AUGUR TBBS Pty Limited
ABN 66 077 857 348
 

HPC Cluster

We have a number of machines in a Cluster of Workstations arrangement. They are grouped by their intended workloads.

We have a HT Condor submit node with 10GbE network connectivity and local SSD.

We have two storage servers, both with 10GbE network connectivity, terabytes of space running OpenZFS for data integrity.

All compute nodes have local storage as well as having access to a storage server.

In order to keep up to date and for reliability older hardware is regularly replaced.


Compute nodes

For CPU-only workloads we use AMD Ryzen 9 5900X processors with 12 cores/24 threads, 64GB of memory and 1TB of local SSD storage.

We have an Ampere Altra with 80 cores/threads, 256GB of memory and 1TB of local SSD storage. We can add SAS storage to it as needed.

For CPU, CUDA or OpenCL workloads we use 3rd generation AMD Ryzen 7 processors with 6 cores/12 threads, Nvidia RTX 3060 Ti graphics, 32GB of memory and 1TB of local SSD storage.


Raspberry Pi

These are suitable for low power compute roles. We have a mix of Raspberry Pi 3 model B+ and Raspberry Pi 4 model B. The Pi 3 nodes are being replaced by the Pi 4 as we can get them.

Raspberry Pi 3 model B+
These have ARM v8 (ARM Cortex-A53) processor with 4 cores, Videocore IV graphics, 1GB of memory and local storage is a 16GB micro SD card.

Raspberry Pi 4 model B
These have ARM v8 (ARM Cortex-A72) processor with 4 cores, Videocore VI graphics, 8GB of memory and local storage is a 32GB micro SD card.

The Raspberry Pi’s have their own internet proxy server, storage server and a GPS time server.


Trademarks

AMD and Ryzen are trademarks of Advanced Micro Devices, Inc (www.amd.com)

CUDA is written by Nvidia, Inc (www.nvidia.com/object/cuda_home.htm)

Intel and Core i7 are trademarks of Intel Corporation (www.intel.com)

OpenCL is being created by the Khronos Group (www.khronos.org/opencl)

Raspberry Pi is a trademark of the Raspberry Pi Foundation (www.raspberrypi.org)

Contact us to discuss your processing needs.



Home   |   Contact Us  |   iOS BOINC Client  |   HPC Clusters