OpenPower Group Puts Initial Hardware Products on Display

1 of 18

OpenPower Group Puts Initial Hardware Products on Display

By Jeffrey Burt

2 of 18

Drawing a Crowd

Journalists and analysts get a look at the more than 1 dozen pieces of hardware—from servers to adapter cards to silicon—created by IBM and other OpenPower members.

3 of 18

First Box Out of the Box

Tyan's TN71-BP012 server will be the first OpenPower system to hit the market when it becomes commercially available in the second quarter. The 2U (3.5-inch), single-socket system, based on the Power8 architecture, was developed in conjunction with Mellanox Technologies.

4 of 18

Tyan's First OpenPower Offering

In October 2014, Tyan launched the GN70-BP010, a 2U, single-socket customer reference platform based on the OpenPower architecture.

5 of 18

Cirrascale Accelerates Computing With GPUs

Cirrascale teamed with Nvidia and Tyan to create the RM4950, the first OpenPower development platform to leverage GPU accelerators—up to four Nvidia Tesla K40s—to increase performance while keeping a lid on power consumption. It will ship in the second quarter.

6 of 18

Bringing OpenPower to HPC

IBM is teaming with Wistron, Nvidia and Mellanox in developing 2U, dual-socket servers—code-named "Firestone"—that will leverage GPU accelerators and will be aimed at the high-performance computing sector. Firestone, which will be released later this year, is the first of three systems the vendors will use to build supercomputers for the Lawrence Livermore and Oak Ridge national laboratories.

7 of 18

Another IBM-Nvidia Collaboration

The companies developed the IBM S824L Power8/Nvidia GPU server, a 4U, two-socket system for high-end workloads from such businesses as banks, energy firms and pharmaceutical companies. The system was launched in October 2014.

8 of 18

IBM, Altera and FGPAs

IBM also is working with Altera in developing memory technologies that leverage Altera's field-programmable gate arrays (FGPAs). The Con Tutto card allows for in-system experimentation with new memory technologies and with combining flash and DRAM.

9 of 18

Inspur Aims for the High End

Inspur's 4U (7-incH) rack server supports two OpenPower processors, up to 64 memory slots and up to 2TB of DDR3 memory. It offers two PCIe expansion slots and comes with four integrated 1G network interfaces. It also includes application optimization capabilities.

10 of 18

A Server for Chinese Telecoms, ISPs

Chuanghe worked with Tyan to develop the OP-1X, a 1U, single-socket Power8-based server that supports the latest version of Linux and is aimed at China's telecom and Internet industries.

11 of 18

Google Gets Into the OpenPower Game

The search giant a year ago unveiled a two-socket Power8 server motherboard, developed in conjunction with IBM, that Google wanted to use to port its software stack to OpenPower. image: Google planar

12 of 18

Bringing OpenPower, Open Compute Together

Rackspace created an open server design and prototype motherboard that combines OpenPower and design concepts from the Open Compute Project. Rackspace will use the new design in its own data centers to run OpenStack services. image: Rackspace board

13 of 18

Bringing the PowerCore to China

IBM and Suzhou developed the PowerCore CP1, a customized Power chip created for the Chinese market.

14 of 18

Leveraging the PowerCore

Zoom Netcom is basing its new dual-socket RedPower servers on the PowerCore CP1 from Suzhou.

15 of 18

Faster Throughput for HPC and the Cloud

Mellanox's ConnectX-4 adapter based on the Coherent Accelerator Processor Interface (CAPI) found in the Power architecture doubles the throughput of its predecessor, leading to high performance and low latency needed for HPC, cloud, Web 2.0 and enterprise applications.

16 of 18

Leveraging Coprocessors for Developers

The Virtex 7 CAPI-based adapter kit brings together Convey Computer's Eagle line of coprocessors with Xilinx's Virtex 7 FPGAs for Power8 systems. Developers can offload functions to the coprocessors, improving performance and reducing power consumption.

17 of 18

Developer Kit Uses FPGAs for Compute-Intensive Algorithms

Nallatech's OpenPower CAPI Developer Kit for Power8, developed with IBM and Altera, was launched in November 2014.

18 of 18

Nvidia, OEM Partners Show How GPUs Are Used in Data Center

SAN JOSE, Calif.—Nvidia CEO Jen-Hsun Huang is promising that his company will be a significant player in the growing field of deep-learning technology, where neural networks can learn from past experience and improve their performance over time. Deep learning has use in a broad array of areas, from self-driving automobiles to medical services to image and voice recognition. At the GPU Technology Conference 2015 (GTC) here, Huang introduced a new powerful GPU, the GeForce Titan X, as well as Digits DevBox, an appliance powered by the new GPU that runs Nvidia's Digits software to help developers and data scientists researching deep-learning technologies. The new offerings build on Nvidia's efforts over the past decade in expanding the use of its GPU accelerators in mainstream, scientific and high-performance computing (HPC). At the GTC, Nvidia showed off its new offerings, while other vendors,...