Slashdot Mirror
Intel Claims Chip Suppliers Will Flock To Its Mobile Tech
MojoKid writes It has been over six years since Intel first unveiled its Atom CPUs and detailed its plans for new, ultra-mobile devices. The company's efforts to break into smartphone and tablet sales, while turning a profit, have largely come to naught. Nonetheless, company CEO Brian Krzanich remains optimistic. Speaking to reporters recently, Krzanich opined that the company's new manufacturing partners like Rockchip and Spreadtrum would convert entirely to Intel architectures within the next few years. Krzanich has argued that with Qualcomm and MediaTek dominating the market, it's going to be tougher and tougher for little guys like Rockchip and Spreadtrum to compete in the same spaces. There's truth to that argument, to be sure, but Intel's ability to offer a competitive alternative is unproven. According to a report from JP Morgan, Intel's cost-per-wafer is currently estimated as equivalent to TSMC's average selling price per wafer — meaning TSMC is making money well below Intel's break-even. Today, Intel is unquestionably capable of building tablet processors that offer a good overall experience but the question of what defines a "good" experience is measured in its similarity to ARM. It's hard to imagine that Intel wants to build market share as an invisible partner, but in order to fundamentally change the way people think about Intel hardware in tablets and smartphones, it needs to go beyond simply being "as good" and break into territory that leaves people asking: "Is the ARM core just as good as the Intel chip?"
That's not really clear how much the cost per wafer matters.
Intel are a process node ahead of the competition so presumably they can fit more transistors on a given wafer, so the worth of a wafer is higher. Also if that's a best-to-best comparison rather than equivalent-node-to-equivalent node then it's not a direct comparison.
I'd assume intel's yields and costs on old but not retired processes are resprctively higher and lower than on the top end node
SJW n. One who posts facts.
Intel has flexed its near monopoly power heavily over the years. ARM licenses its tech and anybody can build compatible chips if they want. If I were an OEM I would be wary of giving Intel back control over my devices.
"A person is smart. People are dumb, panicky dangerous animals and you know it." - K
Of couse, Intel will always play fair and not use its money to sell components below cost, give money to OEMs putting 'Intel Inside' stickers as "advertisement", leverage the CPUs it sells for laptops and desktops to the OEMs which also produce tablets and/or phones.
They would not dare to!
They can also expect Microsoft's help them for that, after the amazing success that was Windows RT for ARMs.
If Intel's product is actually better, they wouldn't have to make such bold predictions because people will want it.
This is just more marketing bullshit.
Anons need not reply. Questions end with a question mark.
Windows 8 tablets run Intel processors, so they're not "trying" to break into the tablet market, they HAVE broken into it. The smartphone market, not so much.
The legacy compatibility can be handled by running a VM. There's not need to support it in hardware at all, but you're right that it takes only a tiny bit of die space. A '386 had fewer than 300k transistors. A P5 had 3 million. An Atom has 47 million. That's skipping some architecture steps in between, but you can see the direction. Each can have a complete implementation of the previous architecture and it's still only a few percent of the processor complexity.
I've said it before: companies that are perfectly happy with ARM chips now are not going to be in a hurry to lock themselves in to sole-source chips.
http://apple.slashdot.org/comments.pl?sid=3202785&cid=41735071
Intel would have to be better than ARM, and not just a little bit better... they would have to be dramatically better, such that the risk of being locked in to a sole source vendor would be worth accepting. It hasn't happened yet and I don't expect it to happen.
It will be difficult for any company, even AMD, to really challenge Intel in the high-end CPU market. But it would take a miracle for Intel to lock down the mobile CPUs market.
Intel's plan:
0) Get everyone locked in to needing to buy chips from Intel.
1) Charge stiff margins for those chips.
2) Profit!
Intel does have some chips in some Android devices, but they aren't charging the stiff margins they would like to charge. I don't think they will ever manage to do it.
Second best would be to not charge stiff margins but at least get a large chunk of the available profits from the mobile space. But I don't think they will be able to push out ARM and gain majority share of the market; they will continue to be a niche player.
lf(1): it's like ls(1) but sorts filenames by extension, tersely
The problems you describe have _absolutely_nothing_ to do with the underlying chip instruction set architecture.
Then what is it? Android OS problem?
The problems you describe have _absolutely_nothing_ to do with the underlying chip instruction set architecture.
Then what is it? Android OS problem?
YES it is an Android problem. barring a manufacturing defect the chip does what it is told, their is nothing you can really do at the hardware level to counteract an OS that gets its knickers in a knot.
The legacy compatibility can be handled by running a VM. There's not need to support it in hardware at all
Ha-ha-ha-ha-ha!
There speaks someone who's never written a PC emulator. Intel already tried it with the Itanium, which is one reason that chip was a disaster.
1) Full 64bit may work on linux and BSDs, but on Windows you need 32bit at least for backward compatibility (or even compatibilty with current 32bit software). Hybrid 32/64bit mode (called x32, with 32bit pointers) could also find use on mobile platform where the cheap offering only have 512MB or 1GB ram (but you lose good ASLR)
3)4) things are working somewhat this way already (chip with 2 cores and GT3 graphics, etc.) with GPU power and features and reliability increasing with each gen. But sometimes we do want a fast CPU and a weak GPU, if that's cheaper or the GPU performance is meaningless (software dev, high quality video encoding, image editing/photo processing on software that doesn't use the GPU..). You're also overlooking that on e.g. ultrabooks you will find the high end GPU, but the (short term average?) power budget for the whole chip is 15 watts. So it's both high end and low performance (relatively), and expensive (the $1000 thin laptop). More people are interested in a $500 laptop with lower end chip.
Ah yes.. once you enter 64bit mode, 16bit mode and 16bit virtual machine (VM86) are unavailable. So after booting your 64bit OS it's as if the chip is 32/64bit only, like most 64bit CPU are I believe (e.g. Sparc, MIPS, PPC..)
Maybe you're running out of RAM and it's crapping out, or some stupid app is hammering the low quality flash with writes, plus garbage collection pauses..
More multi-core won't help you. Better OS (Android 5 or even Windows?), more memory, better flash (such as stacked flash that works like a single chip SSD?) or post flash memory (MRAM or other) will help. Sometimes a 15 year old PC has it better.. Its hard drive may fare much better at writes sometimes (!) and it allows swap space.
Everything went up after the meltdown. You could have put money in just about anything and done likewise... which is exactly what the ultra-wealthy did. Since the ultra-wealthy could afford to take the risk and not sell when things were low, they could then swoop in with cash and buy up things real cheap. If you're poor, you probably weren't able to do so. This phenomenon has contributed to the increasing wealth disparity in the United States. Congrats on not being poor.
You can lead a horse to water, but you can't make it dissolve.
Arm v8 is coming out, and it has a much more sophisticated memory model than any previous chip (it is basically the first chip to clone the c++11 memory model exactly). I expect that means that arm will go from making cooperation between cores less efficient than intel to more efficient than intel.
Yes, as far as I know, Android devices never have virtual memory turned on (you don't want to wear out the flash memory on-board, because when it's worn out the device is bricked).
So when you're low on memory, Android starts killing processes pretty much at random.
On SOME Android devices you can turn on swapping if you root, but on others it's disabled in the kernel.
1) Cut all the legacy crap. Forget emulating x86-32b/x86-16b, just build a straight 64-bit chip
Most of that is microcoded. There's no benefit in removing most of the 32-bit stuff, because it's exactly the same in 64-bit mode. Things like segments are microcoded and if you have a non-zero base you hit a really slow microcoded path on Intel's low-power CPUs.
I am TheRaven on Soylent News
One not so commonly known fact is that many - probably the majority of the tablet 'wins' Intel is touting - come from subsidised engineering.
What does that mean ?
Basically it means - that Intel pays a good chunk of the engineering costs to Samsung or HP to use an Intel processor in a Tablet or Chromebook.
This problem will only be magnified the further down the profit margin stack Intel tries to play.
Problem is this is a company that is used to making a 60% margin on each processor it sells, it's culture can't grasp a few ideas.
First - IA is not a solution people are crying out for in mobile, that ship has sailed, ARM has won. Intel can't get past it's architecture - hence the brainless decision to sell XScale. Still pushing Atom - and spending billions trying to compete with the ARM architecture - when Intel should be spending those billions tweaking an Intel branded ARM SoC - using it's advanced process technology to produce the ultra-power efficient core mobile wants. Tweaking IA to be both - scalable in the top end - and low-power in the mobile domain is a fools errand. Use the right tool for the job. IA in the data centre and ARM for mobile. Hello BK ?
Second it's a company focused on IoT - the internet of things. Great - except that the money is in mobile computing - not in internet enabled fridges and watches with silicon inside. Buzzwords that sound great in internal fora but have no resonance with people outside. Hello BK ?
Both points really reduce down to the following. It's a company that is used to owning all of computing and for political reasons can't make the leap to provide what the market wants. Simple fact is - it's trying to 'educate the market' which is also a complete waste of time. Try providing what the market demands in mobile (hint a low power ARM based on 14 nanometer would be a good starting point) - instead of trying to ram IA down the throats of mobile users.
My need for IA in my phone is roughly analogous to my need for an asshole on my elbow.
Hello BK ?
I personally have owned several ARM based Android tablets and recently bought an Intel Atom based 7" ASUS Android tablet for my kid on sale for about $100. For that kind of money, I was not expecting much, but have been pleasantly surprised at how fast it is. It is as fast, if not faster than most modern Android tablets. And it lasts just as long as my Nexus 7. I was not expecting Intel to have achieve parity at this time. But they have. If they are able to offer a solution that is a little more performant while using less power than the ARM chips, I think they can succeed.
People have shown willingness to spend money on high end phones and tablets. I currently own a Samsung Galaxy S5. A $700 device. Do I wish it was a little faster and lasted a little longer? You bet! There is a lot of room to improve the mobile experience. At the pace that Intel has been developing their mobile solution, I think they can and will succeed in this space.
The problems you describe have _absolutely_nothing_ to do with the underlying chip instruction set architecture.
Then what is it? Android OS problem?
Yes, it's a fundamental problem with trying to run the whole device under a series of "Java" (Dalvik) Virtual Machines. The move to ARTS may help that; we'll see; but as long as the JVM-derived stuff is in the picture it will probably continue to be an issue, even with ARTS (it's just that ARTS has a magnitude better performance over Dalvik).
Truth is like the sun. You can shut it out for a time, but it ain't goin' away. - Elvis Presley (source: imdb.com)
There's a difference between using virtual memory and using a swap file. Android devices ALL use virtual memory and a MMU. However, they usually don't have a swap file.
Linux doesn't normally use a swap file. Rather it normally uses a swap partition which provides a lot better performance than a swap file since the OS can directly manage the layout itself instead of having to necessarily traverse the file-system.
Yes, you can use an actual swap file with Linux instead; it's just no where near the norm (1%) and highly discouraged.
Truth is like the sun. You can shut it out for a time, but it ain't goin' away. - Elvis Presley (source: imdb.com)
Microsoft is the dark horse in this race. Intel's trump card is that their products run on x86. Computing power is getting to the point where mobile devices are able to run Windows 8 quite well. I have a Dell Venue Pro 8 from last year that can run full-on Windows 8.1, and it's based on the old Atom. While the device has flaws, it is still goddamned amazing (and very useful!) to have Windows instead of Android in terms of application compatibility.
The new Intel Broadwell processors promise even better performance and lower power consumption. If Microsoft does not fuck up Windows 10, then this would push sales of Intel-based tablets. Why bother running Android or iOS when you can get desktop applications running on your tablet, even if you pay a slight premium? Microsoft's Surface Pro 3 sold $900 million in the last quarter. Competing devices such as Yoga 3 would only get more users onto the idea of a Wintel tablet.
A NYC lawyer blogs. http://www.chuangblog.com/
Maybe in the future we can get an SD card made of MRAM, then you can use it as swap on an older device. And get virtually free writes for applications and data, if there's no big barrier to applications installed on it.
You have similar examples of exponentially better storage upgrade for old gear, like using Compact Flash cards in an Amiga, or a cheap 2.5" IDE to mSATA adapter for laptops.
I'm not an Apple fanboy, I own both Android devices and Apple devices. Why don't you buy an Apple phone the problem you describe which I have seen on many Android devices does not happen on Apple devices, they rarely crash.
Apple changed processor architectures entirely twice.
Just because it's been done poorly doesn't mean it can't be done well.
The first Itaniums had x86 compat in hardware and were, I believe, disappointingly slow at executing x86 code. Obviously that's something that Intel could have improved if they applied themselves to the problem (and maybe they'd have made it faster if they hadn't been expecting / hoping / planning to replace x86 anyhow).
But given the different philosophies of the architectures, I think it's somewhat plausible that doing an x86 -> Itanium conversion in hardware is just a bit awkward and that software might genuinely give the flexibility to do a better job. Around the same time, Transmeta were selling their chips that exclusively exposed a software-emulated x86 layer for use in laptops. I remember wishing Intel would buy their tech and apply it to Itanium / x86 compatbility.