The mobile world's pointless rush to 64-bits

Credit: Image credit: flickr/Linda Tanner

My question to all of the proponents: what are you going to do with it?

So the rush to 64-bit mobile processors has turned into a stampede. Apple started it with the A7 processor. Qualcomm and Intel have followed. But this charge begs a question: why the heck do you need with a 64-bit processor on a mobile device?

Intel continues spinning its wheels in mud with its mobile efforts, despite prototypes and promises of design wins that never show up. Now at Mobile World Congress it has introduced two Atom designs for mobile, Merrifield and Moorefield, plus a new LTE-Advanced modem from its Infineon group.

Merrifield and Moorefield should both arrive this year, with "SoFIA 3G" and "SoFIA LTE" coming late in the year. That generation will be Intel's first Atom SoCs with integrated modems. If Intel is going to make progress in mobile, that will be the generation that does it, I think. One of the key advantages Qualcomm has is its Snapdragon and Krait SoCs have the modem integrated with the SoC. This means one less part and less power consumed.

Then there's Qualcomm, which announced the Snapdragon 610 and 615. The only difference between the two is the 610 is quad-core while 615 is octo-core. They are compatible with the ARMv8 64-bit instruction set. So Qualcomm has an eight-core, 64-bit processor for smartphones.

Intel and Qualcomm are run by very smart people, and yet I wonder if I'm the kid who sees the emperor has no clothes. I expect most of you know this, but I'll spell it out anyway. Jumping to 64-bits in no real way makes a processor any faster or more efficient or draw less power. There might be some instances where performance improves due to longer registers, but you won't see that on a smartphone.

There is only one real gain from jumping to 64-bit CPUs, and that's addressable memory. A 32-bit chip can't see more than 4GB of memory. In theory, a 64-bit chip can see 16 exabytes, but you will never assemble a server with that much memory for the foreseeable future.

A 64-bit processor is golden in a server. It ushered in the era of virtualization, cloud computing and Big Data. On the desktop, it's convenient to have 16GB. I can run every app on my PC, literally, and not slow down. Convenient but not necessary. I'm sure developers and CAD and other graphics designers are making the most of all that memory more than anyone else.

On a mobile phone or tablet, all that memory is pointless and even counterproductive. The average smartphone and tablet comes with 1GB of application memory, well under the ceiling of a 32-bit chip. ISVs have done a spectacular job of keeping their apps tiny. Every now and then I remember to go into my iPad and iPhone and clean out all the apps minimized in the background and it really adds up.

So let's say someone comes out with an 8GB smartphone. Suddenly, the power needed to initialize and keep active memory has gone up eight-fold. Memory has to be powered even when it's not in use. So the battery life will be significantly impacted, plus you will have more heat, because more memory being powered means more chips drawing electricity.

Then there's this business of four and eight cores. Virtually all smartphone apps are single core. When you add more cores, you have to dial back the clock speed. In a server, of course, this is not an issue. It's easy to make a four-core CPU running at 3.4GHz. Intel does it all the time. But it's impractical and almost useless in a smartphone.

I think the vast majority of us would be just fine with a dual core processor that shuts off the idle core and really cranks on single core performance.

This rush to 64-bit mobility is pointless. Maybe it will be useful in 5-10 years but the current trend in mobility is to make very small, compact and highly efficient apps, negating the need for more than 4GB of memory. And technically, a many-core, 8GB mobile device is very counterproductive, especially when it comes to battery life.

Am I the only person who sees this?

What’s wrong? The new clean desk test
Join the discussion
Be the first to comment on this article. Our Commenting Policies