[News] Princeton researchers launch 25-core open-source CPU

[StyM]

http://www.bit-tech.net/news/hardwar...-25-core-cpu/1

A team of researchers at Princeton University have released a 25-core processor design under a permissive licence, after taping out a physical implementation in the 32nm process node of an IBM fab.

The Princeton Piton Processor, as the device is known, is based on a 32nm silicon-on-insulator (SOI) manufacturing node, making it cheap and easy to manufacture. The chip itself boasts 25 cores running at 1GHz, based around a modified version of the OpenSPARC T1 architecture. The idea, its creators have explained, is to build highly-parallel systems from entirely open components - and with three on-chip networks and directory-based shared memory, it's claimed the Piton can scale to hundreds of thousands of cores in a single system.

'With Piton, we really sat down and rethought computer architecture in order to build a chip specifically for data centres and the cloud,' claimed David Wentzlaff, a Princeton assistant professor of electrical engineering and associated faculty in the Department of Computer Science, of his team's creation. 'The chip we've made is among the largest chips ever built in academia and it shows how servers could run far more efficiently and cheaply.'

The 6mm? core of a Piton processor packs in 460 million transistors, while the high core counts and scalable architecture make it well-suited to data centre use. Coupled with techniques like execution drafting, where similar instructions can be spotted in a queue and executed in a row, the performance is impressive: execution drafting itself is claimed to boost performance by 20 percent, a memory-traffic shaper system adds an additional 18 percent, while other cache memory management techniques offer a 29 percent increase, all in comparison to traditional processors used in a data centre.

'We're very pleased with all that we've achieved with Piton in an academic setting, where there are far fewer resources than at large, commercial chipmakers,' crowed Wentzlaff, during a presentation at the Hot Chips conference this week. 'We're also happy to give out our design to the world as open source, which has long been commonplace for software, but is almost never done for hardware.'

[AliG]

Kinda cool, but unnecessary. It's simply cheaper to scale with GPUs than CPUs for workhorse applications.