Archive notice: This article was originally published on May 6, 2009. Links and embedded videos are preserved as part of the historical record.
The new standard solution was designed to bring GPU computing into existing data-centre infrastructures quickly and easily.
The preconfigured Tesla cluster solution was said to offer up to 30 times the performance of CPU-only systems. This was intended to meet complex and growing application requirements in fields such as molecular research, seismic exploration and financial analysis. NVIDIA also said a Tesla solution used significantly less energy, reducing operating costs.
BNP Paribas’s Corporate and Investment Banking division had recently replaced 500 conventional CPUs consuming 25 kW with a smaller cluster comprising CPU servers and two Tesla S1070 1U systems, consuming only 2 kW. By using Tesla GPUs and accelerating its applications, BNP Paribas reportedly improved energy efficiency by a factor of 190.
“There are 15 to 20 million engineers, scientists and researchers worldwide who have only limited access to computing time on supercomputers,” said Andy Keane, general manager of NVIDIA’s Tesla business. “With our new, preconfigured Tesla cluster, any of them can quickly and easily build a GPU-based supercomputing cluster that significantly reduces energy consumption while accelerating their work.”
The preconfigured Tesla clusters consist of x86 CPU servers and NVIDIA Tesla S1070 systems. Configurations start at 16 teraflops with four Tesla S1070 systems, each containing four Tesla 10-series GPUs. All systems include host servers, InfiniBand switches and the required cabling. They can also be customized to meet specific customer requirements.
The preconfigured Tesla cluster was available from Tesla GPU Preferred Partners including CADNetwork, FluiDyna and Megware.
