Today is January 9, 2009 UTC

Nvidia makes headway with its GPU-based personal supercomputers

"Forget PCs, forget GPUs or even mGPUs ... Nvidia wants to conquer the SuperComputer"


Tags : GPU   News   nvidia   digitimes   super computer   supercomputers  

Posted on Nov 21, 2008, 12:06 PM UTC by Winston Chim
Source : Digitimes

Nvidia makes headway with its GPU-based personal supercomputers

"Nvidia has begun close collaboration with HPC makers including NEC, Cray, Bull and Hewlett-Packard (HP) to integrate Nvidia Tesla GPUs into systems for the high performance computing (HPC) industry, based on its personal supercomputing strategy.

Nvidia argues that its GPU solutions are attractive to HPC developers because heterogeneous computing, where GPUs work in tandem with CPUs, is making breakthroughs in desktop supercomputing possible. Currently, scientific research is carried out on supercomputing clusters, a shared resource that consumes large amounts of power and costs millions of dollars to build and maintain, and researchers must fight for time on these resources, slowing their work and delaying results, Nvidia claims.

GPU-based Tesla personal supercomputers, on the other hand, can deliver the equivalent computing power of a cluster, at 1/100th of the price and in a form factor of a standard desktop workstation, Nvidia stated.

Nvidia Tesla GPU computing processors are being used in industries such as oil and gas, finance, medical and life science, claims Nvidia. In many cases, processing tasks that are simply not possible on CPU-based clusters and workstations are being enabled by the Tesla 10-series products, the chip maker added. Each Tesla processor has 240 cores, providing one teraflop of processing power, along with 4GB of onboard memory. The new Tesla S1070 1U system features four Tesla 10-series processors – for a total of 960 cores and four teraflops of processing power.

Nvidia recently made a breakthrough in its HPC development when the Tokyo Institute of Technology (Tokyo Tech) announced that the addition of 170 Tesla S1070 1U systems to its Tsubame supercomputer helped deliver a combined nearly 170 TFLOPS of theoretical peak performance, as well as 77.48 TFLOPS of measured Linpack performance, placing the system among the top ranks in the world's Top 500 Supercomputers.

GPU-based Tesla personal supercomputers are also available from numerous HPC vendors including Dell, Asustek Computer and Lenovo."

Comments
Login to post a comment

Don't have an account? Click here to register at Forums3D.com

Latest Articles

Win a Coolermaster ACTS 840 Aluminum Chassis
Here's your chance to win a stunning Coolermaster ACTS 840 Aluminum Chassis.
Announcement: Game Servers Sponsored by DCH
Direct Clan Hosting have kindly provided us with a host of game servers for our in-house clan - ARM3D. The game servers will be available to both private and public use.
CPU3D Exclusive: fnaticMSI enters the 3rd Dimension
CPU3D is proud to announce a new media partnership with the ultimate eSports Pro-Gaming Team - fnaticMSI.
CPU3D Exclusive: Win a Limited Edition Gigabyte Chassis
To celebrate the New Year ... we are please to bring you this exciting competition/prize giveaway which is exclusive to CPU3D only. Here's your chance to WIN a Limited Edition Gigabyte Chassis.
CPU3D Product Awards for 2008
It's almost the end of 2008 and it's now time for our annual CPU3D Product Awards.
CPU3D Special: Cardiff Gaming Society
Supporting local events at the grass roots level is something that's important to CPU3D. We check out the recent LAN party hosted by the Cardiff Gaming Society, organised by students at the University of Wales, Cardiff.
CPU3D Exclusive: Interview with Corsair
Corsair's UK Manager spent a day at our CPU3D labs. We took the opportunity to ask a few questions about their up-and-coming products and what they've got in store for 2009.
Join our CPU3D Folding@Home
Help contribute towards the studies into cures for diseases by protein folding, misfolding, aggregation, and other related distributed computing techniques.

More Articles.