REX Computing Unveils New Energy-efficient High Performance Computing System
Server blade designed by MIT Systems Architect and Thiel Fellow follows Open Compute standards and utilizes new many-core coprocessor technology
San Jose, CA, January 28, 2014 - REX Computing unveiled today its first server blade prototype sporting a combination of ARM and Adapteva Epiphany processors at the Open Compute Summit in San Jose. The system showcases the capabilities of low-power ARM based processors in addition to new coprocessors to outperform existing technologies while using only a fraction of the electricity. The prototype will be on display at the OCP Summit (Booth #C10, Grand Ballroom, San Jose Convention Center).
Started in 2013, REX Computing is a growing hardware and software company developing the next generation of computing systems designed to power everything from web servers to supercomputers. Founders Thomas Sohmers and Kurt Keville started REX Computing with the vision of solving major problems related to power efficiency, density, and scalability in data centers. REX is approaching these problems by leveraging low power, highly scalable, and easy-to-program coprocessors alongside general purpose processors to yield over a tenfold increase in energy efficiency for the increasing workloads in the next generation of data centers.
“For decades the enterprise hardware industry has been dominated by large, established companies with little interest in fixing systemic problems throughout the business. With data centers and cluster computing these problems are particularly grim: everything from the basic architectures being used, how software is written, and even what racks are used, has become a drag on innovation throughout the industry,” said Thomas Sohmers, CEO and co-founder of REX Computing and 2013 Thiel Fellow. “Now, thanks to developments like the Open Compute Project, a nimble startup like REX can innovate in key areas without reinventing the wheel. We can keep our costs low without having to worry about designing a new form factor, confident that we have access to the best data center components and designs already known by our customers.”
“By applying the principles of open source software – transparency, flexibility, and community – to hardware, the Open Compute Project lowers the barrier to entry and provides new routes to market for innovative technologies to reach customers large and small alike,” said Frank Frankovsky, chairman and president of the Open Compute Project. “We’re thrilled to see companies like REX Computing building on OCP’s designs to capitalize on new opportunities in high performance computing.”
REX Computing’s prototype, on display at the OCP Summit, is being used to develop the software backend for future systems, in addition to showing the highly scalable nature of the coprocessor architecture being used. Future product plans include the use of the 64-core Epiphany IV coprocessor, allowing up to 4096 Epiphany cores per server blade. Multiple blades can then be networked together in a rack, and then throughout a datacenter.
The system utilizes 16 individual compute nodes, each containing a dual-core Xilinx Zynq ARM processor with an integrated FPGA (Field Programmable Gate Array), and an Adapteva Epiphany III coprocessor. Each compute node consumes less than 5 watts of electricity, with the Epiphany coprocessor alone consuming a mere 2 watts while delivering over 30 GFLOPs (Billion floating point operations per second). The full blade uses less than 150 watts of electricity, contains 32 ARM Cortex-A9 cores, 256 Epiphany cores, and fits in the 2OU tall and 1/3rd conventional server blade width “Torpedo” server chassis of the Open Compute Project’s defined standards.
REX co-founder and CTO Kurt Keville has been developing cluster computer systems at the Massachusetts Institute of Technology for over 10 years, and began working with REX co founder and CEO Thomas Sohmers in 2010. After being selected for Peter Thiel’s “20 Under 20” Thiel Fellowship class of 2013, Sohmers started REX with his colleague to bring the advancements of their research to the computing industry.
OCP Summit attendees who are participating in the third Open Compute Project Hackathon will have access to the (16 core) Epiphany III powered “Parallella board” development kits, in addition to the ability to run their project on the whole 256 Epiphany core server blade on display for the summit. REX will be providing development assistance and will be giving out Parallella boards to the best hacks to utilize the Epiphany coprocessor.
ABOUT:
Founded in 2013 by MIT researchers, REX Computing is developing new cluster computing systems for servers, mainframes, and supercomputers, while focusing on energy efficiency, cost and scalability by using new processor technologies. REX has been on the research forefront of ARM and other lower power cluster computing systems for a decade, and is now taking that research and development and bringing it to data centers worldwide.
|
Related News
- Semiwise, sureCore, and Cadence Showcase Breakthrough in Cryogenic CMOS Circuit Development for Quantum Computing and Energy-Efficient Data Centers
- Ceremorphic Exits Stealth Mode; Unveils Technology Plans to Deliver a New Architecture Specifically Designed for Reliable Performance Computing
- CAES Receives Contract from Vinnova to Advance High Performance RISC-V Space Computing
- Graphcore and SiPearl: strategic partnership to combine artificial intelligence and high performance computing
- PLDA Joins ETP4HPC to share its expertise in High Speed Interconnect Solution IP with the European High Performance Computing Ecosystem
Breaking News
- Ubitium Debuts First Universal RISC-V Processor to Enable AI at No Additional Cost, as It Raises $3.7M
- TSMC drives A16, 3D process technology
- Frontgrade Gaisler Unveils GR716B, a New Standard in Space-Grade Microcontrollers
- Blueshift Memory launches BlueFive processor, accelerating computation by up to 50 times and saving up to 65% energy
- Eliyan Ports Industry's Highest Performing PHY to Samsung Foundry SF4X Process Node, Achieving up to 40 Gbps Bandwidth at Unprecedented Power Levels with UCIe-Compliant Chiplet Interconnect Technology
Most Popular
- Cadence Unveils Arm-Based System Chiplet
- CXL Fabless Startup Panmnesia Secures Over $60M in Series A Funding, Aiming to Lead the CXL Switch Silicon Chip and CXL IP
- Esperanto Technologies and NEC Cooperate on Initiative to Advance Next Generation RISC-V Chips and Software Solutions for HPC
- Eliyan Ports Industry's Highest Performing PHY to Samsung Foundry SF4X Process Node, Achieving up to 40 Gbps Bandwidth at Unprecedented Power Levels with UCIe-Compliant Chiplet Interconnect Technology
- Arteris Selected by GigaDevice for Development in Next-Generation Automotive SoC With Enhanced FuSa Standards
E-mail This Article | Printer-Friendly Page |