For 2013, NCI will provide to the NCMAS 15% of its new Fujitsu Primergy petascale system with a peak performance of 1.2 PFlops. The Fujitsu cluster has 57,454 cores (Intel Sandy Bridge 2.6 GHz), 158 TBytes of memory, and 10 PBytes of (usable) short-term storage. Access to the current peak system, vayu (Sun Constellation cluster), will be available for some months from the beginning of 2013, until users have transitioned to the Fujitsu system, after which it will be decommissioned.

For 2013, iVEC will contribute 15% of its two facilities -- the iVEC@Murdoch supercomputer called Epic and the iVEC@UWA GPU-accelerated system called Fornax. This equates to:

• Epic -- 11,000,000 core hours (2,750,000 per quarter) and 45 Terabytes of scratch space (temporary storage for working data)
  Each node contains two 6-core Intel Xeon X5660 processors and 24 Gigabytes of memory.
• Fornax -- 1,378,944 core hours (344,736 per quarter), which equates to 114,912 GPU hours (conversion from SUs to GPU-hours is achieved by dividing by 12). There is also 45 Terabytes of scratch space available, as for Epic.
  Each node consists of 2 6-core Intel Xeon X5650 processors plus an NVIDIA Tesla C2050 GPU accelerator, and 72 Gigabytes of memory. A job is always allocated a whole number of nodes. When estimating computational-time requirements, one should quote the Service Units requirements based on the number of node-hours the GPU-/CPU-code will require multiplied by 12.
  GPU-based codes are more appropriate on Fornax given the relatively small size of the CPU capability in comparison to Epic. However, Fornax does have a large memory-per-node capacity (72GB), which may make it appropriate for some software that is not GPU-accelerated but that require significant system memory. Applications to use Fornax for non-GPU simulations must clearly demonstrate the requirement for large-memory nodes.

The NCI-SF Bioinformatics seeks applications from Bioinformaticians and Computational Biologists, and will contribute 32% of its HPC facility (8,760,000 core hours). Its associated support team will provide HPC and Bioinformatics consultancy services.
The facility provides heterogeneous hardware to enable Bioinformatics tool users to scale up their work without having to modify their code:

• 335 nodes with 24GB RAM
• 28 nodes with 72GB RAM
• 3 nodes with 1024 GB RAM

The Multi-modal Australian ScienceS Imaging and Visualisation Environment (MASSIVE) is a specialised Australian HPC facility for computational imaging and visualization. The facility caters primarily to two types of users:

1. Those with specific needs in imaging and visualisation including users of synchrotron x-ray, infrared and diffraction imaging, functional and structural MRI, CT, electron microscopy and optical microscopy. The facility provides specific support to researchers to process data "in-experiment", generate large-scale visualisations, and analyse large-cohort and longitudinal imaging studies.
2. Researchers who will make good use of MASSIVE hardware, including GPU hardware.