Yep, its true.
Per the thread's title question, the best possible post-bulldozer scenario is that Intel elects to become a division of UNICEF and thereby voluntarily chooses to manufacture AMD's future processors on a pro-bono basis. You know, for the betterment of mankind and all that.
Because right now AMD doesn't have too many other options when it comes to improving on price/performance. Only Intel's management and decision makers had the wisdom and fortitude year's ago to see to it that they were smartly investing in the R&D that they would need some 7-10 yrs later (now).
And for that they should be punished, nobody likes a smart kid who has planned smartly for his future. Bring in the FTC and bust them up!
[...]then why would anyone stay at 28nm for a long time when we should expect 20/16/14 to improve at least performance wise, but probably when it comes to production efficiency and cost too after a while.
Yep, its true.
Per the thread's title question, the best possible post-bulldozer scenario is that Intel elects to become a division of UNICEF and thereby voluntarily chooses to manufacture AMD's future processors on a pro-bono basis. You know, for the betterment of mankind and all that.
Because right now AMD doesn't have too many other options when it comes to improving on price/performance. Only Intel's management and decision makers had the wisdom and fortitude year's ago to see to it that they were smartly investing in the R&D that they would need some 7-10 yrs later (now).
And for that they should be punished, nobody likes a smart kid who has planned smartly for his future. Bring in the FTC and bust them up!
Even Samsung says the same.
You analogy (no TV, radio, IC,...) fails because you assume that Intel won't innovate and milk its technology.My response is somewhat off-topic.
Analogy:
What if it was 1948, and the transistor had recently been invented, and only Intel (who did not exist at this point in time) could make them, and they were considerably faster/cheaper/lower-power-consumption/better/reliable than everyone else's Electronics, which would have been Tubes(Valves) and mechanical-relays.
To further rub salt in the wound, let's say that Intel ONLY used these transistors, to make mainframe computers (Servers), which although fast and good, were very over-priced, as they were the only ones making (transistorized) mainframe computers, with no realistic competition.
Except AMDs terribly slow/expensive valve mainframes, which were bad compared to Intels.
The world would get no (transistor based) colour (color) TVs, Radios, Integrated circuits (as they need transistors), etc etc.
This would be very bad for the world.
So maybe sooner or later, something needs to be done, otherwise the whole world would suffer.
i.e. Such a big/wonderful thing such as an integrated circuit (the latest ones), needs to be shared, NOT given to just one person/company in the world.
Some comments:
* The graph does not provide any absolute numbers for the design cost, nor does it put it into relation to AMD's total R&D budget. So it's impossible to draw any conclusions from that.
* Price per transistor is not static over time. Likely 20/16/14 nm will come down in price going forward, just as has happened for all other nodes that precede them.
* Since AMD will be sticking to 28 nm for Carrizo/Excavator, chances are they can jump directly to Samsung/GF 14 nm after that for their next APU. Thereby they would save the design cost for 20 nm.
* If your graph and theory is accurate, I assume Intel's one year process tech lead on 14 nm compared to Samsung/GF is of little importance for desktop CPUs. Because 14 nm will apparently not bring much benefit compared to 28 nm on the desktop (not lower price, and not much higher higher frequency either). I.e. AMD might actually be smarter than Intel, by staying on 28 nm instead of 14 nm for now, since the cost is lower.
Nope. AMD confirms 20nm in 2015.* Since AMD will be sticking to 28 nm for Carrizo/Excavator, chances are they can jump directly to Samsung/GF 14 nm after that for their next non-Bulldozer based x86 APU generation after that. Thereby they would save the design cost for 20 nm.
AMD CEO Rory Read also commented on the transition to 20nm and mentioned the old business side of the logic behind the transition. Rory says that you make a move to a new technology at the crossover point between profitability, cost of technology and cost of the product that the company can sell. Chipmakers are not moving 20nm simply because they can, a company moves to a lower manufacturing node to make higher performance products and makes some money selling them.
Rory also answered a question from John Pitzer of Credit Suisse, who inquired about the 20nm rollout. Rory told him that there is a lot of life in 28nm and that we will see some mix of both 28nm and 20nm chips for AMD.
It isn't 1 year. It's more like 2-3 years. And Intel's 14nm has huge advantages compared to 28nm. About 3x higher density and up to 4x lower power consumption.* If your graph and theory is accurate, I assume Intel's ~1 year process tech lead on 14 nm compared to Samsung/GF is of little importance for desktop CPUs. Because 14 nm will apparently not bring much benefit compared to 28 nm on the desktop (not lower price initially, and not much higher frequency either). I.e. AMD might actually be smarter than Intel, by staying on 28 nm instead of 14 nm for now, since the cost is lower.
Its not possible for AMD to make an x86/ARM core on 20nm and below without having higher transistor cost than 28nm, or an IC design cost they can never recoup. So that alone limits the options drasticly.
Getting back on topic, one of the most interesting possibilities is pin compatible ARM and x86 SoCs. This will let OEMs offer Windows and Android devices with the same design, varying only a single component. This will hopefully reduce both design and inventory costs.
Not for their next post-Bulldozer x86 APU generation, which was I wrote. So far they've mentioned sticking to 28 nm for Carrizo. The next jump for AMD x86 big core after that could be directly to 14 nm.
According to this post Apple will push out A9 SoC on Samsung's 14LPE in late 2015. That's ~1 year after Intel on 14 nm.It isn't 1 year. It's more like 2-3 years.
And Intel's 14nm has huge advantages compared to 28nm. About 3x higher density and up to 4x lower power consumption.
The title is "Best possible scenario" not "Should AMD just give up and go home"
Not for their next post-Bulldozer x86 APU generation, which was I wrote. So far they've mentioned sticking to 28 nm for Carrizo. The next jump for AMD x86 big core after that could be directly to 14 nm.
According to this post Apple will push out A9 SoC on Samsung's 14LPE in late 2015. That's ~1 year after Intel on 14 nm.
But the cost per transistor is lower due to the IC design that enables it.Regardless, density does not matter much, if the cost per transistor is higher anyway in the end. Lower power consumption does not matter much either on desktop, while staying at roughly the current TDP numbers.
The title is "Best possible scenario" not "Should AMD just give up and go home"Its not possible for AMD to make an x86/ARM core on 20nm and below without having higher transistor cost than 28nm, or an IC design cost they can never recoup. So that alone limits the options drasticly.
But the cost per transistor is lower due to the IC design that enables it.
Intel's 14nm is not the same as Samsung's 14nm for a host of reasons, including the fact they are delivering a full node's worth of areal shrinking over that of their existing 22nm node.* If your graph and theory is accurate, I assume Intel's ~1 year process tech lead on 14 nm compared to Samsung/GF is of little importance for desktop CPUs. Because 14 nm will apparently not bring much benefit compared to 28 nm on the desktop (not lower price initially, and not much higher frequency either). I.e. AMD might actually be smarter than Intel, by staying on 28 nm instead of 14 nm for now, since the cost is lower.
Here and here for example.And you read that where?
Not much is known about Samsung/GF 14 nm. If you make such bold statements that Samsung 14 nm is comparable to Intel 22 nm, please show us metrics for transistor density, transistor cost, power consumption, etc to prove that. Do you have such metrics for Samsung 14 nm and Intel 14 nm to back your statements up?Its more comparable to Intels 22nm than to 14nm. It shouldnt be a surprise since the fabled 14/16nm from TSMC and Samsung is still 20nm. So no, they are much further behind Intel. Assuming they even come out with any of the claims at time.
If you think going to 3D transistors is enough to name your node 1 step lower, sure, that's ~1 year after Intel on 10nm.According to this post Apple will push out A9 SoC on Samsung's 14LPE in late 2015. That's ~1 year after Intel on 14 nm.
22nm has a up to 2x lower power consumption than 32nm, Broadwell-Y consumes 0.7x the power of Haswell-Y. That was silicon that wasn't optimized however, so final silicon + lower voltage for Atom could bring that close to 0.5x. Note that I said 'up to'.Source for the 4x lower power consumption? And no Intel PR slide please.
Density does matter. 20FF will not be less expensive because it does not have a lower density. 14nm will be ~2.2x more dense according to Intel, while wafer cost increases less rapidly.Regardless, density does not matter much, if the cost per transistor is higher anyway in the end. Lower power consumption does not matter much either on desktop, while staying at roughly the current TDP numbers.
Here and here for example.
Not much is known about Samsung/GF 14 nm. If you make such bold statements that Samsung 14 nm is comparable to Intel 22 nm, please show us metrics for transistor density, transistor cost, power consumption, etc to prove that. Do you have such metrics for Samsung 14 nm and Intel 14 nm to back your statements up?
If you refuse to work inside the possible parameters. Then we can always come up with one fantasy after the other to get to the predefined goal you wish.
It isn't a bold statement.Not much is known about Samsung/GF 14 nm. If you make such bold statements that Samsung 14 nm is comparable to Intel 22 nm, please show us metrics for transistor density, transistor cost, power consumption, etc to prove that. Do you have such metrics for Samsung 14 nm and Intel 14 nm to back your statements up?
Fair enough, but now that you've defined the parameters, I assume you came to this thread to answer the title.
What is your take on "Best Possible Scenario for post bulldozer AMD?"
Possible bigger 28nm dies. Perhaps a few 20nm dies with reduced or stagnat transistor count to enable lower consumption at a somewhat equal production price compared to today.
But competition wise we are not talking performance products. Embedded is the future for them.
so after BD, you expect AMD to stay on 28 and 20 nm?