There has been much written over the past year about Intel and the arrival of the end of Moore’s law – at least as we knew it.
Earlier today, a friend sent me a link to an Ars Technica piece discussing Kaby Lake, and what a letdown it was in terms of desktop CPU performance momentum.
I’m going to let you in on a little secret. The desktop CPU is dead. Don’t tell your friends who are big desktop gamers… they’ll never forgive you for crushing their dreams. But it is true. Gaming and VR will surely continue to have a place in the PC realm. But these aren’t mainstream scenarios – they aren’t the things that the broadest section of normals seek to have their personal computing devices do for them at home. They want a (personal) device that they can use for Web browsing, email, video, and perhaps productivity and some casual gaming. This device is most likely not plugged in all that often, and lives… around the house, not tethered to a dusty desk.
When I started working with Windows 25 years ago, it was an ongoing battle of wits, where Windows was constantly pushing the boundaries of what software asked from hardware. It practically felt like it was Microsoft’s goal… or… responsibility? to keep pushing the CPU requirements. This helped drive a virtuous cycle where Windows demanded a new computer, which demanded a new version of Windows, etc. I’m particularly inclined to recall a Christmas with my ex-wife’s aunt and uncle when we were newlyweds in 1996, when their big gift to the whole family was a new Gateway PC.
The irony of the PC was that, for a long time it wasn’t really a PC (Personal Computer). It was a Family Computer. That’s what my family had when we owned a //e. That’s what my ex-wife’s cousins got that Christmas. PCs were so expensive, they were a family purchase every few years, for the family to share.
Most of my generation may remember a single family phone line, which splintered into call waiting, multiple lines, and finally personal cell phones and the (near) death of the communal home phone line. The arrival of cheap PCs, tablets, and most importantly, very full-featured smartphones that, for many people, have replaced the desk-dependent PC. With the 15th anniversary of the release of Windows XP just behind us, I recall the Fast User Switching feature it delivered, and how it was, in a way, a nod to the future, where the devices around us would be truly personal – in terms of ownership, how they’re chosen/tweaked/replaced, and (sigh) managed or not.
I can’t speak to explicitly why Intel chose to insert Kaby Lake into their release cycle. Much like I can’t speak to why Apple chose to release a MacBook Pro in 2016 that was based off of Skylake (the prior year’s chip). I can’t speak to either one because I wasn’t included in any of the design meetings where the decisions behind both were made.
But I can pull out my handy-dandy Jump to Conclusions mat, and suggest the reasons why both of these things happened.
The market is changing underneath Intel, and underneath the Mac. I firmly believe that Apple invests in the Mac proportionally to the revenue the Mac returns – and if Apple chose to assign additional millions of dollars/year in Mac engineering R&D, it would not result in sales growth that reflects the investment. The scenario appears somewhat similar for the iPad lineup. However, Apple invests significantly in the iPhone, where R&D in equates to sales out. The device we’ll likely see in 2017 for the 10th anniversary of the iPhone will most likely reflect this. The Mac lineup, on the other hand, will be a bit of a cobbler’s child… getting hand-me-down technology.
Apple is building the device that addresses their mass market. But wait, why am I talking about Apple so much, in a blog post about Intel?
Intel is facing the real complexities to the end of Moore’s Law as we know it. There’s a finite end to how far you can take Intel’s “tick-tock” cycle to shrink CPUs using today’s technology.
More importantly, chip performance or “horsepower” isn’t the main thing most consumers shop for anymore. This is akin to people today who still love massive V-8 engines when most manufacturers are working on squeezing performance out of turbocharged V-6 or even inline fours in order to achieve better energy efficiency (and more importantly for their federal compliance, fleet vehicle efficiency). Consumers want battery life, quiet, and minimal heat. These are things that do not equate to monster innovations in raw CPU performance.
The move towards electric automobiles is of course another great analogy to take forward here – with ARM taking the role of electrics, and Intel being conventional fuels. Intel is working diligently to do performance limbo, and take their x64 architecture as low as it can go, to deliver great battery life while delivering good performance and great video playback. (See the earlier Ars Technica reference and the mention on 4K video in Kaby Lake.) Here Intel is focused on building distinct innovation into CPU releases that directly benefit the broadest section of the market, not just people looking for raw performance gains YoY.
In the consumer “PC” market (accepting the squishy, ever-evolving definitions of “tablet” and “PC”), any chance for Intel’s broader market strength will come from trying to compete with rapidly increasing performance from ARM chips. (Witness the Snapdragon 835 chip, which Microsoft plans to run x86 legacy Windows [Win32] applications on. We’ll see in time how that works out.)
For people who care more about speed and raw performance from their processors, the future isn’t likely full of roses. It’s likely to look more and more disappointing, much like the future for “motor heads” is. They’ll have to shop at the highest end of the market, or – longshot – hope for another vendor to address it with (likely expensive, niche, products).