I enjoy reading the post from engineers on this board. Intelligence is important, but a linear view of the world does have its drawbacks. They can't wait until the Intel Atom processors for phones come out so that they can prove once and for all that there is a superior ARM processor already available in the market, or just about to come out. ARM wins, case close, Intel fails.
Had this been true Intel would have been out of business years ago. Intel has a long history of not having the best processors in the market from time to time. If this logic is indeed correct AMD would be going like gangbusters and Intel would be laying off right now.
Intel uses this phrase, so you ARM fan boys please write this down. "Best in class". When it comes to processors for phones this is something to remember. They simply want the user to have a good experience. Sure there are going to be people who want the fastest processor at any given moment and perhaps it may be ARM. For the rest of the world, a good experience is important with a phone, more than speed. Honda Civics aren't the fastest cars for the money, but they certainly sell a lot of them.
When it comes to competition Intel does things at times which may look counter productive. Was it really a good idea to put a GPU under the die of the processor in Sandy Bridge? In some ways it makes the layouts of the boards less complex, but in reality it was a move to deprive Nvidia and ATI/AMD of funding. Expect to see more computer builds without discrete graphics cards in the future, not such welcome news to an already dying market.
Yes, in the big pictures cell phone processors don't really matter. Perhaps Intel really does care less about them, or do they?
If you have a business degree like I do you get your obligatory accounting courses. One of the more interesting topics other than the hot blond who sat in front of me all semester was the term "fixed costs." When you open a fab there are a lot of fixed costs, the machinery, the land, the building. Staff and materials are variable costs, we'll just focus on fixed costs.
If Intel takes on lets say 10 billion dollars in debt to open the fabs and they sell exactly on processor at $200, well it does very little to clear up that debt. But when you are pumping through 90K wafers a month each little processor gets to take a few more pennies off the fixed cost liability. A less fancy way of saying it is "each processor sold helps pay for the factory."
The ramp up in the fabs, or the doubling of output by 2016, however you like to phrase it is un-freaking believably expensive. Not many companies can afford to pull it off. As Intel uses the the phone processors to not only pay off the building, but to help increase the margins of the more expensive processors by diluting the fixed costs. Oh, I'm sorry, ARM fans get annoyed when I mention HUGE PROFITS in front of them. Sorry, I'm not an engineer, I'm an investor, I like the sound of HUGE PROFITS.
Speaking of ARM, Intel is trying to deplete ARM's financial resources. Each ARM fab needs 100 million 28nm processors to sell before they hit break even. Yikes! The journey into 28nm will happen for ARM. But, they will be there for a while. The road to profits is going to be longer. Yes it would help if we had a rip roaring world economy and hopefully that will happen.
What is interesting is that each move is designed to set keep ARM from taking over like we've all heard so much about. Yes, this is not a zero sum game, but keeping competition at bay is something Intel has a lot of expertise with.
So, when I see someone point out ARM has the fastest processors and they will take over, I kinda get a smile.
Atoms come in a wide range of TDPs from 0.65W to 13W
There is actually a Z600 Windows PC phone now so it obviously isn't 10W ;-)
'I normalized the Atom result down to 0.9 GHz, which is probably unfair to it as scaling should be better than that as bandwidth/clock would go up at lower clockspeed, and computed the ratio of it to the 0.9 GHz A9 result to get an idea of relative ipc and here's the result'
I posted the actual links to the results so you can do your own comparison (again below). I didn't check the results of the normalized results, but they looked dubious at first glance(?)
That isn't true even at 45nm.
0.65W Z500 Atom
1.3W Z600 Atom SoC (the gpu is already PowerVR even though it has an Intel codename)
The problem is performance/power and performance/cost due to the large SoC die size (400 sq mm at 65/45nm MCM and 150 sq mm at 32nm) for only one core even if it does HT.
I am dubious at those claimed Max TDP numbers as some of the systems I've seen with those chips are well north of 10 watts. Perhaps the rest of the chipset?
This means you won't admit to anything that's in the future which is a convenient way to pretend that Intel's fabrication isn't going to reel ARM in.
There is no way to have a meaningful discussion under your limitations so let's just skip it.
You are trying to compare apples and oranges. You can not compare Intel's high performance processes in X years against a low power process in X years. You and I would just be speculating on 'bull' from both sides. If you are talking about AMD then I would agree with you.
Yes, Intel has a process lead in HIGH PERFORMANCE. It's yet to be seen if this HIGH PERFORMANCE process lead can be brought to bare in the low power sector.
If everything you are saying is true Atom wouldn't have any business. Yet Intel has 450 new Atom customers in the last 12-18 months with 200 of them moving from non-Intel platforms.
"But all is not well in Santa Clara: their Atom group is down 32% to $269mn, a tiny fraction of what it was during the netbook rush of 2009 and 2010"
Fundamentally Intel has to start sacrificing Atom margins to match ARM pricing from 32nm onwards to get real phone traction. Atom is a very small part of Intel's business so it doesn't matter if they bring in less money at higher volume. However any volume it picks up will be at ARM's expense so that will be of long term competitive benefit. Time to start getting serious price-wise with Atom as the main x86 iCores can still continue to bring in the real premium money above Atom whilst no-one is buying single-core Atom SoCs at $30-40 when you can pick up quad-core ARM SoCs at that price now.
OK, I'll take your word for it about Transmeta...
I remember the noise about x86, but was too swamped at the time to care
"So both ARM linux and Transmeta x86 low power notebooks have already been tried with no success, it could just be that battery life is not that important for a fully featured notebook "
The available ARM chips just don't have enough juice...
Heck, I sure don't want one for a notebook...
At least today I don't...
But one thing constant is change...
"ARM killing Intel fantasies down with a crash "
It's not a zero sum game...
Just as much the fantasy INtel will kill ARM...
'But Transmeta tried out x86 INtel...'
They had a lower performing part at lower power just like ARM at the moment so your x86 reference is just a strawman and actually helped as it allowed greater software availability.
So both ARM linux and Transmeta x86 low power notebooks have already been tried with no success, it could just be that battery life is not that important for a fully featured notebook which would kind of bring all those ARM killing Intel fantasies down with a crash ;-).
Okay, well I got a little misled by you starting off with
"Suppose ARM wins everything and x86 becomes extinct"
followed by "Look at how much trouble INTC would be in."
That wasn't actually your conclusion.
It could have been a little clearer but I understand now where you were going. My apologies...