ARM to bash 'non-issue' Intel with multi-core chip ARM’s big reveal today of its latest processor design came with the clear message that the company will use its market weight and mobile heritage to fend off challenges from Intel. Speaking here at the ARM Developers Conference, CEO Warren East began the hyping exercise around the Cortex-A9 processors. ARM plans to provide …
It's a lovely concept. Taking on the UMPCs, which do - in fairness - suffer from hot and hungry CPUs (or just Not Very Good Ones) is a very worthwhile task as the market for these devices matures/converges with PDAs.
But what OS is going to power these? The UMPC's benefit to the consumer is easy, instant access to mature and well-developed (or at least, functional) applications on an OS that they are familiar with. Will the multi-core chip be supported by derivatives of Windows Mobile?
In which case, will we see manufacturers migrate from what is already becoming an expensive platform, where unsubsidised hardware like the HTC Advantage costs £650 for a 256MB ROM, 128MB RAM, 8GB HD device with a 5" screen and a 624MHz CPU that spends most of its time at 102MHz in a struggle to get an 8hr runtime. The same money will, whilst losing the telephony capability, get a 1.2GHz VIA C7-M powered Ubiquio 711 with 1GB RAM, 40GB HD, and Vista which will do about 3 hours of snappier, more versatile operation. Neither device really fits in a pocket for most users, after all.
Will we see Linux, perhaps? Devices that are slightly unfamiliar to the mass market, and at a fair guess will take two to three years to be developed and marketable (remember how long it took for UMPCs to be developed and ready for launch).
Symbian? Palm?
By the time this CPU is ready to hit mass market in a UMPC-style device, Intel's 45nm chipsets will be there, LED backlights will be cheaper, HD density will have improved, and who knows, maybe MS will have managed to drum up some enthusiasm for Vista. Where ARM's clever little chips will of course continue to make the world tick and people communicate, I think that Intel and Microsoft's approach to the PDA will ultimately succeed by sheer brute force, by giving the consumer the ability to carry their desktop not in terms of "productivity" but in terms of actual functionality and familiarity. OQO's Model 03+ will be sporting a 1.5GHz variant on Intel's A100 chipset, 2GB RAM, a 160GB HD and an HD screen for £999, FlipStart will be selling the V2.0 with a Core 2 Solo and 80GB HD, and Ubiquio's OEM sources will be mass producing OQO performance in a bigger box for half the price. Sony will probably be shipping 1.86GHz Core 2 Duo systems for £2000, and all of them will have HSDPA or equivalent comms.
Of course the ARM chip would offer theoretically better performance for a given clockspeed, or better battery life, and would be vastly more efficient with a decent OS on top of it... but it's not going to change what the market wants. And the market is still unsure it wants anything at all, but when it does figure it out, it wants "desktop, in pocket".
iPods don't need multi-core. They work perfectly well as they are, and that's not a "no-one needs more than 640K RAM" statement - once the iPod needs that kind of CPU power, it is becoming a new device.
There is one mass market device I can see, and it's mentioned here. If Apple, instead of as rumoured shifting to Intel, deploy this multi-core CPU in an iPod Touch, with more real RAM and storage memory, and a suitable port of OS X on ARM... then perhaps Apple could develop the PDA that does become the device everyone wants.
Before that, I think we need iSync and .Mac support on Windows, and open development on the platform. The question everyone asks is "WIll it run MY apps". Until the answer is yes, very few consumers are interested and only the geeks and those truly new to computing will attempt to work around the limitations to gain access to the hardware. For an Apple PDA to truly succeed, it has to be available to both Windows and Macintosh users.
..of getting way better than ARM and competing products.
It really seems that no one among competitors management has noticed but Intel it's not the same it was during the P4 Prescott (and failed Tejas) power hungry core era. Since the team behind Centrino got full Intel resources and focus inside the Corporation, everything changed. x86 Intel products are the best ones on the market and 2008 seems to be full of improved products able to deliver way higher performance with way less Watts.
So I really think that Intel might have done a smart thing selling its XScale ARM division and starting from scratch with a new ultra low power set of x86 products. We will see if the team behind Centrino technology has been able to design a better architecture than what XScale was already capable of.
I'm an ARM shareholder and although I don't see low-power x86 taking away all ARM's markets it will be a big long-term threat for at least half the market.
As small devices have increasing amounts of memory and storage it makes sense to bring them into the x86 line where the three mainline desktop OS's are. For hardware companies it gives you access to a larger set of possible apps and developers for things like in-car entertainment, PDAs, games terminals, point-of-sale devices, home tablets etc.
The OS/App companies are already thinking about different ways of delivering applications without the bloat install problem - such as application streaming - www.stream24-7.com, or SaaS, or browser delivered apps. Bringing mobile devices into this world long-term makes total sense.
And if you read and understand the register for a few weeks you stand a very good chance of actually getting a clue why you've missed the mark by a country mile.
Re: Operating System.
Nobody said that the multi-core processor _has_ to be used purely symetrically.
The way multimedia operations are handled is changing, a few years ago a so called "MPEG4 Codec chip" was actually a traditional CPU with a SIMD DSP strapped to it.... With improvements in technology we now do the whole thing in software on a normal CPU. Some of those cores may be used for nothing more than running the current codec, and communicate with the OS via semaphores!
There is actually a pretty large choice of OS for ARM-based PDAs:
Windows Mobile, Symbian, Palm OS, Linux, netBSD, iPhone OS X, RISC OS, ...
RISC OS is not currently used on any PDAs, but it can easily be adapted to do so (it has been used on a touchscreen tablet PC, which isn't far from a modern PDA).
The DSPs are cheap, are the multi-core ARM processors going to be as cheap? I see where you're coming from, but I don't see the benefits over the established hardware solutions to be perfectly honest. You get a single chip which has the ARM core, DSP, RAM - the whole lot - and costs very little whilst doing precisely what you want it to.
I can see it being useful in phones in a similar way, but I don't know enough about how the technology/loads are balanced - how much is software, how much is hardware, in a typical multimedia capable phone.
As for Centrino, I just see it as building a better hammer. Like I said before; brute force to produce what the customer wants - if the desktop OS isn't running quickly enough on a mobile device, make the mobile device faster. It's not the solution I'd prefer (I used to be an Archimedes user... still have about 20 of 'em) but it is the solution the market will lap up.
Windows Mobile already supports ARM9 etc. so why shouldn't it support these processors?
I believe vendors are growing wiser in regards to writing portable software anyhow these days.. and then there's Java and ARM's Jazelle etc.
As far as processing speed/power consumption goes, I believe it is the ARM instruction set that goes a long way to helping ARM chips perform the way they do. Ultimately, I don't think the x86 architecture will hack it in the race for high-speed/low power because it will always be constrained by its roots, where this scenario was not really on the agenda.
Matt.
"Ah, but perhaps ARM will relent on its anonymity a bit with the Intel threat looming. Why not shove an ARM logo on every iPhone that goes out the door?"
I suppose the Plan must be then to wait until "Beta Inside Intel" markets ITs Wares.
A Processor Names War Postponing Advanced Enriched Intelligence Use is not Logical.
@call me scruffy: And I mean it. If you think that general purpose CPU software performance will make DSPs obsolete then you are completely wrong. Intel has designed its upcoming new architecture with the CSI/Quickpath bus to allow for 3rd party DSP/multicore-DSP co-processors/accellerators to be installed on motherboards.
The whole Larrabee high-end Intel GPU project is based on micro-x86 cores that work in a way similar to IBM/Sony Cell SPEs...
And if you think that specialized DSP products from Texas Instruments, Analog Devices and other manufacturers are going to get obsolete and a general purpose ARM o x86 CPU will do better, then you are wrong.
@Alistair: .. do you really think that Intel sold its own ARM XScale division to waste money ? The team behind Centrino surely knows what it's doing, so far they proved to be real professionals delivering top-notch products in a shorter time than ever before. So unless they do some major mistake I bet their upcoming ultra-low-voltage CPUs will be a tough competition for ARM manufacturers.
As Matt Butland pointed out, there are very fundamental reasons why anything pretending to be x86 will eat batteries for breakfast.
Intel flogged it's XScale division because it had an "epiphany" about what type of market it was targetting. That was discussed on el-reg about six months ago.
"Ultra Low Voltage" is a matter of fabrication technology, not processor architecture, an in order to displace ARM it's
a) got to offer more _effective_ instructions per second, per watt,
b) got to be cheap
c) symbian has to be ported to x86
As for TI DSPs, I suggest you go and have a good look at OLAP, oh look! An ARM core in the middle of a TI die!
Single cycle mutliplier, low branch over-head, SIMD capability... what more do you actually WANT in a DSP?
And just ask the PFY... Archies Rule.
Correct. Lest ye forget the StrongARM and its offspring, the XScale, which were developed by DEC Semi from older ARM processors. Intel inherited these beasts when it bought DEC Semi, much to the annoyance of Acorn users that had SAs loaded, suddenly to find.... Intel Inside! ARRRRGH!!!
Indeed intel has upped its game considerably in the last 2 years, and while many devices run on the ARM processor, most of these devices are very limited in their functionality. Intel is moving into new markets, as is ARM, they are wanting to be involved in more than just phones and pocket pc's and no doubt they will be excellent. It is an excellent idea to deliver the x86 instruction set to small devices, and it will give manufacturers the power to offer people various operating systems on their product instead of one customised one. How hard is it to replace the operating system on your smart device? It is this flexibility that will really appeal to developers and other computer enthusiasts, and it will drive innovation in this market in a way we haven't seen before. Plug and Play on your phone. The ability to port programs directly from the desktop. PROPER browsers. Soudns great.
I can't wait for the first x86 phone/UMPC, cos I will be queing up to buy one.
C
I don't see that ARM will lose all it's markets. Mobile phones and peripherals markets where 3rd party apps aren't as important and cost and wattage issues dominate. Multicore makes sense given disparate data input streams, I don't see Intel getting into these, at least not without serious improvement in battery technology.
But there are plenty of mobile applications coming up (and current applications which could be used more mobile) where power consumption is valuable as it allows portability, but not as important as having access to desktop quality apps type functionalitiy. For instance, in the web-tablet market, ARM-based processors lack oomph and they need a dedicated software fork to keep the browser up-to-date. Much easier to use x86 and plug in standard Firefox.
And having multiple OS options, is not the same as having compiled out-of-the-box applications ready to stream or run on a portable device. Multiple OS is a valuable option to manufacturers as it allows development flexibility and control, but not to users who just want to flick and go.
It's not to say you won't have an x86 core with ARM-containing support chips in key peripherals such as DSP but it's the apps that drive the CPU, not the other way about.
The line about "I don’t actually believe there is any struggle to be had really,” he said. “It is almost no contest. We are an order of magnitude ahead. We have more than 200 semiconductor licensees"
I find rather distressing. I remember many years ago, when I worked for IBM (so that's over 10 years ago at least) the stories about the PC CPU choice.
Journalist to IBM - "Why did you chose the Intel part over the Motorola 68000 line?"
IBM - "Motorola didn't seem to want to sell it to us."
Journalist to Motorola - "Why didn't you market the 68000 line to IBM for the PC?"
Motorola - "We didn't think we had to."
The rest is history.
I like competition, I hope ARM do well. I also like the idea of using less power ( and that should be 'fewer' watts, not 'less ) since power does tend to imply release of CO2 somewhere along the line.
There are many MANY more devices out there that are getting processors in them - washing machines, cars, PVRs etc than the whole market for smartphones.
And ARM has a huge market shares in those areas because the processor is cheap (we use a PXA270 running embedded linux), and does the job well. In fact we are not worried about power consumption, but still went for the Arm architecture as it was the cheapest option in our case. Its difficult to see an X86 architecture device being as cheap to make as an ARM device, and that could be a real problem for Intel.
James
It makes perfect sense that people would want to shove x86 into the the embedded area and bust a billion guts in the engineering effort to make it happen.
But x86 is junk, so why bother?
And x86 binary compatiblity is irrelevent, because we are not living in 1979. As Alisdair Rawsthorne (the boss of Transitive Technology) once said to me: "Let me put it to you that Instruction sets are not the future."
ARM is the proof of that: most ARM processors run at least 2 instruction sets and the future ones will run at least 3 (ARM / Thumb2 / Jazelle). While Intel spend zillions on making creaking 70s ideas work in the 21st Century, Arm spend orders of magnitudes less on defining computing's true future: One that uses less money, less energy, less CO2.
So, scrap the obsession with x86, let go of the past that hinders so much! Embrace the future with every Arm available!
We might see a replay of the Poqet PC story: a near-pocket-sized monochrome MS-DOS system running out of some combination of ROM and non-volatile RAM (before there was flash memory, there was battery-backed RAM in an early PCMCIA format). It ran on either AA or AAA batteries (I forget which).
The Poqet was very clever, with suspend/resume functionality that nobody else had as far as I was aware. Fujitsu bought it, Windows came along and the Poqet died.
The analogy I want to draw here with ARM is that highly-optimized niche players are highly vulnerable to technology shifts, whether in hardware or software. Unless ARM can bring some other things to the table like (to pick an example out of the air) greater resistance to EMP that will viably differentiate it, it may well go the way of Poqet regardless of engineering excellence.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
