Tag: Mellanox

IBM pushes for UK raw power

IBM data centreIBM said it is to create a UK Power design centre at Hartree in conjunction with Nvidia, Mellanox and Science and Technology Facilities Council (SFTC).

Power is shorthand for IBM’s microprocessor family, and the centre will focus on high performance computing for modelling, simulation, and big data analysis.

The UK government had recently announced investment into the Hartree Crntre at Daresbury, in the belief it will boost economic growth.

Dr Peter Allan, acting director of the Hartree Centre said the new design centre is aimed at helping industry and academia use IBM and Nvidia’s expertise in supercomputing.

Dave Turek, an IBM VP of high performance computing, said the centre will extend existing acceleration and design centres in Hermany and France, based on the OpenPOWER technologies.

Qualcomm is testing its server chips

qualcomQualcomm has been explaining how it is going to get its chips into the server market and is apparently testing its Server Development Platform (SDP)

Anand Chandrasekher, senior vice president of Qualcomm demoed its SDP and chatted about the company’s progress within the server market.

Apparently Qualcomm is sampling its chips into tier-one data centres using a pre-production 24-core SoC based on the ARMv8-A instruction set and built using advanced FinFet technology.

Qualcomm claims that the use of its chips in server set ups can reduce costs by more than 40 percent.

Qualcomm is “partnering” with Xilinx to create heterogeneous computing solutions for data centres using Qualcomm’s server processor and Xilinx FPGAs;

Another of its chums is Mellanox which is designing cost effective platforms for servers and storage using interconnect for data transfer and analysis with Qualcomm’s server CPU and Mellanox’s Ethernet and InfiniBand interconnect;

Chandrasekher said the release of Qualcomm’s evaluation system is a milestone.

“As data centres evolve to support the exponential growth and innovation in data, connectivity and cloud services, Qualcomm Technologies is creating an ecosystem to meet the needs of these next-generation data centres with Qualcomm-based server technologies,” he said.

Customers were testing the Server Development Platform and porting their software. Qualcomm was incorporating their feedback into the product offering with the goal of ensuring system and software readiness by the time we are in full production.

 

IBM takes aim at exaflop computing

IBM client centre, Montpelier, FranceBig Blue said it is collaborating with French high performance agency GENCI to create computers that perform at least one exaflop.

An exaflop represents a billion billion calculations in a second.

IBM said that the fastest systems in the world now operate at between 10 and 33 petaflop and that’s only around one to three percent the speed of exascale computing.

The two will collaborate for 18 months and create complicated scientific applications at speeds of over 100 petaflops.

They will use technologies of over 140 members of the OpenPOWER foundation, all of which concentrate on using IBM’s POWER microprocessors.

Some examples of how this may pan out is connection Nvidia GPU accelerators to the POWER processors using Melanox 100Gb/s Inifiniband switches.

IBM will use its design centre in Mellanox to test and bring into existence much faster applications.

GENCI and IBM appear to be confident that they’ll move the supercomputing industry much closer to the exascale level.

Nvidia teams up with Big Blue and Mellanox

IBM client centre, Montpelier, FranceMellanox, Nvidia and IBM have opened a design centre in France dedicated to research into using the POWER platform for commercial, industrial and scientific applications. The centre is situated in its existing building (pictured).

In case you haven’t been following the story so far, IBM has wheeled heaps of companies into the OpenPOWER foundation – and the centre in Montpelier will give open source software developers technical assistance to develop high performance computing (HPC) applications.

Mellanox keeps itself out of the limelight but it is responsible for Infiniband, while Nvidia has its Tesla advanced computing platform. IBM brings its licensable POWER architecture in the picture.

The French centre is the second such establishment – the three opened a centre in Germany last November and are the founding members of the foundation.

The aim is to help solve big challenges scientists and others face using HPC architectures and technologies.

IBM’s VP of HPC, Dave Turek, said the opening of the centre shows his firm is committed to open source collaboration. Nvidia brings its NVLink graphics processor interconnect to the party, while Mellanox’ Infiniband offloads data to the network level.

Mellanox to buy Voltaire for $218 million

Mellanox Technologies has announced that it is to acquire data centre fabric manufacturer Voltaire for approximately $218 million.

The all-cash deal will see Mellanox pay $8.75 per share in Voltaire, which Mellanox believes will strengthen its position in the server and supercomputer sectors.

Mellanox was founded in 1999 and makes connection devices for servers and supercomputers, including the InfiniBand adaptors for China’s first petaflop supercomputer in May of this year. It claims to be the company of choice for server and supercomputer connectivity, particularly for the Fortune 500 firms.

Voltaire was founded in 1997 and is also big in the server and supercomputer game, making computing fabric for data centres, supercomputers, and cloud systems. It supplies more than 30 percent of the Fortune 100 companies and many of the Top 500 supercomputers, making it a perfect match for Mellanox.

As part of the acquisition, Voltaire’s CEO, Ronnie Kenneth, is expected to join Mellanox’s Board of Directors in 2011, subject to approval at the next Mellanox meeting.

With both companies combined Mellanox will have roughly 700 employees and a revenue base of $217 million for the year ending September 30. Mellanox also expects a cost synergy of at least $10 million by 2012.

The deal was approved unanimously by Voltaire’s Board of Directors and is expected to close within the first quarter of 2011.

Oracle’s Ellison declares Word War III on HP and SAP

The formerly somewhat insouciant CEO of Oracle – the man who possesses one pair of shoes that are worth more than an average workers’ lifetime wardrobe – lost it tonight.

Not his pair of shoes, his cool.

In other news, Oracle made a strategic investment in Mellanox, perhaps more important than this personal matter. It took a 10.2 percent stake in Mellanox, the outfit that runs Infiniband – a venture started by Intel – Intel is perhaps Oracle’s temporary friend. The software giant “has no plans” to take over Mellanox.

In a slightly somewhat unprecedented attack on HP, Larry Ellison said that the inking company should change its tagline from “Invent” to “Steal”. It’s not the first time. He lit out against HP and ex-SAP CEO Leo Apotheker last night, too. Last night he appeared to be claiming that Apotheker was hiding in a foreign jurisdiction because he was too frit to be subpoenad in Oracle’s continuing case against the German combine.

Larry Ellison is taking it very personally that Ray Lane, his former henchman at Oracle in the 1990s and now chairman of HP, is defending Leo Apotheker, formerly the chief SAPman.

Said Ellison in a statement:  “HP Chairman Ray Lane has taken the position that Leo Apotheker is innocent of wrongdoing because he didn’t know anything about the stealing going on at SAP while Leo was CEO. The most basic facts of the case show this to be an absurd lie. Oracle sued SAP for stealing in March of 2007. Leo became CEO of SAP in April of 2008. Leo knew all about the stealing. In fact, Leo did not stop the stealing until seven months after he became CEO. Why so long? We’d like to know. Ray Lane and the rest of the HP Board do not want anyone to know. That’s the new HP Way with Ray in charge and Leo on the run. It’s time to change the HP tagline from ‘Invent’ to ‘Steal’.”

Perhaps Ray Lane would like to speak about his time at Oracle. We invite him, as we invited Carly Fiorina before, to use online journalism as a forum where things can be hammered out. We know where you got your shoes.

 

Mellanox builds China's first petaflop supercomputer

Mellanox has announced that its ConnextX-2 adaptors will be used in a new petaflop supercomputer called the Mole-8.5, which will use the InfiniBand adaptors to give a transfer speed of up to 40GB/s.

The ConnextX-2 40GB/s InfinBand adapters come with GPU-Direct technology and an IS5600 648-port switch with FabricUT fabric management software and fibre optic cables, which it is providing for the Institute of Process Engineering (IPE) at the Chinese Academy of Science for what will be the first petaflop GPGPU in China and will offer one of the fastest speeds in GPU-to-GPU communications.

A petaflop is 10 to the power of 15 (1,000,000,000,000,000) floating point operations per second, a term used to measure a computer’s performance, particularly in the scientific community which makes heavy use of floating point calculations. IBM’s military supercomputer “Roadrunner” was the first to reach the petaflop speed in May 2008, but the Mole-8.5 is the first of this speed in China.

TechEye spoke to Brian Sparks, Senior Director of Marketing Communications at Mellanox, who told us that the advantages of the new speeds capable depend on the application being used. “In some cases we see double performance increase from going to 20GB/s to 40GB/s.”

The Mole-8.5 supercomputer is already in use at IPE to conduct scientific simulations in chemical engineering, material science, biochemistry, data and image processing, oil exploitation and recovery and metallurgy, but the new adaptors will up the performance and open new avenues for scientific research.

“By incorporating Mellanox 40Gb/s InfiniBand with GPU-Direct technology, we have been able to conduct scientific simulations using GPUs at performance levels that we would never have been able to achieve using a different interconnect,” said Dr. Xiaowei Wang of IPE. “The new Mole-8.5 Petaflop cluster, with industry-leading interconnect performance and efficiency, enables us to shorten the time it takes to run applications that are critical in the process of scientific discovery.”

We asked Sparks what the term Mole-8.5 means and he said it “came from IPE” and Mellanox is “not sure what it means”. The Mole-8.5 system was designed to achieve high efficiency in real applications with a low cost in establishment and power consumption. By using Mellanox InfiniBand adaptors GPUs compute at a much faster rate, increasing the performance of applications run on the Mole-8.5 supercomputer.

Sparks told us that Mellanox is not resting on its laurels and is now working on “even more application productivity and system efficiency with new technologies such as CORE-Direct that offloads the MPI communications from the CPU and GPU-Direct which offloads the CPUs involvement when handling GPU memory data from the CPU to the interconnect.”

We’re not sure what all that means but we are happy to tell Mellanox that a mole is an animal from the Talpidae family in the order Soricomorpha. It could also refer to a spy working secretly from an organisation or country. Hope this helps!