ARM Community: Hot Chips Takes a Walk Down Memory Lane - Part 1 - ARM Community

Recently I gave a keynote presentation at the Hot Chips conference at Stanford University entitled “ARM Processor Evolution: Bringing High Performance to Mobile Devices”. In my talk I covered three main topics. First, how computing has evolved over the last 30 years from its desk bound origins to the ultra mobile world we know today. Second I talked about ARM’s role in that evolution and some of our early experiences of mobile devices. Finally, I spoke about some of the challenges I see ahead of us. In this blog I will give a recap of my presentation. Here in part 1 I’ll cover the history, and in part 2 look to the future.

Back to the 80’s
There is no doubt that personal computing has come a long way in the last 30 years. In my presentation I started by looking at the leading mobile devices from the early 80’s, namely the Osborne 1 and the Motorola DynaTAC. Launched in 1981, the Osborne 1 is widely recognized as the first portable computer. Weighing in at a solid 24.5lbs, it would be a brave man who carried that on a round-the-world business trip, especially as it is slightly too large to qualify as carry-on on many airlines today! However, luggable it was, and its 4MHz Z80 processor allowed it to run the CP/M operating system and a bunch of ‘office’ applications, such as word processing and spreadsheets, on its tiny 5” screen. By today’s standards it’s hard to imagine anyone using the Osborne 1 in anger, but it started the quest for smaller and smaller portable computers and was a commercially successful product.

A couple of years later, in 1983, along came the Motorola DynaTAC, the world’s first portable cellular phone. Like the Osborne it was large and heavy; a far cry from the sleek device you carry in your pocket today. However, once charged you could have 30 minutes of mobile talk time, and that was revolutionary. While you couldn’t send text messages, pair your BlueTooth headset or download apps (since none of those things had been invented) it did exactly what it was supposed to do – allow people to make phone calls without being plugged into the wall.

Unfortunately the DynaTAC also came with a revolutionary price tag - $3,995, rather more than you would pay for a phone today! As a result, whilst its iconic design is instantly recognizable by most people over the age of 40, the DynaTAC was a business tool and a status symbol, affordable only by those with expense accounts.

1990: ARM is born
Fortunately as time has moved on mobile devices have become smaller and cheaper with considerably more functionality and battery life. For the past 20 years (nearly 21 now) I believe ARM has played a significant role in that evolution by designing technology that has enabled high performance computers to run off small batteries.

ARM was originally a joint venture between Apple, Acorn Computers and VLSI Technology Inc. When the company was formed, the short-term goal was to build a processor for the Apple Newton. The Newton was another revolutionary product - designed for mobile computing and with handwriting recognition. It was a great device; I personally used one for many years to organize my life. But the product was ahead of its time and was ultimately canceled after a few short years. Meanwhile at ARM, we had fortunately found other applications for our low power, 32-bit embeddable processors.

Specifically we found an opportunity to have our processor designed into digital cell phones but there was one small problem, code density. Code density is the measure of how much memory your program occupies. At the time mobile phone OEMs wanted to move to 32-bit processors to add performance to support the new features enabled by digital cellular. However when they compiled their code they found the program size was large and hence the memory costs for the end device were prohibitive. In tackling this problem we worked out how to compress the 32-bit instructions into a much denser 16-bit format we called Thumb. With the ARM Thumb™ instruction set we were thus able to enable high performance without compromising code size.

The first processor incorporating this technology was the ARM7TDMI™, a small core that became the de facto standard for GSM cellular modems. The processor was small enough to be integrated with a DSP and other logic to form a digital baseband chip capable of running the protocol stack, the phone’s UI and ‘apps’ (by which I mean Snake). Originally 4.2mm2 in 0.5um technology, ARM7TDMI has subsequently been ported all the way to 28nm and alongside modems been used in a myriad of products resulting in nearly 15 billion cumulative shipments.



Back to Today
20 years on and cell phones, laptops and PDAs have merged into one all-powerful mobile computer. ARM processors, now much more sophisticated than the ARM7TDMI, power many of this new generation of device. As we have evolved our products and delivered more and more performance, the themes of small size, low power and efficient system design have remained central to our activities.

One of the leading devices of 2011 is the Samsung Galaxy SII. This features a GHz+ dual ARM Cortex-A9 processor and our Mali graphics processor all built using ARM Artisan Physical IP. These technologies combine to deliver a visually stunning communications/entertainment/personal/work device that would have seemed like science fiction 30 years ago. The path from big, expensive and limited to small, low cost and ubiquitous has required some significant technology advances in our designs and in the industry in general.

Underpinning this evolution has been the progress of semiconductor process. Roughly every 2 years the semiconductor industry has delivered a new process that enables twice the transistors per unit area of silicon than the previous generation and at the same time delivered more performance and lower power. This feat of science and engineering, known as Moore’s Law, is nothing short of miraculous and has enabled the creation of all the electronic devices we take for granted today. It has enabled low cost electronics to transform our lives and soon, with the ‘Internet of Things’, be embedded into almost everything around us.

Portable computing has been a major beneficiary of these advances in semiconductor processing. Today on a 28nm process we can integrate a multi-GHz dual-core Cortex-A9 along with a Mali graphics core and have plenty of room left over for everything else the SOC designer needs to complete his product. 20 years ago that would have required many separate chips resulting in a larger circuit board, more cost and more power consumption. That technological progress means we can have devices like the Galaxy SII and in real terms only pay about 1/20^th what the DynaTAC cost.

As a consumer, and as an engineer, I couldn’t be happier. Today’s Smartphones are amazing devices that are transforming how people interact with each other, with the Internet and with the world at large. But really, these devices are still in their infancy - just imagine what another 30 years of technology development is going to deliver! The market is big and growing but servicing it is not going to be easy. I’m not saying that the technology advances of the last 30 years have been easy, but I think there are some real challenges ahead. In Part 2 I’ll look at this opportunity and discuss some of the challenges I see ahead of us.

Thanks for reading.

Simon Segars, EVP and General Manager - Physical IP Division, ARM, Simon joined the board in January 2005 and was appointed EVP and General Manager of the Physical IP Division in September 2007. He has previously been EVP, Engineering, EVP, Worldwide Sales and latterly EVP, Business Development. He joined ARM in early 1991 and has worked on many of the ARM CPU products since then. He led the development of the ARM7 and ARM9 Thumb® families and holds a number of patents in the field of embedded CPU architectures. He is on the board of two not for profit organizations, the SOI Industry Consortium and the EDA Consortium.