Mobile Virtualization

According to Electronista, ARM’s next generation of chips for phones and tablets should start shipping in devices at the end of this year.

These chips are based on ARM’s big.LITTLE architecture. big.LITTLE chips aren’t just multi-core, they contain cores that are two different implementations of the same instruction set: a Cortex A7 and one or more Cortex A15s. The Cortex A7 has an identical instruction set to the A15, but is slower and more power efficient – ARM says it is the most power-efficient processor it has ever developed. The idea is that phones will get great battery life by mainly running on the slow, power-efficient Cortex A7, and great performance by using the A15 on the hopefully rare occasions when they need its muscle. Rare in this context is relative. Power management on modern phones involves powering up and powering down subsystems in microseconds, so a ‘rarely’ used core could still be activated several times in a single second.

The Cortex A15 and the Cortex A7 are innovative in another way, too: they are the first cores based on the ARMv7-A architecture. This is ARM’s first architecture with hardware support for virtualization.

Even without hardware support, virtualization on handsets has been around for a while; phone OEMs use it to make cheaper smartphones by running Android on the same CPU that runs the cellular baseband stack. ARM says:

Virtualization in the mobile and embedded space can enable hardware to run with less memory and fewer chips, reducing BOM costs and further increasing energy efficiency.

This application, running Android on the same core as the baseband, does not seem to have taken the market by storm. I presume because of performance. Even the advent of hardware support for virtualization may not rescue this application, since mobile chip manufacturers now scale performance by adding cores, and Moore’s law is rendering multicore chips cheap enough to put into mass-market smartphones.

So what about other applications? The ARM piece quoted above goes on to say:

Virtualization also helps to address safety and security challenges, and reduces software development and porting costs by man years.

In 2010 Red Bend Software, a company that specializes in manageability software for mobile phones, bought VirtualLogix, one of the three leading providers of virtualization software for phones (the other two are Trango, bought by VMWare in 2008 and OK Labs.)

In view of Red Bend’s market, it looks as if they acquired VirtualLogix primarily to enable enterprise IT departments to securely manage their employees’ phones. BYOD (Bring Your Own Device) is a nightmare for IT departments; historically they have kept chaos at bay by supporting only a limited number of devices and software setups. But in the era of BYOD employees demand to use a vast and ever-changing variety of devices. Virtualization enables Red Bend to add a standard corporate software load to any phone.

This way, a single phone has a split personality, and the hardware virtualization support keeps the two personalities securely insulated from each other. On the consumer side, the user downloads apps, browses websites and generally engages in risky behavior. But none of this impacts the enterprise side of the phone, which remains secure.

iPhone 4S not iPhone 5

Technically the iPhone 4S doesn’t really pull ahead of the competition: Android-based phones like the Samsung Galaxy S II.

The iPhone 4S even has some worse specifications than the iPhone 4. It is 3 grams heavier and its standby battery life is 30% less. The screen is no larger – it remains smaller than the standard set by the competition. On the other hand the user experience is improved in several ways: the phone is more responsive thanks to a faster processor; it takes better photographs; and Apple has taken yet another whack at the so-far intractable problem of usable voice control. A great benefit to Apple, though not so much to its users, is that the new Qualcomm baseband chip works for all carriers worldwide, so Apple no longer needs different innards for AT&T and Verizon (though Verizon was presumably disappointed that Apple didn’t add a chip for LTE support).

Since its revolutionary debut, the history of the iPhone has been one of evolutionary improvements, and the improvements of the iPhone 4S over the iPhone 4 are in proportion to the improvements in each of the previous generations. The 4S seems to be about consolidation, creating a phone that will work on more networks around the world, and that will remain reliably manufacturable in vast volumes. It’s a risk-averse, revenue-hungry version, as is appropriate for an incumbent leader.

The technical improvements in the iPhone 4S would have been underwhelming if it had been called the iPhone 5, but for a half-generation they are adequate. By mid-2012 several technologies will have ripened sufficiently to make a big jump.

First, Apple will have had time to move their CPU manufacturing to TSMC’s 28 nm process, yielding a major improvement in battery life from the 45 nm process of the current A5, which will be partially negated by the monstrous power of the rumored 4-core A6 design, though the Linley report cautions that it may not be all plain sailing.

Also by mid-2012 Qualcomm may have delivered a world-compatible single-chip baseband that includes LTE (aka ‘real 4G’).

But the 2012 iPhone faces a serious problem. It will continue to suffer a power, weight and thin-ness disadvantage relative to Samsung smartphones until Apple stops using LCD displays. Because they don’t require back-lighting, Super AMOLED display panels are thinner, lighter and consume less power than LCDs. Unfortunately for Apple, Samsung is the leading supplier of AMOLED displays, and Apple’s relationship with Samsung continues to deteriorate. Other LCD alternatives like Qualcomm’s Mirasol are unlikely to be mature enough to rely on by mid-2012. The mid-2012 iPhone will need a larger display, but it looks as though it will continue to be a thick, power hungry LCD.

Apple iPad has proprietary processor

Well, the Apple iPad is out. Time will tell whether its success will equal that of the iPhone, the Apple TV or the MacBook Air. I’m confident it will do better than the Newton. The announcement contained a few interesting points, the most significant of which is that it uses a new Apple proprietary processor, the A4. Some reviewers have described the iPad as very fast, and with good battery life; these are indications that the processor is power efficient. Because of its software similarities to the iPhone, the architecture is probably ARM-based, with special P.A. Semi sauce for power and speed. On the other hand, it could be a spin of the PWRficient CPU, which is PowerPC based. In that light, it is interesting to review Apple’s reasons for abandoning the Power PC in 2005. Maybe Apple’s massive increase in sales volume since then has made Intel’s economies of scale less overwhelming?

The price is right, as is an option to go without a 3G radio. The weight is double that of a Kindle, and half that of a MacBook Air.

I am disappointed that there is no user-pointing camera, because as I mentioned earlier, I think that videophone will be a major use for this class of device.

Update 3 February 2010: Linley Gwenapp wrote up some speculations in his newsletter.