I called this one in a piece I wrote about Mellanox...I think that as Nvidia gears up to build its own HPC platform with Denver/Tesla, it will work with Mellanox on the InfiniBand side of things to team up against Intel and its QLogic assets.
LEIPZIG, Germany, Jun 17, 2013 (BUSINESS WIRE) -- Mellanox Technologies, Ltd., a leading supplier of high-performance, end-to-end interconnect solutions for data center servers and storage systems, today announced it will showcase the world's first technology demonstration of ConnectX-3 FDR 56Gb/s InfiniBand solutions running on the SECO ARM development platform featuring NVIDIA Tegra processors. The SECO development platform is powered by an NVIDIA Tegra quad-core ARM processor. The RISC-based ARM architecture is advancing far beyond today's billions of client devices, and is rapidly moving towards the mainstream technical computing and high performance computing (HPC) market segments.
Together with NVIDIA, Mellanox will demo the ARM architecture with InfiniBand at the Mellanox exhibit booth #326 during the International Supercomputing Conference (ISC) in Leipzig, Germany from June 17-20.
"This particular technology demonstration represents a significant development milestone for adoption of Mellanox's InfiniBand solutions in new CPU platforms such as NVIDIA Tegra-based ARM platforms," said Gilad Shainer, Vice President of Marketing at Mellanox Technologies. "As future generations of 64-bit ARM solutions come on-line, applications will continue to demand ultra-low latency communications and scalability that only Mellanox InfiniBand can provide."
"Mellanox and NVIDIA are working together to bring all the benefits of a modern HPC network to ARM-based platforms," said Ian Buck, General Manager of GPU Computing Software at NVIDIA. "This technology demo, coupled with support for ARM platforms in the latest release of the CUDA parallel programming toolkit, provides the foundation for developers to build out the ARM HPC application ecosystem."
Breakthrough will speed system-to-system data center links
By Rik Myslewski in San Francisco, 11th April 2013
Intel has demoed what it says it "believes" is the world's only silicon photonics module that uses a hybrid silicon laser – a breakthrough that should allow such advances as vastly improved system-to-system interconnects in the data center.
The demo of the 100Gb/sec module was presented via a video during Intel CTO Justin Rattner's keynote presentation at the Beijing Intel Developer Forum on Thursday.
The chip incorporates modulators, detectors, waveguides, and circuitry – and that aforementioned laser – all in garden-variety silicon rather than pricey hand-built, gallium arsenide–based photonics modules such as are currently available.
Since integrated silicon photonics modules could be fabbed using existing equipment and tested using a hybrid of conventional and Intel-developed techniques, such speedy modules could be created much more cheaply, and therefore eventually work their way into far more devices and systems
[From another interview, The photonics technology was demo’d at IDF Beijing recently, and Rattner says Intel has engineering samples, was building up yields, and “We’ll be in the market within the year.”]