The National Oceanography Centre in Southampton is one of the world’s leading centres for research and education in marine and earth sciences. The oceans play a key role in the climate system but the difficulties of sustaining observations of anything other than the ocean’s surface make good modelling of the ocean critical to advancing our understanding of potential changes in our environment. To simulate these effects, ocean models must use grids with spatial resolutions of less than 10km horizontally and have a vertical resolution of less than 10m. The computing power required to achieve model integrations of worthwhile length (at least 30 model years) and adequate resolution (1/12o longitude= 9.3km at the equator) is enormous. Backed-up with 320 Intel Itanium 2 processors and 1.2 TB of memory organized in a cluster of twenty 16 way NovaScale servers, TeraNova enables an efficient simulation with 608M grid cells. “TeraNova demonstrates impressive scalability and copes extremely well with the intense I/O requirements. “I was pleased to be able to try out the Bull NovaScale servers on a demanding problem” said Dr. Andrew C. Coward. “With sustained access to such facilities we could accelerate our research. My experiences show that Bull can offer a complete solution that meets our needs and can back this up with strong support services.” NOC accessed the TeraNova platform in 2005 to run its 1/12° model with additional relevant physics packages. One of these packages involves a representation of mixing within the ocean, taking into account the local gradient in terms of density and mix preferentially within stratified layers (as does the real ocean). The run achieved on TeraNova is probably the highest resolution global ocean model ever integrated with such complete physics. “Bull is very pleased that new important results in Science are achieved thanks to our NovaScale servers. Bull is committed to bringing HPC solutions with exceptional performance/price ratio to the research community, thus enabling new discoveries” commented Jean-François Lavignon, Director of HPC solutions at Bull.