JEMC said:
You are a lot more optimistic than me . That's why I'll only say two things. 1-Despite how bitter Nvidia is for being left out of the console business, to the point where they take every oportunity they can to dismiss it, there is one thing that's true: the profit margins from consoles are on the low side. Yes, AMD is in desperate need for money, but they will make more money selling their own graphic cards and mobile parts than from the components made for Nintendo. It's because of that reason that expecting them to hinder their own Polaris and 14nm business for Nintendo (or Sony or MSoft) doesn't make sense.
2-It's been confirmed by everyone that 14nm production costs are a lot higher than the 28nm costs, and the resulting chips are also more expensive. Take a look at this graph: As it clearly says, it's how it costs to produce a wafer with each manufacturing process. Look at how much a 28/32nm costs: $6,000, now look how much a 14nm costs: $18,000. Three times more, but they won't get three times more chips (specially now that the 14nm process is still not perfect while the 28nm one is very mature).
I know that I won't make you change your mind, but don't be surprised when someone opens a NX (because Nintendo never tells those details) and we find out that its components are made with the 28nm process. |
I don't think I'm being optimistic, just using information available to see what seems logical and see whether a rumor could have some legitimacy too it.
That rumored handheld, with performance close to XB1 requires what Polaris is able to offer on the GPU side. An AMD Puma CPU, scaled down to 14nm would also offer the necessary CPU performance to match XB1 in that form factor, but since AMD aren't scaling Puma down to 14nm and Zen as far as I'm aware is also 14nm. It's not really wrong to say that Zen would offer better than 2X the performance of Puma per watt, because new architecture, made on the 14nm node, compared to 28nm.
As for your points on console profit margins, well there's money to be made there, it's still pretty big business, otherwise AMD wouldn't have got into it.
Nintendo doesn't have to make much from hardware on launch models, since the console business model has always been about making profit from software. A potential Nintendo handheld is way more likely to sell big numbers than a console will.
The console business isn't going to hinder the PC market for graphics cards, CPUs or SOCs/APUs that AMD will sell themselves, because the two audiences are not in competition with each other. Console and Handheld gamers aren't PC customers really, they're their own thing and by targeting the dedicated gaming console/handheld business AMD increases their own sales. It actually makes perfect sense to provide products for this area of the market.
As for costs per chip, well as I've said before it's about what's required from a performance to energy perspective. AMD have even said themselves that their launch Polaris 10 product is aiming at a price of under $349 for a graphics card with performance comparable to an R9 290 (I previously remembered it at a 290x, but I was wrong about that, still the point stands though). Nintendo doesn't need something that big for this hypothetical handheld unit, they need a cut down version of the demo unit they used to show the architecture off at the AMD reveal.
Nintendo would be using more chips from a single wafer, so you have to know how many you can make from a 14nm wafer to say whether it's cost effective for AMD or whoever.
Finally for the record I'm just trying to see whether this rumor made any sense, more from a technical perspective (performance and power consumption, ability for a handheld power source to handle it).
For the record a Carrizo style APU could also fit the bill, the 15 watt was able to have 8 compute units, so it's not far from Xbox One levels of performance, 4 extra CUs, some more CPU cores could still fit the power envelope needed, although the rumor doesn't say that this hypothetical NX handheld was in the exact same power level as XB1, just that it was closest to XB1, which as I said before fits either half the power of XB1 or half way between XB1 and PS4. The latter doesn't really make sense for Carrizo, since the power requirements would put it above what a notebook level battery could handle.
Getting close to XB1 would be important if Nintendo wanted to have really easy porting of 3rd party games to NX handheld. To prevent cut down versions of those games you need something at least on XB1's level.
This is all speculation anyway right now, I was just bringing one of the rumors to the thread and explaining how it could work in reality.