ARM And the Hacktop Pro Project

05 Apr 2013

Since the invention of silicon-based computers, we have seen what has become known as “Moore’s Law”, being that over a two year span a comparison between the initial and final states of computer power show a twofold increase: either the price has fallen 50% or its compliment that the per-dollar performance available has doubled. This trend held steady for a long time but it has a limit. The lithography technology which we use to fabricate chips hits a wall at about a 5nm feature size and the rate at which Intel is reducing the size of their chips suggests that before the decade is out we will reach the limit. Furthermore, as chip size decreases chip designers have to worry more and more about the quantum mechanical behavior of their designs because as it turns out with some probability electrons can “tunnel” through the structures which chip designers use to control their flow.

This means that we will either see consumer chips designed for deployment in atmosphere near sea level begin to deploy some of the hardware error correcting and detection techniques which NASA and the military have been using for years to defend against cosmic rays & high energy radiation simply to accurately perform day to day computations. The alternative is that as an industry we shift away from building smaller and faster cores to building more cores and trying to reap benefits from parallel software and systems. We are not blind and Intel’s introduction of quad core desktop and laptop class processors are the premonitions of the parallel shift.

When we have to go to such lengths to achieve performance, it seems to me that the question should be asked: are we looking for performance in the wrong place and do we as an industry need to reconsider what we “need” vs. what hardware do consumers seem willing to give us to run on. In short, I think that one could build a “better” processor than the i7 chips which power my machines. When it comes to claiming superiority, I will be very careful. For 20 odd years, programmers have been writing code targeted to x86 processors. College sophomores learn to code to the x86 and write operating systems for the x86 because it is ubiquitus. AMD hasn’t built a competing architecture, and ARM is the only one which it seems at present stands a chance against the x86 due only to the low power profile of its chips compared to the historically high power profile of Intel product.

In the last year or so, the Rasbery Pi has broken onto the scene powerd by a sigle ARM core. We’ve seen webservers, music and file servers, even desktops run off of this tiny, arguably underpowerd computer and yet it can reasonably run Emacs, Firefox and an entire Linux deskop suite at 1080p resolution. The $35 pricepoint of the Pi leads me to question the wisdom of of purchasing another ~$1500 Intel “ultrabook” when my Samsung Series 9 dies. In fact I almost built what I was going to christen the “hackbook pro” by taking a Motorola Atrix keyboard, screen and touchpad combo and case-modding a Pi into it. The build was going to run me $250 pi included and for it I’d get hardware only physically less attractive than an equivalently priced Intel powered “Netbook”.

This suggests that a low power ARM core suitably wrapped in a well designed 15” or 13” laptop chassis with remotely attractive design would be competitive certainly in terms of price with the Intel “Ultrabook” product line. How so? The ARMtop or shall we call it the Hackbook Pro would presumably have lower CPU and motherboard power costs than the equivalent Intel hardware and consequently should be able to boast a battery life well in excess of the four to six hour state of the art of Intel machines. However I must stress the importance of the 15” form factor. As a college student I probably encounter about a thousand unique laptops a day, many of them macbooks. The unfortunate reality of the