Intel on the Windows/ARM port: it could be a good thing for us - Ars Technica

On Friday, while I was at the Intel booth checking out some QuickSync and Intel Insider demos, Intel Executive Vice President Dadi Perlmutter dropped in for a moment. I got the chance to ask him some direct questions about the recent ARM news—the Windows/ARM port and NVIDIA's Project Denver announcement—and his answers are definitely worth sharing. He claimed that the Windows/ARM port could actually be a good news for Intel, and I think he has a point.

One of the ongoing themes in the last two years of our coverage of Microsoft is that the company's mobile strategy is a mess. The smartphone situation has just begun to turn around with Windows Phone 7 but for tablets and even netbooks, Windows is too big, too bloated, and has a monstrously unsuitable UI. So everyone's in agreement that Microsoft needs to do a lot more on the tablet front than just insist that some combination of desktop Windows plus a stylus is good enough. But part of Microsoft's problem is that there's currently no x86-based tablet hardware that can compete in battery life with ARM tablets.

Perlmutter basically said (and I paraphrase from memory here) that Microsoft has a long ways to go on the tablet software front, and that an ARM port will help them get there. And, by the time they get their tablet software sorted—a fresh, tablet-centric UI and much-needed power optimizations—Intel will be ready with an Atom-based chip that can compete directly with ARM in the milliwatt power draw range.

Right now, Atom is significantly more power-hungry (and more high-performance) than the ARM Cortex A8-caliber hardware that's showing up in Android tablets. But Intel will continue to close that gap with each new process generation. So, by the time Moore's Law delivers an Atom SoC with the same power profile as A8- and A9-based SoCs, Microsoft will either be ready with a tablet-worthy OS, or it will be further along than it would've been if it had been waiting for Intel to catch up.

So is this really how it's destined to play out? Will the ARM port of Windows pave the way for later development of an Intel-based Windows tablet, with ARM just keeping Intel's seat warm until smaller Atoms arrive? I have no idea, but the possibility is worth considering.

NVIDIA's speech vs. actions gap

As for competing with ARM across mobiles and now desktops, Perlmutter was clear that this wasn't the first time in his 30 years with Intel that the chipmaker had faced competition. He kept reiterating that competition from ARM, NVIDIA, or anyone else was nothing new, and that the company takes all competition seriously. Indeed, Perlmutter went out of his way to emphasize that Intel isn't claiming that dominance over ARM is a foregone conclusion, and in these remarks I could hear echos of former Intel chairman and CEO Andy Grove's famous dictum: "only the paranoid survive."

On the topic of NVIDIA's Project Denver CPU, Perlmutter pointed out that the announcement marked a possible disconnect between the GPU maker's past rhetoric about how much CPU performance does or doesn't matter. Specifically, he reminded me that NVIDIA has been claiming that, in the era of visual computing, only the GPU really matters anymore—CPU performance isn't a big deal.

"If they really believe that, and are designing a low-power, low-performance CPU core, then I'm very happy," said Perlmutter. But if NVIDIA is investing the money and effort that it takes to design a high-performance CPU core, then it suggests that the company's talk about how little the CPU matters nowadays is just hot air.

If NVIDIA really is trying to design a high-performance ARM core, Perlmutter again stressed that such competition would be nothing new for Intel. In fact, he insisted that NVIDIA probably could design a worthy high-performance CPU core. But the company will then have to successfully combine the results with a high-performance GPU on an SoC, and then integrate that into a larger system with all of the supporting logic and related engineering required.

Truth be told—and this is my take, not Perlmutter's—NVIDIA can probably do the system design, too. Their high-end GPUs are essentially compact systems-on-a-daughtercard, with separate power and cooling systems and support logic. So the company has plenty of experience in designing very compact, high-performance systems. But it's point taken about NVIDIA's anti-CPU rhetoric—if the CPU is irrelevant, why make one?

The answer, of course, is that both the CPU and GPU matter, as do the connections between them. We'll see how well NVIDIA is able to combine the two in the coming years, as both Project Denver and the Windows/ARM port move closer to market.