Can ARM Kill Intel?

http://www.electronicsweekly.com/blogs/david-manners-semiconductor-blog/2010/11/can-arm-kill-intel.html

by ssp • 16 years ago

48 59 16 years ago

electronicsweekly.com

lovamova• 16 years ago

I wouldn't be surprised if Apple will make Mac OS 11 run on ARM only.

roc3• 16 years ago

I would -- that would be a terrible idea.

"Merging" the desktop and mobile-touch universes is a fool's errand. They shouldn't be unified for the very same, very good reasons that the iPhone was served by a fork in the first place.

Frankly: you're looking for a fundamentally different experience at the desk top than you are on a mobile. And ARM is a bad choice for the types of tasks that people are still buying traditional PCs and Macs to do.

runjake3• 16 years ago

It may be a fool's errand but that's the way Apple is going. And I do not see them having any lack of market success with their approach.

I don't think it's really a matter of desktops becoming like mobile devices. I think it's "desktops as we know them will be going away (except in specific use cases)".

My wife's Macbook has seen better days. When it finally dies, we're going to replace it with an iPad, a bluetooth or docked keyboard and a camera connection kit. It does everything she needs. I imagine this will be the same when my kids get old enough.

And why is ARM a bad choice? Have you heard about performance complaints from iPad users? Have you seen Rage HD on iOS yet?

lyime2• 16 years ago

Try running Photoshop CS5, Starcraft 2, AutoCad on ARM. I am not saying it's not going to happen. ARM is focusing on the embedded market. They are not going to be able to compete with the likes of a Quad Core Nehalem or Sandy Bridge.

rbanffy5• 16 years ago

ARM processors have, so far, been targeted to mobile, low-power applications for a very long time. When it was launched, powering a desktop computer (the Acorn Archimedes), its speed was comparable to then current 386s with FPUs.

There is nothing that prevents building an ARM processor that performs as well as a current x86. In fact, due to its simpler implementation logic, an ARM implemented with the same technology as a x86 would probably run faster or leave more silicon for caches, inter-core communication, more cores and specialized logic.

mooism2• 16 years ago

Not while anyone wants to run Windows in a VM, they won't.

jawee• 16 years ago

If that were the sole issue, I can't imagine getting a hardware virtualization solution like QEMU up to par would require too much work.

nextparadigms• 16 years ago

You can bet Microsoft will make Windows 8 ARM compatible. They don't want to lose the tablet market and since Intel is having such a hard time getting into that market, they can't simply rely on them.

And if you're going to suggest WP7 for tablets, don't. Microsoft will only do that in worst case scenario if Windows 8 will prove to not be good enough for tablets. They don't want to get $10 per license when they could get $50, and that's especially true if touch/mobile computing will replace desktop computing.

cosmicray• 16 years ago

When you talk about Apple, and you talk about ARM, remember that Apple bought two companies: PA Semi and Intrinsity (to get chip design talent). The PA Semi acquisition came with IP for a POWER chip, the PA6T. Thus far Apple has announced nothing using that chip (and possibly never will).

Apple certainly has the ability to make OS X run on POWER, but doing that might people fall over from dizziness.

gonzo5• 16 years ago

Daniel W. Dobberpuhl (the founder of PA Semi) was also the architect for StrongARM (when he was at Digital).

Intel only got involved because DEC sued Intel for patent infringement (over the Pentium processor), and lost. Intel grabbed a lot of DEC's technology in the settlement, including StrongARM (which became XScale).

vimes656• 16 years ago

In the last 2 years we have seen many ARM netbooks prototypes showing better battery performance and cheaper price than Intel netbooks. But I still can't buy an ARM netbook. Where are they? What is holding them back?

msbarnett• 16 years ago

Lack of an operating system with mass-market appeal in the netbook space, as far as I can tell.

rbanffy5• 16 years ago

I assume the Windows OEM prices have been tuned so that it's more expensive to develop and manufacture ARM-based netbooks that run Linux than it is to stick with x86 for the current generation.

Also, the processor is not responsible for a huge part of the power consumed on a netbook. Disk, screen and memory seem to be a big part. If you make a netbook that can run Gnome and Firefox decently, it will have a hard-drive (or SSD) and a couple gigabytes of memory. The LCD is the same.

graphene• 16 years ago

heh, just got mine delivered today, and fired it up just to post this reply :-)

It's a Toshiba AC 100 with an ARM-based Nvidia Tegra processor and Android operating system in a standard netbook form factor. I don't believe it's available in the US though, if that's where you are.

First impressions: it's very light, to the point of wobbling around when I type with it on my lap. Many of the peripheral keys are very small for my Western fingers, although I feel I'm getting used to them already. It's fanless, and has only flash storage, so it's eerily silent and apparently gets up to 8 hours of battery life. The trackpad doesn't have enough vertical space to be comfortable, and there are neither function keys nor a delete key, although there are a whole row of special keys which presumably can be xmodmappped.

The online reviews mention that Android provides a less than elegant interface with the lack of a touchscreen, and I agree; I was planning to install a Linux or BSD OS on it all along. As I understand it, that's pretty tricky; should make for a nice weekend project :-)

As for why there aren't more of them around already, my best guess is that OEMs don't want to piss off Wintel by selling what is effectively a PC which doesn't use their platform, for fear of losing the discounts and rebates they get when they do. I hope I'm just cynical and there's a better reason; for casual use the system resources definitely seem adequate. Acutally, as soon as I get a proper OS and WM installed on this thing, I can easily see it taking up half of my computer time, and it will be brilliant for having a quick hack on the go...

[edit]

After the first half hour or so of usage, it's clear that the Android as currently installed is definitely not a good fit for a non-touchscreen, full-keyboard device: After 5 seconds of idleness, the screen dims, then blacks out completely, just like a phone would; navigating through screenfulls of icons using keyboard or mouse becomes tedious very quickly, and to top it off I've run into a few bugs as well.. This thing is begging for a linux/bsd makeover, hopefully I will uncover a real beauty...

GHFigs• 16 years ago

Netbook makers are getting too good a deal on Windows to switch back to shipping flavors of Linux, at least for the US market. There is far more demand for "cheap Windows laptop" than just "cheapest laptop", if that makes sense.

That said, the rumor is that Chrome OS is only going to ship on ARM, and will do so Real Soon Now. I'm not convinced that Chrome OS is going to be a hit, but I can see more people being willing to give up Windows for it than for a typical netbook Linux distro.

InclinedPlane• 16 years ago

Amateurs discuss tactics, professionals discuss logistics.

Intel has so much FAB capacity that they can't easily be killed even by superior competing products.

grogers• 16 years ago

They also usually have the smallest process technology of the time. If necessary, Intel could fab ARM chips and dominate the ARM market from the inside.

mjschultz• 16 years ago

Precisely what I was thinking. Does anyone know if there is a reason Intel hasn't just acquired ARM? I assume Intel believes that they might still be able to compete and win in the market, but that seems to be less true every day.

(I assume the answer is that it would raise a bunch of anti-trust flags everywhere, but I'm not positive about that.)

gonzo5• 16 years ago

Not without paying ARM Holdings, they couldn't.

epynonymous3• 16 years ago

agreed, i don't think the combined likes of samsung, ibm, toshiba, sony, amd, etc have more superior fabs than intel.

westbywest• 16 years ago

Also, considering the sheer inertia in the economics of the larger fabs like TSMC's and Intel's (and by "large" I'm including both quality of product and production capacity), having an excellent silicon design in no way guarantees you market leverage. Having an excellent design and optimized arrangements to fab it quickly in large numbers do.

yread• 16 years ago

Plus they have a profit margin of 25% on revenue of 40B. So they could be selling everything 20% cheaper and still earn money (and lots of them). And that's including all the settlements Intel had to do this year. It's just a big Chipzilla

crander• 16 years ago

I prefer to think of this in a Christiansen disruptive technology context. Technologies grow upmarket beyond the needs of its user base and create a disruptive opportunity for downmarket technologies.

I recently made a comment about this on an Intel blog article (SC10: Jeff’s Notebook - Innovation and Disruption: How a Successful Company can be gone in 10 Years): http://intel.ly/dZAxKk

meastham• 16 years ago

Has ARM's design licensing model ever been demonstrated in the high end (server/workstation) CPU market? It seems to me that the amount of engineering expertise required to push that sort of high performance, fully custom design through to fabrication would make it difficult.

andrewcooke2• 16 years ago

Isn't that how SPARC worked? Of course, you can argue it failed after a while... https://secure.wikimedia.org/wikipedia/en/wiki/SPARC

adulau• 16 years ago

ARM is not really friendly with competitors especially those not paying the high license fee. Do you remember the nnARM project at OpenCores? this has been removed on request by ARM limited. If ARM wants to "own the market", they should allow in some ways competitors to build compatible CPUs.

rbanffy5• 16 years ago

> they should allow in some ways competitors to build compatible CPUs.

They allow you, or anyone, to build compatible CPUs. You just have to pay for the privilege.

lyime2• 16 years ago

"People in the mobile phone architecture do not buy microprocessors," says Hauser, "so, if you sell microprocessors you have the wrong model. They license them."

Someone is building them. In iPhone's case its Samsung. The phone's Apple-designed A4 processor, which was built by Samsung, costs $10.75 each. Most of that is going to Samsung, part of that goes to ARM. The ASP of ARM 10-20x lower than that of Intel.

gonzo5• 16 years ago

Someone seems to have forgotten that ARM means "Advanced RISC Machines, Ltd" and that Acorn, Apple and VLSI were partners in A(dvanced) R(isc) M(achines).

Of the three (Acorn, Apple and VLSI), only Apple is really still around. Apple must already have significant equity position in ARM being a founder.

VLSI was bought by Philips Semi, which was later spun out as NXT. NXT is a user/licensee of the ARM architecture but not even the biggest one by a long shot.

Intel's revenue from ARM (and yes, they still make some Xscale parts) is probably bigger than all of NXT's revenue.

Acorn went belly-up as a going concern in 2000 with the remains bought by Morgan Stanley.

The ARM "JV" IPOed (as ARM Holdings).

In any case, I doubt that Apple pays much, if anything for its ARM license.

And I doubt we really know what Apple pays for A4 parts.

Nelson69• 16 years ago

I don't see how intel is even a target. PowerPC owns gaming. There are still tons of MIPS chips in networking applications as well as some other verticals. ARM owns phones, largely because every phone has a custom chip in it. It's not clear if it will grow beyond that.

Why not suggest ARM will defeat PowerPC or MIPS? That sort of seems plausible, Intel is so insanely good at what they do and they backed out of ARM a few years ago. I can't imagine they did that without understanding how they'd compete.

epynonymous3• 16 years ago

i think you're talking about console gaming perhaps, xbox and wii use powerpc variants, but ps3 uses the cell processor and nds uses an arm processor. in terms of mobile gaming, powerpc is nowhere to be found, this is owned by arm. in terms of pc gaming, intel and their we're-not-totally-a-monopoly division, amd, own that realm. though arm is starting to make inroads into low powered servers so it'd be interesting to see where this goes in the long run.

krakensden• 16 years ago

Cell is a PowerPC variant.

http://en.wikipedia.org/wiki/Cell_(microprocessor)

davidj• 16 years ago

Intel builds ARM processors. This article sucks.

eru• 16 years ago

Sources?

a-priori2• 16 years ago

More accurately, they did build ARM chips -- the XScale line. But they sold that division to Marvell in 2006, although they still retain an ARM license.

http://en.wikipedia.org/wiki/XScale

repiret• 16 years ago

http://www.intel.com/design/iio/index.htm

Although I doubt they'll make them for much longer.

hackermom• 16 years ago

ARM could probably kill x86 given some time, if that was what the writer implied - and let's just wait a few years and see how things currently brewing will develop before we start shooting off our downvotes, shall we? I am willing to bet that an often-blamed-for-being-fruity computer manufacturer will shortly migrate from x86 to ARM. Just wait.

Killing Intel themselves, though, just won't happen. If anyone can dominate the CPU market wielding ARM in their hand (haha!), it's Intel. The rest of the world has a good few years to go before catching up to their level in terms of manufacturing.

epynonymous3• 16 years ago

from the article, "stifled innovation", "monopoly", i wonder how true that is, where would we be if any of this was true, would we have a more awesome chip in our laptops today?

i'm pretty excited about gpu's, but they're only good for parallel instruction processing, some tasks still require the usage of a highly spec'd cpu.

whatever happened to transmeta?

jaen• 16 years ago

In the server markets, you look at performance per watt and performance per rack unit and ease of setup - for general purpose computation, Intel still dominates those metrics. (with some exceptions [1]). Being ahead in fab technology, it can take the performance hits due to the legacy x86 architecture.

You can replace 1 beefy server with 8 small nodes, but you can not replace the convenience of administering that one server.

Also, retargeting performance-sensitive JITs/compilers/maths libraries such as GCC, LLVM, Java, V8, *BLAS is no small feat either - x86 gets support first, because it is the market leader, and also gets tuned the most, since it's the oldest backend. A feedback loop that's hard to break, so don't expect Intel to go out of business any time soon.

[1] http://www.cs.cmu.edu/~fawnproj/

Can ARM Kill Intel?

Related Stories