Post-PC: Why Intel Can No Longer Live in Denial

Editor's Note: This is an expanded version of Jason Perlow's Pro-Post PC arguments in our Great Debate Series.

Indeed, we are living in the beginning of what Steve Jobs has coined as the "Post PC" age. But what does "Post-PC" actually mean?

If we are to interpret from his comments that the iPad is a "Post-PC" device, then it could be inferred that the tablet and light computing devices are to replace the Personal Computer.

But I think "Post-PC" encompasses a broad set of technologies that will eventually lead us into the next generation of Personal Computing as a whole, not just Tablets and Smartphones.

A Post-PC device is any computing device that does not have to fully rely on localized processing to provide the user experience or does not rely on the Wintel architecture in any way shape or form. This includes multiple form factors including Tablets, Smartphones, Chromebooks as well as Thin Clients, WebTops and Smart Terminals.

Even devices such as the Apple TV and the Roku and the XBOX 360/Kinect that are designed for pure content consumption are "Post-PC" devices, but they still encompass the broader definition of Personal Computing.

Why Intel x86 is a "legacy system" and reaching the end of its useful lifespan

When I think of the "PC" I am still thinking very much in terms of the x86 Intel platform -- the one which has been in use for 30 years this month.

While processor clock speeds, MIPS, storage density and improvements in graphics and I/O have improved by over a factor of several thousand, and transistor density on the processor die itself has multiplied over 1000 times, ultimately the Intel systems architecture that we run today is not much different than we ran on the original 5150 PC.

Indeed, several generations of new instructions to support many new features have been added, and in 30 years, the instruction word length and memory addressing has gone from 16-bit, to 32 bits to 64-bit, but x86 carries the very same legacy baggage in terms of core instruction set that it started with in 1978 with the original Intel 8086-8088.

In short:

• Increasing reliance on Cloud Computing, Virtualization and server-based applications (Web/VDI) will further reduce the need for heavily localized processing or reduce it to a very niche market of users that require powerful content-creation workstations.
• The Wintel architecture is 30 years old and has reached a level of diminishing marginal returns in its ability to continue to scale in both processor performance and also the ability to compete as a best of class "Green" microprocessor architecture in terms of heat generation and power efficiency when compared to other architectures such as ARM and PowerPC.
• The movement toward SoC's (Systems on a Chip) and very large scale integration and component consolidation will move the industry towards ARM-based and similar low-power systems architectures that will manifest themselves in a wide variety of device and system form factors.

I discussed this at length with Scott Raymond in our "Blade Runner" Series when we defined what the Personal Computing platform for 2019 could possibly look like.

Intel's challenges

Intel needs to have its "Come to Jesus" moment. Right now, they are continuing to re-iterate the benefits and proven longevity of x86, and with their recently-announced Tri-Gate manufacturing techniques, that they can allow x86 to live on for another decade.

Unfortunately, as demand for PCs wane and tablets and VDI increases, the x86 is going to find itself relegated more and more in the data center as a Cloud platform, where it will have to compete with IBM's POWER platform, System z Mainframes as well as competitive offerings from Oracle/Fujitsu.

In Public and Private Clouds, customers will expect to buy Software as a Service or self-provisioned offerings on shared private infrastructure. In many cases, this will sit on strategically outsourced infrastructure hosted by large IT services firms such as IBM and HP's EDS and Cloud-specific firms like Amazon AWS/EC2/S3 and Microsoft's Azure platform.

Intel's traditional customers will not necessarily even own their own infrastructure anymore, so it will limit the overall size of their server market.

Consolidated infrastructure means higher density and less physical servers -- and ultimately, less x86 silicon footprint.

This will occur at the same time when end-users are discovering they really don't need full-blown PCs in order to be productive or to consume their favorite content.

The next decade is going to be very challenging for Intel if they cannot come up to a strong competitor to ARM, POWER and the Mainframe.

The latter two competing platforms are very important because the Itanium, which was developed at Intel and to be HP's "Big Iron" alternative to POWER and Sun/Oracle SPARC is being largely abandoned by the industry.

Itanium is now only sold in Integrity UNIX systems sold by Hewlett-Packard and a few other remaining niche boutique vendors that still support IA-64 in the few HPC environments that haven't already migrated to commodity 64-bit x86 Linux blades.

Microsoft no longer offers an Itanium version of Windows Server and Oracle has abandoned all software development on the platform, and the various Linux vendors and distributions have stepped down their support for it dramatically.

So this leaves x86 as Intel's only Cloud deployment platform.

While x86 can continue to be a useful technology in the Cloud long after it loses relevance in the Personal Computing world, at economies of scale it cannot ultimately compete with Big Iron systems hosting extremely large multi-tenant Clouds.

As if Big Iron isn't already a concern, we have already seen some indication that even ARM may penetrate into the datacenter, which is x86's stronghold.

Ultimately, If Intel cannot come up with an answer to any of these platforms, it has to at least become a leading manufacturer of ARM-based chips, something they haven't had to do since selling their XScale ARM assets to Marvell in 2006.

What the future means for Personal Computing in a Post-PC world

While Intel has demonstrated fabrication techniques such as Tri-Gate that could conceivably allow x86 to live on for at least another decade, at least in the high-end server space, it would be much more practical to apply those manufacturing techniques to a much more power-efficient architectures.

These architectures include ARM or even the PowerPC, which is used in the XBOX 360, The Nintendo Wii, the Sony Playstation and IBM's System z and POWER 7 computers, such as the ones that were the brains behind the "Watson" computer cluster and Deep Analytics software that beat the humans on the Jeopardy! quiz show.

The ARM architecture, which runs at the core of the iPad and Android tablets and every major smartphone platform, is also a target for Microsoft's upcoming Windows 8 operating system. So even from the perspective of the Redmond software giant, the x86 is entering its final years as a viable Personal Computing platform.

ARM is crucial to this debate because it is central to the systems architecture of many, but not all Post-PC devices. Certainly, it is possible to build a Post-PC device using Intel's Atom chipsets, but they are at a clear disadvantage when it comes to power consumption.

Performance with ARM chips on Personal Computing devices is an issue which often comes up when we talk about how it stacks up against x86. I would like to address that.

At this time, Intel has a clear clock speed, core count and instruction width and memory bus advantage when compared to ARM, but that gap will close quickly in the next five years.

However, performance matters only if typical end-users actually perceive a significant degradation in performance. There are plenty of end-users that have powerful PCs that bought completely overkill when they could have spent 1/2 or 1/3rd the money to address their actual needs.

With a smart device such as a tablet, a smartphone or a Chromebook, not only do they provide more than adequate experience in terms of regular end-user acceptance but they are also less complicated to maintain and deploy in the enterprise.

Additionally, in a number of cases, thin/smart devices might actually provide superior performance and resiliency to their PC counterparts when backed up by beefy server infrastructure, such as in large VDI implementations.

Soon, ARM processors will have 64-bit memory addressing and then 64-bit instructions. In the next two to three years, we will also see quad, six and eight-core ARM SoCs, which will allow them to completely replace the Intel chips used in today's PCs, although arguably a lot of software will need to be re-written to support higher levels of parallelization with higher core counts.

But Microsoft is introducing a completely new software architecture with Windows 8 and is ditching much of its legacy baggage anyway. If you've seen the Metro UI and the new APIs, it bears very little resemblance to the Windows of old.

The company has already gone on record to say that ARM-based Windows 8 tablets will not have Win32 compatibility whatsoever -- they will use entirely new apps that run on the Metro UI using the new WinRT, or Windows Runtime.

I have also argued in the past that Apple is almost certainly leading towards a unified systems architecture for Mac and iOS that is based on multi-core ARM chips, once they become powerful enough to handle the same types of workloads that typical users encounter today.

The PC is dying, but Personal Computing is not

The PC architecture -- which began its life at IBM and spawned an entire industry an is commonly simply referred to as "Wintel" is almost certainly in its final decade in the consumer space. But does this mean Personal Computing is dying with it? Certainly not.

By porting Windows 8 to the ARM architecture, Microsoft has already prepared itself for the "Post-PC" transition. It will re-write popular software products such as Office using the new Metro UI and WinRT APIs for ARM and other chip architectures such as PowerPC if necessary, and permit Windows 8 to scale from the lowest power ARM tablets to the most powerful parallel-processing Cloud-optimized super-clusters.

This is certainly not a new activity for Microsoft, as it had to port Windows NT to multiple systems architectures during the 1990s. At one point, Windows NT also ran on the MIPS R4000, the IBM PowerPC, The Intel i860, the DEC Alpha and also Intel's Itanium.

Today, only 32-bit and 64-bit x86 remains, but the NT kernel that we still use in both Client and Server Windows today has always been designed to be portable. What is old is new again.

Windows will continue on -- using Cloud and server-based technologies such as Microsoft's RemoteFX on VDI using thin clients and with local applications running natively on Windows 8 ARM-based smart devices.

And certainly Google's Android will also continue to evolve and will be able to run more demanding workloads.

Mobility is also an important part of the Post-PC scenario. It is telling that the largest growth sector for the PC market in the last 10 years has been in laptops and notebooks. But for notebooks to become more efficient in terms of battery life the fundamental systems architecture of its underlying OS and microprocessor platform must change.

Smartphone and Tablet usage has exploded because they provide superior battery life to a laptop computer and provide end-users constant access to their critical applications, regardless if they are in the office or Starbucks connected to Wi-Fi or on the train or in the ar using a 4G mobile data network.

Will all of those next-generation devices be tablets or smartphones? No. I find it very hard to accept that office workers will be able to transition completely away from mice, keyboards and large monitors to touchscreen tablets and still remain productive.

Tablets will become an important part of our overall computing experience, but they will not be our only window into Personal Computing, nor will they be the preferred environment where content creation actually occurs.

We will begin to see more and more reliance on smartphones and tablets versus traditional PCs in the next 18-24 months. Beyond that, we can expect Apple to move to a consolidated ARM-based systems architecture for its phones, tablets, set-tops and its Mac-based systems by 2014, with a total consolidation of iOS and Mac OS X product line by 2016.

I expect that in late 2012 and early 2013, we will see quad-core Windows 8, Android and iOS-based ARM tablets. And we will also see LCD HD monitors built by companies like Samsung with built-in RemoteFX RDSH clients which will allow the "Screen" to replace a local PC with VDI sessions.

I also expect to see ARM-based Windows 8 and Android-based laptops in the 2015 and 2016 timeframe as well, which will be manufactured by the usual suspects -- Dell, Lenovo, whoever ends up owning HP's PC business, as well as the Chinese and Taiwanese smartphone and tablet contract ODMs wishing to branch out into the ARM-based "Ultra-laptop" business.

So have we entered a Post-PC era? Yes, without question. The x86 has absolutely been issued its walking papers.

Is Wintel going to become extinct? From the perspective of the majority of end-users, yes. But will the traditional Personal Computing experience of having a mouse, a keyboard and a point and click GUI go away in favor of touch? No.

However, the platform delivering it to us in the next decade will almost certainly have no resemblance whatsoever to the PC we are using right now. That much I can guarantee.

Is Intel in denial when it comes to the x86's lifespan as the leading Personal Computing systems architecture? Talk Back and Let Me Know.

Disclaimer: My Full-Time Employer is IBM. I write as a freelancer for ZDNet. The postings and opinions on this blog are my own and don’t necessarily represent IBM’s positions, strategies or opinions.