Linked by Thom Holwerda on Wed 5th Jan 2011 22:09 UTC
Windows And this is part two of the story: Microsoft has just confirmed the next version of Windows NT (referring to it as NT for clarity's sake) will be available for ARM - or more specifically, SoCs from NVIDIA, Qualcomm, and Texas Instruments. Also announced today at CES is Microsoft Office for ARM. Both Windows NT and Microsoft Office were shown running on ARM during a press conference for the fact at CES in Las Vegas.
Thread beginning with comment 456143
To view parent comment, click here.
To read all comments associated with this story, please click here.
Ravyne
Member since:
2006-01-08

Totally agree there. It's natural that everyone get excited by announcements like this, but ARM is still ARM - its got a lot of things going for it but has a LONG way to go to compete with x86 on pure performance.

ARM is good enough at what it does that I think it could easily become a serious player for systems where power use is critical, but it will have to undergo quite a lot of changes to compete on pure horsepower.


ARM may not be directly competetive now in performance, but they've been under 20+ years of evolution towards low-power, embedded applications. A single, current-gen ARM core alone draws maybe 500mw at load. Intel's most frugal Atom draws, IIRC, 4w at idle and twice that under load. You can lay down 16 ARM cores in the same thermal envelope as a single Atom core (though doing so would be of dubious use). My point is, though, that the comparison isn't all that fair, since all current ARM processors are fighting with both hands behind their back.

Even so, ARM performance has grown by leaps and bounds in the last 5 years, coming from PII levels of performance with Arm 9 and 11, to being nearly on par with Intel's Atom with the A-9. Thus far they've made those advances without throwing power consumption to the wolves, but imagine if someone came along with the 'radical' idea of even a 10 or 20W power envelope on an ARM implementation. Imagine indeed -- this is exactly what nVidia promised to do today, aiming at the desktop and server markets.

The ARM ISA isn't what's holding ARM back -- its been the power/thermal requirements of their core markets (SOCs, Embedded). Given power and die-size to burn, there's no reason ARM won't make a processor just as beastly as AMD or ARM (experience in doing so notwithstanding).

ARM has a similar problem to Intel in that they utterly dominate all the current markets where they compete -- this is why ARM is eyeing intel's turf and vice versa.

Intel may have a massively larger market cap, but ARM has volume that Intel can only dream about -- to give you an idea of that, it took McDonalds 21 years to sell a billion hamburgers -- and 3 billion ARM cores were produced last year alone. When ARM itself (A-15) and others (nVidia) want to push ARM to the limits, they'll find the market waiting.

I'm not an EE, but I would bet that mechanics of making a built-for-speed chip use less power (i.e. Atom/Bobcat) is a lot easier to tackle than the other way around.


Which is easier, to take a rich man who drives a fast car and convince him to drive a run-of-the-mill sedan, or to put a poor man into that same sedan?

Define easier.

Reply Parent Score: 2

galvanash Member since:
2006-01-25

Which is easier, to take a rich man who drives a fast car and convince him to drive a run-of-the-mill sedan, or to put a poor man into that same sedan?

Define easier.


By "easier", I mean less technically challenging... Atom/Bobcat are fundamentally more similar to their ancestor designs than current designs. Intel and AMD are taking the route of removing complexities in order to achieve lower power use (Intel by going in order, AMD by sharing execution resources with a single frontend). The point is the complexities they are removing in many cases are what make their higher end parts perform so well.

ARM has never had those types of complexities in the first place - most of the special sauce on ARM (Thumb for example) is to optimize things for the embedded space - smaller binares, better performance with smaller caches, etc. No one has ever tried to make an ARM core where performance was the primary goal - all existing cores where designed for power envelopes an order of magnitude or more smaller than high end x86 parts.

I'm not at all saying you can't make a very fast ARM core - I'm just saying it isn't as simple as just ramping up the clock speed and doing some minor reworking - a 3Ghz ARM might be possible with current designs - but even at 3Ghz it would have a long way to go to reach the performance levels of a similarly clocked i5/Phenom core, let alone match them when they can legitimately run at 4Ghz or more themselves. Im just saying it will take a lot of work to make ARM competitive if you factor out power use, and I have no reason at all to believe that nVidia could accomplish such a feat.

Also, I want to stress I am talking about single threaded performance, i.e performance per core, not overall performance. ARM can scale up very well by just throwing more cores at the problem, but that is not the same thing.

Reply Parent Score: 2