Intel Reveals 48-Core Chips For Cloud Data Centres

Intel is showing off an experimental 48-core processor that officials say could hold 10 to 20 times the processing power now found in a Core-branded chip.

Intel officials demonstrated the processor – which they are calling a “single-chip cloud computer” – at an event in Palo Alto, California on 2 December.

The experimental processor comes from Intel Labs’ Tera-scale Computer Research Program, which is looking at ways of adding tens or hundreds of cores onto a single piece of silicon.

Such technology offers the promise of a variety of capabilities, from computers that can read the brainwaves of its users to highly efficient, extremely powerful data centres, according to Justin Rattner, head of Intel Labs and Intel’s CTO.

“With a chip like this, you could imagine a cloud data centre of the future which will be an order of magnitude more energy efficient than what exists today, saving significant resources on space and power costs,” Rattner said in a statement.

Intel and other chip makers, such as rival Advanced Micro Devices, have pursued multicore processor development as a way of growing the performance of their chips without having to crank up the frequency. Both Intel and AMD currently have six-core processors, and both are moving to chips with eight and more cores in 2010.

The new processor comes two years after Intel demonstrated an 80-core processor.

Other vendors also are pushing multicore processing. In October a five-year-old company, Tilera, which had already produced chips with 36 and 64 processing cores, announced a chip with 100 cores.

In addition, Nvidia, which makes graphics processors, unveiled plans for a 512-core GPU that will be able to run general-purpose computing workloads. Nvidia and AMD have been driving GPU technology – which can hold many more cores than traditional CPUs – into the mainstream computing space.

Intel’s 48-core chip – which is about the size of a postage stamp – is based on standard Intel Architecture, so that it can run today’s widely used Linux and Microsoft applications. Microsoft already has the Intel prototype in hand and has been working with it, according to Dan Reed, Microsoft’s corporate vice president of extreme computing.

“Our early research with the single chip cloud computer prototype has already identified many opportunities in intelligent resource management, system software design, programming models and tools, and future application scenarios,” Reed said in a statement.

In addition, researchers from Intel, Hewlett-Packard and Yahoo’s Open Cirrus collaboration have started porting cloud applications to the chip using the Hadoop Java-based framework.

Intel expects to have more than 100 of the chips to give to dozens of IT companies and research facilities for research into new applications and programming models.

At the same time, Intel officials said they will incorporate some features of the chip in a new line of Core-branded processors in early 2010 and in six- and eight-core chips planned for later in the year.

Included in the 48-core chip is a high-speed on-chip network and new power management techniques that enable all the cores to operate well at as little as 25 watts, or at 125 watts when running at maximum performance, the officials said.

The high-speed network will greatly reduce the amount of communication time between the cores, with data only have to move millimeters on the single chip rather than tens of meters between computer systems. That will increase performance and efficiency within data centres.

Intel officials said they use the single-chip cloud computer description because the processor’s design in many ways mirrors the parallel programming techniques used in software for data centre clouds.

---

To download APC’s white paper on the meaning and implications of the terms Power Factor, Crest Factor and Surge Factor, as used in specifying UPS, click here

---