P E T A F L O P S
for High Performance Computing (HPC)
PetaFlop computing power is achieved with a combination of high clock speed multi-core processors, high bus speed and vast numbers of high-speed computing cores on Compute Accelerators and GPUs performing efficient parallel processing. Where multiple nodes are connected this is often performed with low latency and high transfer rate adaptors and switches.
PetaFlops for HPC Hardware
We base our solutions upon the latest HPC technology from HPe, Lenovo, Gigabyte, HP, Nvidia and AMD.
HPe, Lenovo and Gigabyte have all developed compute servers specific for high performance computing and massive storage servers for data analytics. The compute servers feature enormous power supplies and a large number of IO slots to support multiple compute or GPU cards. We have several that have adopted Nvidia’s NVLINK/SXM2 bus.
We have a great range of GPU-accelerated High-Performance compute hardware from Hpe, Lenovo, Gigabyte HP, Nvidia and more.
Please call us on 1300 00 8100 to discuss your specific requirements and to take advantage of our vendor-independent advice.
We are also able to assist you with affordable systems for pilots and to leverage GPU-assisted hardware in the cloud where appropriate.
Petaflops for HPC Software
Most High-Performance Compute systems run Linux with docker or similar and containers for each workload.
Many of the common applications and development environments used for AI and Scientific Research have been GPU Enabled. Nvidia provide a large number of application and development environment container specifically optimised for the latest GPU-enabled compute servers
Recent advancements including Nvidia vComputeServer allow this software stack to run virtualised on VMWare ESXi with the capability to virtualise GPU resources across virtual machines. Dynamically changing GPU resources means that multiple high-performance GPUs may be allocated to demanding workloads and partial GPUs allocated to less demanding workloads