by Anton Shilov
07/05/2010 | 10:13 PM
Advanced Micro Devices has announced that some of its partners started to build a cluster based on the twelve-core AMD Opteron 6100-series microprocessors that features AMD FireStream compute cards for highly-parallel computing tasks. This is the first cluster to feature code-named Magny Cours central processing units along with FireStream accelerators.
The cluster to be located in Goethe University, in Frankfurt am Main, Germany, will combine one thousand five hundred and forty four (1544) twelve-core AMD Opteron processors (a total of 18 528 cores) and 772 AMD GPUs (one GPU card per 2P node). The cluster will be build by ClusterVision, an Amsterdam, the Netherlands, based company. The machine will serve a variety of purposes from physics to chemistry, life sciences and computer research.
To help bring the whole set of servers together into a cohesive cluster, Bright Cluster Manager software will be utilized and Mellanox quad data rate InfiniBand will serve as the interconnect between the nodes. To help maximize data center space, the cluster will be built on a SuperMicro “twin” platforms that provide 2 motherboards in a 1U rack server chassis. This allows the 772 total servers to be located in 386 physical servers.
“We are very happy to see this project progressing with AMD technology because it helps us continue our efforts towards bringing GPU and CPU together into a more seamless computing model. Just as we have predicted that in the future, the possibility for GPU and CPU coming together in a single compute model is looking more and more probable, some organizations are not waiting. They see the potential for greater performance and are optimizing their software today to be able to take advantage of this technology convergence that is headed our way,” said John Fruehe, the director of product marketing for server/workstation products at AMD.
At present AMD FireStream and competing Nvidia Tesla accelerators are only used in high-performance computing space and not on general-purpose servers. However, as more HPC specialists learn how to take advantage of graphics processing units (GPUs), the latter get higher chances to open up new markets for them.