This is the first large Opteron cluster that I have seen published. What is especially interesting is the very dense blade arrangement and the use of the new 10 gigabit Infiniband technology. This combination should set new performance numbers for computing that requires really fast communications.
A quick look at the Appro site also indicates the ability to scale these clusters into thousands of processors. The design and construction of the Hyperblade Servers is worth a look - 80 dual processor blades in 5 chassis in one tower. Check at Appro Cluster for full details and click on the various tabs for information.
Excerpt from the Appro press release:
64-Bit Processing Power and 10 Gbps Interconnect Yield Powerful, Large-Scale Clustering Solution To Support Broad Spectrum of Computational Research Projects
Texas A&M University has purchased two Appro HyperBlade Clusters consisting of 130-nodes with a total of 260 AMD Opteron processors. The High-Performance Computing (HPC) Solution includes Appro BladeDome Remote Management Software and InfiniCon's InfinIO 7000 Shared I/O and Clustering System. HPC configurations based on Appro and InfiniCon technology are easy to scale; servers can be added with plug-and-play ease to build a clustering fabric that can support thousands of nodes.
Appro's HPC cluster solution will support multiple research activities at the university such as hydrology, groundwater and oil and gas reservoir simulations. Texas A&M's projects were funded in part by a grant from the National Science Foundation.