Slashdot Mirror
Apple, ARM, and Intel
Hugh Pickens writes "Jean-Louis Gassée says Apple and Samsung are engaged in a knives-out smartphone war. But when it comes to chips, the two companies must pretend to be civil because Samsung is the sole supplier of ARM-based processors for the iPhone. So why hasn't Intel jumped at the chance to become Apple's ARM source? 'The first explanation is architectural disdain,' writes Gassée. 'Intel sees "no future for ARM," it's a culture of x86 true believers. And they have a right to their conviction: With each iteration of its manufacturing technology, Intel has full control over how to improve its processors.' Next is pride. Intel would have to accept Apple's design and 'pour' it into silicon — it would become a lowlymerchant foundry.' Intel knows how to design and manufacture standard parts, but it has little experience manufacturing other people's custom designs or pricing them. But the most likely answer to the Why-Not-Intel question is money. Intel meticulously tunes the price points for its processors to generate the revenue that will fund development. Intel's published prices range from a 'low' $117 for a Core i3 processor to $999 for a top-of-the-line Core i7 device. Compare this to iSuppli's estimate for the cost of the A6 processor: $17.50. Even if more A6 chips could be produced per wafer — an unproven assumption — Intel's revenue per A6 wafer start would be much lower than with their x86 microprocessors. In Intel's perception of reality, this would destroy the business model. 'For all of Intel's semiconductor design and manufacturing feats, its processors suffer from a genetic handicap: They have to support the legacy x86 instruction set, and thus they're inherently more complicated than legacy-free ARM devices, they require more transistors, more silicon. Intel will argue, rightly, that they'll always be one technological step ahead of the competition, but is one step enough for x86 chips to beat ARM microprocessors?'"
The war between CORE and ARM raged across thousands of worlds, ravaging the galaxy. Neither would waver in their belief in their own supremacy. For each side, the only acceptable outcome is the complete elimination of the other.
"For all of Intel's semiconductor design and manufacturing feats, its processors suffer from a genetic handicap: They have to support the legacy x86 instruction set, and thus they're inherently more complicated than legacy-free ARM devices"
Oh shut up. This argument comes up every time there's an ARM vs Intel debate. And you know what? Intel is pushing hard and successfully into ARM's territory and ARM has yet to hit back with any chip that can compete with Intel in servers or high end laptops or etc. And that's WITH Intel's huge profit margins. ARM certainly doesn't have the profit margin's to spare in any price war. Intel is a huge monster to defeat, and its supposed handicap means far less worry for programmers, unlike trying to support the million and growing ARM SOCs out there and the nightmare that is.
Intel's published prices range from a 'low' $117 for a Core i3 processor.
What about atom? You know, the processor produced by Intel, specifically for the same markets that ARM are dominating now.
Intel has made ARM processors in the past (xScale), and, apparently, still retains an ARM license. Intel has manufactured RISC chips, as well (i960, for example). There is absolutely no reason why Intel wouldn't/couldn't produce an ARM chip, if they wanted to. There's just no reason to do so.
Also, using the Core i3 as an example of Intel's "low-end" is not very fair. Intel's low-end chips are the Pentium and Celeron, not the i3. The Atom is the closest thing to a competitor to the ARM chips. Pricing for Atom chips varies extensively, from $20 to $100, depending on features,
Won't run shit is interesting. With Windows forking into an ARM and x86 (or AMD64/IA64 whatever want to call it) versions, the writing may be on the wall for Intel. If one of the ARM guys can produce chips that will do the 150-200 dollar price bracket as well as Intel chips can on windows this becomes a whole other ball game.
I'm not sure where anywhere near there yet. But with Qualcomm feasting on the remains of AMD, Samsung producing millions of parts a year and a few others with them it's entirely possible that within the next 10 years ARM will be a major competitor to x86. Which is why MS is forking - it's going to confuse the hell out of consumers and is, from an end user perspective a terrible idea to go out and buy a Windows RT anything on friday (windows 8 launch day) but MS plans to support their ugly bastard for a long time, so who knows. And in 3 or 4 years when we see Windows 9 roll around we may have enough software that has been compiled and for and runs on both that your 'won't run shit' assertion would no longer apply.
If Gassée is right about "architectural disdain" then it's kind of ironic. Intel itself exhibited the same disdain for x86 architecture when they initially refused to make their first 64-bit chip, the Itanium, backward compatible with it. It was only after AMD demonstrated that the architecture still had legs that they brought it to the 64-bit world — after wasting billions on Itanium development.
Those that forget history, yada yada.
Apple's the one currently manufacturing their A6 chips for $17, while the comparable Intel chip retails for much more?
Isn't this more a statement of how well Apple's vertical integration of chip manufacturing went?
PS: I don't reply to ACs.
If you measure operations per second, the x86 chip will win. If you measure operations per second per watt, the ARM chip will win.
This is a boring sig
It's been pretty much proven that the "x86 legacy baggage" or however you want to put it does not seriously affect Intel's Atom for phones.
http://www.anandtech.com/show/6330/the-iphone-5-review/10
Razer i, which has an Atom processor, beats A6, the best performer in the ARM field, most of the time in non-GPU tasks (one area it is lacking is GPU power), while power consumption is average for a phone. Android adds additional overhead not present in iOS, too.
If anyone can work miracles and cram x86 into a phone, it's Intel. As ARM designs have to start dealing with greater complexity, Intel can apply their immense experience with x86 and improve performance without dramatically increasing power consumption.
With some more work, I can see Atom beating the hell out of any ARM design in the same power envelope. I'll give it one or two generations.
Intel and AMD x86 processors moved on to using micro-ops and risc like operations internally years ago. The only disadvantage nowadays is a small translator that converts x86 machine code into micro-ops. Compared to the actual logic or cache on the cpu the number of transistors that the translation takes is minimal and not a big deal especially when you consider the size of cpus nowadays.
"When you sit with a nice girl for two hours, it seems like two minutes. When you sit on a hot stove for two minutes, it
Hmmm... According to this article it doesn't come with macros or VBA so businesses can't run their crapola apps on it. That mostly defeats the purpose. Still it's less bad than I expected.
Comment removed based on user account deletion
Here is a 50$ ARM general purpose multicore-CPU example for matching 999$ performance of fastests Intel Core i7 (e.g. i7-3770K 3.9GHz (peak), 4 CPU, 8 threads, 2 SIMD ALU/CPU = 8 SIMD ALUs = 64 FLOPs/clock -> 3.9*10^9Hz * 64 FLOP/s = 249.6 GFLOPS:
For increasing integer and load/store performance, it could be achieved with pipeline and issue/execution modifications, using more functional units. The limit is to keep the OooE simple enough for avoiding wasting transistor in executing tons of instructions unnecesarily.
The old 80386 based on the "complex" x86 instruction set had 275,000 transistors. Intel is now making chips with 2.6 billion transistors and somehow what they once implemented as one functional unit within a budget of 0.000275 billion transistors is holding them back?
Certainly they would rather do a few things differently had they been worried about 2013 back in 1978. Transistor count is the least of the matter. What buggers up x86 is the number of active transistors handling the instruction stream at each instruction cycle. There's no way to align variable-length instructions without active transistors (regardless of whether the transistors involved amount to a wart on a small toe of a juvenile mosquito).
The x86 story bugs the hell out of me. Considered how well it actually held up for 45 years (and counting) it's one of the ugly duckling success stories of all time (hint: it wasn't so ugly after all).
It was also a founding member of the Steve Jobs reality distortion field. I'm concerned his posthumous aura will continue to glow with the uplift of falsehood. He should be credited more for what he accomplished than the lies he polished to get there.
It wasn't just Steve, it was the entire RISC consortium manufacturing an Achilles heel out of whole cloth. Far closer to the truth of the matter is that x86 has a much higher design cost than an orthogonal clean-sheet alternative. The design cost was a small multiple. Intel's resources were a large multiple. It didn't go well for RISC. The much vaunted DEC Alpha had a metal connect layer for single-cycle carry-add propagation that forever segregated it from the mass-consumer price point. It was the instruction set. No, it was the instruction set aided by a titanium stent.
Also, the RISC design advantage does not extend to the memory cache and system bus design. These are a bear to design well for any instruction set. The RISC people moaned about the exceptional Pentium Pro performance level on server workloads (it was the first memory bus from Intel that didn't totally suck). Well, Intel broke into the server market with their crappy old x86 instruction set by grafting it onto a titanium alloy cache hierarchy and bus controller (with multiple dies grafted into the same chip package at enormous expense). Cache latency and branch prediction absolutely dwarf instruction set as the big thing to worry about since around this time. If Steve hadn't grabbed onto the inferiority of CISC around this time, it might have died a timely death.
In low power applications, ARM has a real advantage, enough to win a huge market share at race-to-the-bottom price points. How much does the cost of a CPU influence a handset? How much everything else? I've put $300 Intel CPUs in $2000 boxes. I've put $250 Intel CPUs in $1000 boxes. I've put $60 CPUs in $500 boxes. A $16 CPU in a phone that retails for $600 for just a few months, before landing in the discount bin? I'm sure Intel wants a huge slice of that.
One reason Intel has held their ground is that the Cortex-A15 (out-of-order superscalar multiprocessor) is starting to look a lot like the old Pentium Pro. Sure the instruction set is modern and clean (though it took ARM surprisingly long to come up with the mixed 16/32 bit instruction encoding format due to misguided ideological purity; how many active transistors does it take to determine whether the next 32 bit chunk from the instruction stream is one lump or two? More or less than the number of active transistors in the icache devoted to storing common instructions bloated to 32 bits just because?). But all the rest of the issues are pretty much the same: branch prediction stalls, cache snooping, and memory path latency.
From Intel's perspective, an ugly instruction set is good for business. (Then they went on a jag thinking that if ugly is good, atrocious is better, and the Itanium was hatched with a jackhammer from a mastodon egg.)
After another three die shrinks, when half the processor implements on-demand power management, and most of the other half provides task-specialized execution units, is the instruction set going to matter a hill of beans for anything other than legacy lock-in?
It depends. Is my carriage playing Modern Warfare 3 or Angry Birds?
What a bold prediction, you understand of course that Intel has buried every single competing architecture from the past? Intel has a process advantage, even if they have to spend 10% of their die on decoding/rearranging they still have a significant transistor lead by remaining a process ahead AND still use lower power. Not only that but because x86 is nothing more than an abstraction layer at this point the internal architecture of their chips is free to move with the winds of computing in the best direction for the balance of power use, processing capacity and weight. They've had almost 2 decades to improve this abstraction layer to the point of perfection.
People like you forget how long it takes to design and build a microprocessor. From design to hard silicon is almost 5 years. So the designs Intel releases this year were planned out in 2007. Given the ARM didn't start to make an impact (on Markets Intel considers themselves part of) until 2006-7 we are JUST starting to see an Intel design philosophy that emphasizes power as a critical function. Haswell is probably the first chip that Intel hasn't tried to tack power efficiency on add-on at tape out. I fully expect Intel to demonstrate that x86 under their lead has the ability to compete directly with ARM on their best footing, power consumption.
So watch and learn young padawan. Intel has the best process engineers in the business and if things in the foundry business keep going like they are (TSMC and Global Foundaries have both been very very late moving forward on process while Intel hasn't missed a stride) they are going to be two steps ahead on process in the next year or two and that would be an advantage not even the best ARM design could beat even if Intel bungles their design. I fully expect that if Intel wants it they could take the whole ARM chip market. The only reason they haven't up till now is it would destroy their margins. So we will watch them balance their designs to retain the high margin products and forgo the cheap. This could ultimately be their undoing but once power efficiency becomes a priority of their designs which begin with Haswell, Intel will be in a position to take the ARM chip market any time they want.
Don't ever discount the power of the foundry.
Intel wants to be the only company that can meet your needs. That way, they can make you pay premium prices for their chips. This is perfectly understandable; that is what is best for Intel.
Apple wants to be vertically integrated. They want full control over everything they do. Partly this is so they can keep as much as possible of the money they collect; partly this is so that they can guarantee excellent quality and excellent availability. This is what is best for Apple, and it isn't bad for their customers either.
Intel does not want to become just another ARM source, competing on price with all the others. But Apple will never lock themselves in to depending on Intel for mobile chips, when ARM chips have been shown to be more than adequate. And Apple would not be investing in custom ARM chips if it was planning to adopt Intel mobile chips.
People keep pointing out that Intel's mobile x86 chips are competitive with ARM. That won't cut it. Intel's chips would have to be better, and so much better that the risk of depending on Intel is worth it.
That was the case for the PowerPC to x86 transition! Intel's chips were so much better than PowerPC for laptops that it was worth getting into an entangling relationship with Intel. AMD was not able to guarantee delivery of the massive quantities of chips Apple was planning to sell, and Intel was, so AMD wasn't really an option... but at least they served to keep Intel from trying to charge totally outrageous prices for their chips; there was always a credible threat of going to AMD.
Hmm. It's looking like AMD is going to crater in spectacular fashion soon. I wonder if Apple will make a serious attempt to buy what's left of the company. That would enable Apple to make its own x86 chips! Eh, probably not. AMD is behind Intel on process, so switching to AMD chips would mean taking a hit on performance, power use, or both.
The "SemiAccurate" web site thinks that Apple will transition to using ARM chips for laptops, not just for mobile devices, once ARM chips are good enough (which they will be soon). So, transitioning away from x86 and to, say, multi-core 64-bit ARM chips is another way Apple can untangle from Intel.
Apple may not be in a big hurry to actually complete the transition away from Intel chips; just a credible threat of switching to ARM chips might be enough to negotiate good prices on x86 chips. That would leave lower power consumption as the main reason to go to ARM, but a laptop's display is probably the worst power drain, especially with a Retina display.
steveha
lf(1): it's like ls(1) but sorts filenames by extension, tersely
Also note that Apple has people paying $2500 and up for the Mac Pro, and $1000 and up for laptops. But mobile devices are closer to $500, and the Android competition is hitting the $200 price point.
There just isn't as much room to pay top dollar prices for Intel parts in the mobile space.
So even if Intel mobile x86 parts are slightly faster than the ARM chips, will Intel be happy selling at prices competitive with ARM prices? History suggests "no". The cheapest Atom chips are around $20 but Intel makes those suck, just as much as Intel can get away with.
Intel is the master of segmenting markets. Different chips at different price points have different features enabled. Cheaper chips are as crippled as possible, to encourage you to buy a more expensive chip. For example, Intel doesn't support virtualization features on their less-expensive chips; and Intel mostly reserves support for ECC RAM to only the Xeon processors.
(In contrast, AMD puts full functionality in all their parts; they are #2 and they are trying harder to please the customer. That is how you can get an HP Proliant MicroServer with a 1.5 GHz dual-core AMD Turion processor for $320 at Newegg, with full support for virtualization and ECC RAM. I cannot imagine a MicroServer with equal or better Intel parts hitting that price point.)
Intel will try to balance the functionality it allows into the mobile chips against the price it can get. Apple just wants the best chips for the cheapest price. These two goals are not in alignment.
lf(1): it's like ls(1) but sorts filenames by extension, tersely
What a bold prediction, you understand of course that Intel has buried every single competing architecture from the past? Intel has a process advantage, even if they have to spend 10% of their die on decoding/rearranging they still have a significant transistor lead by remaining a process ahead AND still use lower power.
We'll see if that really will happen. But it's very true about the process advantage. If Intel chose to build and ARM CPU, it would blow the doors off all the other ARMs because they can build faster/smaller/lower-leakage transistors. But it might be more expensive, and the ARM market is very price sensitive. I imagine there are Intel people who have considered this carefully and concluded that they can't build an ARM processor that's cheaper AND better than the competitors.
You are forgetting 2 non technical matters.
A: Why let Intel 'win' so they can turn around and stuff higher prices clean up your backside? One supplier markers are best for the supplier.
B: Anti-trust. Once you start taking over entire markets all of a sudden you have even more government flashlights up your ass. Only a dumb person wants total control unless he has TOTAL CONTROL, otherwise powers stronger then you take too much interest in your daily work.
Intel is not playing the same game. With everyone else (except AMD) making ARM devices on older process nodes, Intel should not make ARM chips because that would create the perception of competition and force TSMC and GF to advance their process. So long as all the foundry customers appear to be competing with each other it looks like a close race and there may be less pressure to advance. The further Intel stays away from their products, the less those guys will feel like they are competing with Intel and they will not worry about the process gap so much - they're still close to their "competitors" capability after all.
Everyone seems to have forgotten what business they're in. Those who can design have gone fabless while those who can fab now have more than enough customers to not care about process advancement so long as they can keep up with their perceived competition. In fact, all those customers probably slow them down with countless designs that each need scheduling and a design tweak or two. Meanwhile Intel turns that crank every 2 years. The longer people forget that they're all in the same business, the wider the gap is going to get.
Meanwhile I wonder which ARMs even have instructions like divide or reciprocal square root.
Anon Intel Employee here.
Very. Fucking. Paranoid
Re-look at some of the acquisitions from a "If we buy it, you can't" perspective and see if they make more sense
It doesn't give you a smaller desktop. You have an app for your phone, a different app for your desktop, and yet another app for your tablet. Windows Phone and Windows RT are orphans. Do you see what I'm saying now?
Yes, I think I get what you're saying. Thank you for your clarification; perhaps I can make some of my own.
When iOS was first released (2007) it didn't run any sort of legacy programs from existing touch screen smartphones, and yet people really seemed to enjoy using those devices. When Android was first released, and was still just a phone OS, people seemed to enjoy using those devices. When the iPad first came out it could run existing iPhone apps; but even Apple said that the experience for many apps wasn't good and that app authors needed to (and still do) optimize their apps for the larger screen. There's hardly an article on the web about Android tablets that doesn't mention how there's a pithy of tablet optimized apps, and that while the tablets can run all Android apps, most of them really suck on the tablet. So I don't think that not having the ability to run legacy programs is a nail in the coffin for any new device/platform.
So right now a Windows RT computer does have limited appeal, due to not having much of an ecosystem. That was kind of my point by saying "3 years from now", the ecosystem may come. The ecosystem may come easily because if developers write apps for the Windows 8 Store (targeting desktop) the apps will also light up on Windows RT; and they will light up without any "tablet optimization" step that iOS and Android apps suffer from. The screen sizes are the same. So I think that there will be crossover soon, at least from the point of view of the end user; because they'll be able to find the same apps on a Windows RT computer as they do on their Win8 laptop/desktop.
If RT could run Phone apps, I wouldn't be typing right now.
Yes, Microsoft doesn't have a runtime that runs across Win32, .Net, WinRT, Phone and Xbox. But they do have portable assemblies which do allow for an assembly to run within .Net, WinRT, Phone and Xbox. So someone would still need a separate app/program for the different runtimes, but if the business logic is the same for all of the apps there only needs to be one portable assembly.
Except that for ARM to scale up to surpass x86 performance, it would need out-of-order scheduling, multiscalar execution, instruction pre-decoding, hardware loop unrolling, speculative execution, memory/register renaming, multiprocessor cache coherency and all that fun stuff that Intel has. After that has been implemented, let's see if the design is still far more elegant. I'll bet it's going to look pretty similar, with x86 instruction decoding becoming an almost irrelevant issue compared to all the other things. And, by the way, have you taken a look at the Cortex architecture ? It's getting less and less elegant. You are right about the compatibility, though. ARM is a mess, with a dozen different, incompatible, architectures in the core alone.
That 10% becomes a lot more important going forward. The current buzzword in the semiconductor industry is dark silicon. To keep within the same thermal envelope (power dissipation per unit area), you need to have more transistors idle and in a low-power state in every subsequent generation. If you add complex vector instructions, for example, they're great because they give a big speedup when they're in use and draw almost no power when they aren't. The same with things like AES encryption. The instruction decoder, however, is something that you can't ever turn off. Xeons try to: they cache decoded micro-ops in tight loops, but this means that they have some extra SRAM for the micro-op cache and a micro-op decoder that must always be powered, and these between them take more power than an ARM decoder, and a big fat decoder that must be active all of the time.
Intel had an advantage over other RISC architectures (and Itanium) in terms of instruction density, which meant that they needed to waste a lot more die space with instruction cache than x86 to get the same fetch performance, but ARM is already about as dense as x86 and Thumb-2 is typically 5-10% denser, so Intel is on the losing side of this comparison for the first time.
The process advantage is something that Intel has had over AMD, but it's not something that they have to the same degree over some of the foundries that produce ARM chips. They're on 22nm, and the faster ARM SoCs are made on a 25nm process: that's nowhere near the kind of process advantage Intel is accustomed to. In terms of fab R&D, the industry is almost split into two camps, Intel on one side and everyone else pooling resources on the other side.
I am TheRaven on Soylent News
Haswell will be available for $20 in its underclocked ultra-low-power configuration?