Wednesday, June 18, 2014

Intel to Integrate Xeon + FPGA On Chip to Accelerate Data Center Processing

Intel announced plans to integrate its Xeon processor with a coherent FPGA in a single package, socket compatible to a standard Xeon E5 processor. The goal to accelerate software defined infrastructure and support scale-out, distributed applications. Intel's Diane Bryant and Tom Krazit discussed the plans at the Gigaom Structure’14 conference in San Francisco.

In the past year, Intel has already delivered 15 custom silicon devices to meet specific needs of the end customers, including Ebay and Facebook.  The company expects to deliver more than twice the number of specialized products in 2014.

In a blog post, Bryant writes that the new Xeon+FPGA solution "provides yet another customized option, one more tool for customers to use to improve their critical data center metric of Performance/TCO”. She also highlights work underway with Facebook on the Open Compute Project and ETSI's NFV ISG to develop many proof-of-concept demonstrations.

In February, Intel introduced its new Xeon processor E7 v2 family featuring triple the memory capacity and double the compute performance of the previous generation processor family, allowing much faster and thorough data analysis.

The Intel Xeon processor E7 v2 family is built for up to 32-socket servers, with configurations supporting up to 15 processing cores and up to 1.5 terabytes (TB) of memory per socket. 

To reduce data bottlenecks, the Intel Xeon Processor E7 v2 family features Intel Integrated I/O, Intel Data Direct I/O and support for PCIe 3.0, achieving up to four times the I/O bandwidth over the previous generation and providing extra capacity for storage and networking connections.

Intel said Big Data analysis and Internet-of-Everything require vastly increased in-memory processing, which places and analyzes an entire data set – such as an organization's entire customer database – in the system memory rather than on traditional disk drives.