as launch dates go. This gives a studio a year's time to finish up whatever they're working on for next generation. Considering that there hasn't been any leaks of games in development for next generation, and that multiple studios have stated that they're in the dark when it comes to next generation, 2012 seems highly unlikely.

EA, UBI, EPIC have had dev kits for a while...

Smaller players might not, but those are the big ones that we know of.

Software will be limited at launch as is always the case and it will likely not push the hardware to it's limits ... as is the norm for a launch.

As for what to expect:

2012 28nm using the same die budget projects to roughly 4b transistors.

A pretty substantial jump in processing power.

 

[TheChefO]

Nearly 4000 posts and you still don't know that, here in B3D, guesses with less than 92% probability of happening are severely punished with an agressive session of balls busting??!




j/k

Based on the reactions, you'd think I was suggesting that nextgen will be based on alien technology from the future which will run on stemcells ...

 

[McHuj]

Didn't AMD scrap their 32nm fusion cpu's that they were supposed to produce in GF and instead redid the design for GF as they supposedly had somewhat better working production line?

Rumor is that it was the 28nm notebook class Krishna and Witchita APUs. The performance ones, Trinity and Llano, are 32nm and nothing has changed with them.

 

[Shifty Geezer]

Guess is without evidence.

I've shown evidence.

60% yields 1 year after 40nm

As hoho says, of what sort of processor? Without those details, the number is meaningless, making it useless as a piece of evidence. eg. Take 3000 mm^2 of wafer. That could be 10 300 mm^2 processors, or 300 10mm^2 processors. If the latter and they have 60% yields, that's a lot of defects across that 3000 mm^2 of wafer. And that would mean defects in 10 large processors made on the same area, rendering them useless.

Based on the reactions, you'd think I was suggesting that nextgen will be based on alien technology from the future which will run on stemcells ...

No, the reaction comes from you presenting the flakiest arguments and believing they are based on decent evidence. No-one would mind the notion of stemcell alien consoles if supported by decent evidence. Plus your repeating the same arguments over and over and diluting the discussion from other contributors.

Present a document/quote that shows yields of CPUs and GPUs of the size suitable for your expectation of next-gen components are up to 60% already and you'll have made your case. Otherwise learn to understand when people are asking for real evidence that you need to do so or let the conversation go at that point, rather than just add more conjecture and circumstantial references.

 

[TheChefO]

The only reason yield was even brought up is some were suggesting it would be prohibitive on 28nm for 2012.

I've brought in countless quotes and links suggesting this is not the case, and 28nm is coming along just swell.

On top of that, I've brought in quotes/links where the previous 2 xbox consoles have used the latest and greatest nodes within the same year as GPU's utilizing the same new node.

Shortages at launch? What else is new?

So hardware wise, not only is it possible, it's probable.

Then, I brought in sales figures from other successful consoles showing how low and behold, even though hardware was selling well, the HW maker decided to introduce a nextgen system anyway in the following year.

I've brought in quotes and links suggesting that MS has been thinking of forward compatibility in games since a few years ago.

I've brought in evidence of big Devs having dev kits for almost a year.

In fact, at this point I've brought in so much evidence which points to 2012, you all will have to convince me that 2012 WON'T be the year MS launches xb720 ... present some real evidence.

As far as predicting what's under the hood ... well I brought that into the discussion as well ... based on 28nm of course:

Assuming 32/28nm launch in 2012 would yield 8x trans count, this would amount to a budget of roughly 4billion (497m x8 = 3,976m) if we are to assume equal budget/process node.

This leads to some pretty interesting potential hardware:

With that budget, MS could extend the xb360 architecture to the following:

10MB EDRam (100m) => 64MB EDRam (640m) Enough for a full 1080p frame buffer with 4xaa

3 core xcpu (165m) => 9 core xcpu (495m) - or an upgraded 6 core PPE with OoOe and larger cache along with an ARM core (13m trans)

This leaves a hefty 2.8b trans available for xgpu which could accommodate 3x AMD ~6770 (1040m) ~3 teraflops or 4x AMD ~6670 (716m) ~2.8 teraflops.

Such small, modular chips would enable good yields on new(er) processes until they were mature enough to combine together and eventually integrate to an APU.