Moore’s Law—which states that the number of transistors on a chip will double approximately every two years—remains the driving impetus of the IT world and, increasingly, of the business world too. It certainly drives Intel, the industry leader, and its competitors and it has led to decades of data center price/performance density gains.
The ability to double the number of transistors on each chip boosts productivity and performance while cutting the cost per transistor and, in turn, driving down the cost the products that use them. Where once chips were about clock speed and numbers of transistors, today’s latest chips incorporate many more capabilities, from power reduction to graphics processing to cryptographic security and more. A tablet or smartphone like the one you’re probably carrying in your pocket will have faster processing, more storage, greater display resolution, and consume far less energy than the standard desktop PC of a decade or more ago for about the same price.
For data center managers Moore’s Law translates into greater density at lower cost. They can shrink the required floor space and cut energy consumption while increasing the processing performance of the data center. They can literally do more with less from a processing standpoint. (Unfortunately, for data center managers Moore’s Law does not apply to people.)
Intel appears intent to lead the industry to new processor heights. Chip bloggers report that the company revealed an ambitious roadmap during its annual investor meeting. Specifically Intel plans to shrink its chip fabrication process down to a mere 5 nanometers (nm) sometime after 2015, with a 10nm process set to release in 2015 and work having already begun on the 7nm and 5nm processes. Typical processors today are in the 32nm range and Intel’s current top processor, Ivy Bridge, is 22nm. The next Intel chip, planned for 2013, will drop down to 14nm. Intel roadmap is here.
There are other chip makers, such as AMD, but the only other chip maker attracting this kind of attention today is ARM, which had shipped over 15 billion ARM processors as of Jan. 2011. Now ARM is aiming its chip designs beyond the mobile device market. According to analyst firm IHS iSuppli, by 2015, ARM integrated circuits are estimated to be in 23% of all laptops. ARMv7, its latest processor, mandates a hardware floating point unit, making it a full function chip.
The next step for ARM, ARMv8, already announced in 2011 but not yet available, represents the first fundamental change to the ARM design. It adds a 64-bit architecture and 64-bit instruction set. ARMv8 allows 32-bit applications to be executed in a 64-bit OS and for a 32-bit OS to be under the control of a 64-bit hypervisor. Products using ARMv8, however, remain a way off.
Although chip trends are central to data center success, data centers don’t buy chips. They buy products from system vendors who package the chips together with other hardware, software, and services to deliver the devices they actually deploy, notes Mark Teter, CTO at Advanced Systems Group, a system integrator. Unless data center managers make a deeper dive than usual into a product’s specifications they won’t know how multiple threads are being handled or how core pipes are organized nor should they. All they need care about are the capabilities, price/performance, density/footprint, and energy consumption. Let the chip vendors worry about the underlying details.
Still, what is shaping up on the chip front is the collision of two major chip camps, one conventional (Intel) and one mobile (ARM). As ARM extends into laptops, Intel has declared plans to focus on mobile chips. The shrinking of mobile chips down to 22nm and even 14nm is big. Watch this battle, like two galaxies colliding, shape up over the next few years. Data centers can enjoy a spectacle that will only accelerate the benefits of Moore’s Law.