9:07 PM
Tech Plus
The future of computing isn't x86. It's not Windows-based either, condemning both halves of the 'Wintel' alliance to death row. But then it's been a long time since anything from Microsoft looked like the future rather than the past.
As it happens, Microsoft was partly responsible for the announcement that inspired my smell-the-coffee moment. I speak, of course, of the Windows-for-ARM announcement at the Consumer Electronics Show in Las Vegas back in January, where Microsoft head honcho Steve Ballmer revealed that the next version of its core Windows OS would support ARM processors as well as x86 chips.
The other half of the equation was Nvidia's Project Denver, an attempt by the graphics specialist to create a high-performance ARM processor for desktops and servers. It's a catastrophic cliché, but when I put the two announcements together, I immediately knew my world of computing would never be the same again.
Don't get me wrong. I'm not suggesting that Windows for ARM itself is going to be a big deal. On the contrary, it's probably best characterised as a lumbering giant's desperate attempt to remain relevant in a fast-changing world. Likewise, I've no idea whether Project Denver will be the success Nvidia needs it to be.
Established trend
Instead, it hit home because everyone instinctively knows that any trend behind-the-curve Microsoft is picking up on must already be well established and widespread. And Microsoft has seen that the future of computing might be ARM instead of x86.
You might argue that this is hardly a profound observation. After all, ARM chips already dominate most mobile and embedded applications including the increasingly ubiquitous smartphone segment. But what the Microsoft and Nvidia announcements underlined was that ARM has a real shot at extending up into the traditionally performance-orientated desktop and server segments - x86 territory, in other words.
Likewise, Windows for ARM and Project Denver put Intel's efforts to move down into ARM's traditional mobile and embedded segments with the Atom processor into new context. Previously, Atom seemed like a sensible tactic to get Microsoft into an ultramobile market experiencing explosive growth.
What I hadn't truly considered was that Intel, and by extension AMD, was at risk in its core desktop and server market. The way I see it, the likely scenario goes like this.
As transistors become smaller and smaller, the issue of outright performance in all computing devices becomes less critical. Already, you could argue that a decent quad-core desktop PC processor provides all the general purpose performance most consumers are likely to need. The fastest ARM chips remain a long way off the performance of any quad-core x86 CPU.
But with the latest out-of-order Cortex A9 core and the emergence of multi-core ARM designs, it's only a matter of time before that gap is closed. That's not to say I expect processor technology to remain static - quite the contrary. But instead of ever more general purpose cores being added, future chips will sport a mix of special-purpose circuitry designed to accelerate specific tasks.
Think graphics, media encoding, encryption and maybe even features like facial recognition, all accelerated in hardware. What future processors won't be are wall-to-wall general purpose cores.
Strong ARM
Massively multi-core is not the answer. To my mind, this scenario plays to the strengths of ARM chipmakers. ARM itself doesn't make chips - it licenses the IP needed to make ARM-based processors, leaving Samsung, Qualcomm, Nvidia and up to 15 other outfits to bring in their own tech and that of third parties to create a product tailored for an application.
Intel and AMD's traditional approach has been a one-size-fits-all chip designed, manufactured and distributed by a single company. That demands both be experts in all areas of chip design, and compete with specialists in areas like graphics and multimedia. More recently, AMD has spun off its manufacturing arm, leaving it positioned halfway between ARM and Intel in terms of business model.
Similarly, Intel revealed a deal with Taiwanese chipmaker TSMC to build Atom processors containing third-party IP. But the implications are clear enough. Desktops, laptops and servers are currently dominated by processors from just two companies.
In five years or so, AMD and Intel could well look like the victims in a cheap horror flick, covered by and being consumed by a swarm of small nasties. Some kind of future, eh?
 |  |
| |
| |
| |
| |
| |
| |
| |
0 comments:
Post a Comment