Appro introduced the new Appro Xtreme-Cool Supercomputer liquid-cooling architecture. The system will have limited production availability starting in early Q1 2013.
According to the company, Appro Xtreme-Cool offers the same features and benefits of the current Appro Xtreme-X air-cooled solution with improved energy savings, lower total cost of ownership, and faster return on investment by requiring fewer or no air conditioning units in the datacenter. It uses warm water liquid -cooling heat exchangers with no chillers, reducing typical energy consumption used to cool the data center by 50% while dropping PUE lower than 1.1. It also produces 80% heat capture to the warm water for possible heat reuse, the company says.
The system is composed of blade nodes that are typically installed in a cluster architecture rack, and the liquid cooling installed in the node is connected to the Coolant Distribution Unit (CDU) via tubes with drip free quick connects. The liquid cooling is specifically designed to be interconnected in the node, directly cooling processors and memory in an isolated, low-pressure secondary loop to more efficiently remove heat compared to air-cooling. Integrated remote power and temperature monitoring and reporting is also provided. In addition, all processors, GPUs, coprocessors and memory are Field Replaceable Units (FRUs).
It can be configured as Fat Tree or 3D Torus architecture with interconnect options for single or dual rail, InfiniBand or Ethernet. The system is composed of two processors per node, supporting approximately 80 nodes per rack in a standard 42U rack based on Intel Xeon processor E5 Family. It also supports hybrid processing based on Intel Xeon processors paired with Intel Xeon Phi coprocessors.
It uses quieter low-power fan system operation for a better work environment and an option for a 480 volt power distribution with a choice of 208 or 277 volt power supplies for further energy savings, the company says. The system features Appro's HPC Software Stack, including Appro's Cluster Engine.
It is designed to address medium-to-large datacenters HPC deployments with up to 25 petaflops of computing performance.
For more information, visit Appro.
Sources: Press materials received from the company and additional information gleaned from the company's website.