Tag: supermicro

Submerge your supercomputer in liquid

Yellow-Submarine-HeaderA team of boffins have discovered that if you take your supercomputer and immerse it in tanks of liquid coolant you can make it super efficient.

The Vienna Science Cluster uses immersion cooling which involves putting  SuperMicro servers into a dielectric fluid similar to mineral oil.

The servers are slid vertically into slots in the tank, which is filled with 250 gallons of ElectroSafe fluid, which transfers heat almost as well as water but doesn’t conduct an electric charge.

The Vienna Science Cluster 3 system has a mechanical Power Usage Effectiveness rating of just 1.02, meaning the cooling system overhead is just 2 percent of the energy delivered to the system.

This means that 600 teraflops of computing power uses just 540 kilowatts of power and 1,000 square feet of space.

Christiaan Best, CEO and founder of Green Revolution Cooling, which designed the immersion cooling system. “It is particularly impressive given that it uses zero water. We believe this is a first in the industry.”

Most data centres cool IT equipment using air, while liquid cooling has been used primarily in high-performance computing (HPC). But cloud computing and “big data,” could make liquid cooling relevant for a larger pool of data centre operators.

The Vienna design combines a water-less approach with immersion cooling, which has proven effective for cooling high-density server configurations, including high-performance computing clusters for academic computing, and seismic imaging for energy companies.

SuperMicro shouts about CAD factory

Super Micro is debuting its SuperWorkstation kit in the gambler’s Disneyland of Lost Wages.

Super Micro says the server grade systems have been made in mind to give professionals easy maintenance on tool-less designs, along with swappable space for HDDs or SSDs, as well as optical drives and cooling.

Spokesperson Don Clegg said the SuperWorkstations offer a very wide array of CPUs and GPUs, powered by the company’s very own high efficiency power supplies. Plus they’re quiet which.

Among Supermicro’s kit will be the catchy-named Ultra High Performance 4 GPU SuperWorkstation SS7046TG-TRF. Better right that oen down. This machine really shows off, Supermicro says, the iray renderer in Autodesk 3ds Max 2012. It’ll feature up to 192GB of memory and runs on Nvidia Quadro 6000, as well as three Tesla C2075s, which all helps with designing pretty 3D art and more. Intel’s inside, with dual Xeon E5600/5500 CPUs. 

On the other end of the spectum, SuperMicro also has an entry level SuperWorkstation with the similarly catchy SW5037A-T. It’s got Sandy Bridge i3, i5 and i7 support along with a Quadro CPU.

The SW5037A-T is a cheaper option for lighter workloads but still rather good, says SuperMicro, for AutoCAD users.

You can buy SuperMicro’s hardware all over the world, but you’d better check with them first on their website, here

Intel spins shrinking the die in data centre tale

Chip giant Intel outlined its server plans for the future today. It claims it spans the “full range” of needs for data centres based on Itanium (remember that?) and Xeon chips.

The Intel Xeon will dominate the server market by 2014, it said, but will also offer “density optimised” CPUs to help your average corporation along. It is pushing the idea of “micro servers” – machines that use one socket, share a chassis, and can provide four times the servers per rack.

Over the next five years, Intel estimates, micro servers will gobble up to 10 percent of the server market segment. You cannot underestimate Intel – it has already gobbled up the CPU market big time. Sun Microsystems used to own the workstation market. Where is Sun now? Eclipsed by Oracle.

It says it has a number of big players on its side for the micro server concept including Quanta – a huge Taiwanese ODM, Tyan – owned by Mitac now, Dell,  Seamicro and a company called Supermicro which we’ve barely heard of.

Gio Coglitore, director of the unlikely sounding “Facebook Labs”, endorsed Intel’s server roadmaps.

Intel claims that its server CPUs will give up to 2.2 times better system performance per watt over its last generation, while CPU power reduction will be up to a third – that is comparing the E3-1220L (30 Watts) compared to the Xeon L3426 at 45 Watts.

A corporation would surely be foolish not to want to upgrade its data centres to the newest microprocessor, Intel is hinting.

The Itanium was hardly mentioned. Earlier today, we reported massive shutdowns of data centres in Japan. But Intel claimed it wasn’t affected by the earthquake and tsunami that followed. AMD, as usual, was unavailable for comment on Intel’s claims.

Supermicro launches Intel-based uni-processor motherboards

Server firm Supermicro Computer has launched its latest uni-processor (UP) motherboards based on Intel’s P67 and Q67 chipsets and optimised for Intel’s second-generation Core i3 to i7 processors.

The C7P67 and C7Q67 boards employ Supermicro’s server technology with Intel chipsets to deliver high performance and a long list of features, including USB 3.0 and SATA 3.0. They also feature optional disk on module (DOM) support.

Speed won’t be an issue either, as they tout speeds of  5Gb/s or 6Gb/s, depending on whether USB 3.0 or SATA 3.0 is used. Transfer speeds and file conversion speeds will be particularly fast.

The UP systems are aimed at both the work and gaming markets, with Supermicro promising that they are ideal for personal workstations or for gamers who want high-definition video and 3D performance.

A gaming tournament held by Turtle Entertainment, the guys behind the Electronic Sports League, showed off some of the capabilities of the Supermicro systems at CES last week.

AMD proclaims wave of Firestream GPUs

X86 and graphics company AMD today announced its newest Firestream GPU compute accelerators, the Firestream 9350 and 9370, for use in servers and enterprising clouds. The smaller 9350 offers 2.0 TFLOPS single precision and 400 GFLOPS double precision floating point performance, whereas its larger relative, the Firestream 9370, achieves 2.64 TFLOPS single precision and 528 GFLOPS double precision performance.

Both cards come packed with GDDR5 memory, 2GB on the 9350 card and 4GB on the 9370. The 9350 also merely needs to occupy one lonely PCIe slot. Standards such as OpenCL, DirectX 11 and OpenGL are naturally supported by the latest streams of fire. The cards require 150W and 225W of power, respectively.

“Heterogeneous systems in which high-performance GPU and x86 CPU technologies work in tandem can deliver enormous computational power,” said Patricia Harrell, a director of CUDA, sorry, stream computing at AMD.

AMD states it will start shipping the cards in the beginning of this year’s third quarter, so they can be expected sometime in July. Partners and OEMs will offer them in rack servers and expansion systems. AMD did not go out of its way to mention One Stop Systems and Supermicro.

The lesser X86 chip maker recently announced its newest Opteron CPUs, and boy-oh-boy would its Firestream cards just work swell with those server brains. A software development kit (SDK) is also available so “the developer community can harness the combined compute power” of a Firestream-enhanced Opteron system.

AMD did not comment if crossing its streams would lead to total protonic reversal, stopping all life instantaneously and exploding every molecule in a server by the speed of light.

AMD intros Opteron 4000 series for the cloud

AMD is making a bid to capture the data centre and cloud market with a series of low cost processors that also, it claimed, are highly power efficient.

Gina Longoria, director of product management for AMD’s server business, said the Opteron 4000 is designed specifically for cloud and data centre customers – and for SMBs, because AMD claims it offers exceptional value. It claims it’s the world’s lowest power per core server processor.

The Opteron 6000 was launched in March but that’s targeted at performance per watt and expansion for 2P/4P.   But the Opteron 4000 is aimed at energy efficiency, designed for customised machines by OEMs. A number of OEM customers will introduce power efficient machines.

Customers need high density and to fit a lot of servers into a small space, said Longoria. OEMs will design units that meet the need of people without a lot of space. The 4000 uses DDR3 and will be upgradeable to Bulldozer cores in 2011.

AMD claims this series is the first “true cloud server platform”.  These Lisbon chips have six or four core options, 3MB or 2MB of L2 cache, 6MB of L3 cache, IO Virtualisation, Hypertransport 3.0, with  two X16 links of up to 6.4GT/s per link.

The chips have an integrated DDR3 memory controller with 1333MHz  support and up to 21.3GB/s memory bandwidth – power bands are 32W, 50W and 75W.

The chips have four power states – a sleep state, a reduced state when  a temperature limit is reached, remote monitoring, and lower memory voltages of 1.35V compared to the standard 1.5V.

AMD makes some ambitious claims for how much data centres with 10,000 servers can save – with up to 8.7kWH per year, and an estimated  $992,000 yearly savings.

The 4100 series, claims AMD, covers all the bases for low power including chips, chipsets, and memory, as well as software.

The chip can offer a 2P server system below $100. AMD claims that its entry level prices, compared to the Xeon 5600 is half the price, or more.

Dell, GatewayAcer, Tyan, MSI, Gigabyte and Supermicro will produce products in the next month or two. The Lisbon four and six cores that are in the 400 series, and codenamed San Marino and Adelaide, will be displaced by Valencia six and eight core systems next year.