Home  |  NARL  | 中文 
Sun, September 24, 2017

High-performance Computing

The NCHC provides many kinds of highly stable and efficient computing facilities for Taiwan's government, academia, and industry.

Computing is the foundation of scientific research. As Taiwan’s only national-level supercomputing center, we provide HPC services to science, research, and other interdisciplinary fields. We plan the assembly of advanced computing mainframes, high-efficiency PC clusters, and various engineering and scientific application software and databases. These resources provide domestic researchers a fast and stable HPC environment.

ALPS Taiwan's Largest Open Supercomputer

The Advanced Large-scale Parallel Supercluster (ALPS) or Windrinder, is a supercomputer built in May, 2011, that offers an aggregate performance of over 177 TFLOPS.

ALPS possesses a total of 512 general computing nodes, 32 large memory compute nodes, and over 25,600 cores. Each general compute node contains four AMD Opteron 6174 processors, and each large compute node contains four AMD 6136 processors as well as 256 GB memory.

ALPS - Acer AR585 F1 Cluster
Performance 177.1 TFLOPS
Compute node
General compute node:
• Total 512 nodes
• CPU –AMD Opteron 6174 2.2GHz ×4
• Memory –DDR3 ECC 128GB
Large memory computing node:
• Total 32 nodes (Two nodes upgraded to 512GB memory in 2013)
• CPU –AMD Opteron 6136 2.4GHz ×4
• Memory – DDR3 ECC 256GB
Interconnection Intel Qlogic QDR Infiniband ×2
Storage Lustre 2.5.39, total disk space 1074 TB
IBM Cluster 1350 is an x86-based parallel computing cluster system built in 2007. It has 512 computing nodes (i.e. 2,048 cores) and a General Parallel File System (GPFS). Its topology is 100% non-blocking FatTree with an InfiniBand DDR interface for interconnection. Iris ranked the 35th on the 2007 TOP500 list.
In 2009, 128 compute nodes (i.e. 1,028 cores) were added to Iris via Ethernet. The group of compute nodes that utilize Infiniband is called 1350. The other group of nodes that utilize Ethernet is called 1350A. The entire system, in total, has 640 compute nodes, 3,072 computing cores, and 70Tb of storage.
IBM Cluster 1350 is suitable for general purpose computing. Its primary user domains include atmospheric science, typhoon/flooding, computational chemistry, computational fluid dynamics (CFD), and computational engineering.
Iris—IBM Cluster 1350/1350A
Performance

24.9 TFLOPS

Compute Node IBM X3550 server x 512 ; each server includes:
IBM X3550 x512;each server includes:
.CPU-Intel Xeon 5160 3.0GHz x2
.Memory- 16GB DDR2 ECC

IBM X3550 x128;each server includes:
.CPU-Intel Xeon 5450 3.0GHz x2
.Memory- 16GB DDR2 ECC
Interconncection 1350: Voltaire DDR Infiniband
1350A: Gb Ethernet
Storage 70 TB

Formosa 5 Cluster Taiwan’s Largest GPU-based Cluster System

Carrying on the heritage of the Formosa series, the NCHC continues to innovate by successfully building the Formosa 5 system-- the largest cluster computer in Taiwan designed specifically for GPU computing!
As demand for computing resources has steadily increased, so have the demands for new and different platforms and hardware. By taking advantage of many-core parallelism, the CPU working with the GPU has become the mainstay of HPC architecture. Formosa 5 is made up of 88 computing nodes, each equipped with two 6-core central processing units, three graphic processing units, and 96 GB of memory, all sharing an InfiniBand 40 Gbps high-speed network! After tuning, the NCHC's Formosa 5 achieved an optimal performance of 89.94 TFLOPS.
Formosa 5 began operations in August, 2012, and thus began providing Taiwan’s academic communities with the latest in GPU processing.
Formosa 5 Cluster
Performance 89 TFLOPS
Compute Node
88 Intel Xeon X5670 2.93 GHz dual socket; 12 cores/node;
96 GB DDR3-1333 MHz ECC/REG RAM;
3 Nvidia Tesla M2070 GPUs/node;
120 GB Intel SSD per node
Interconnection Mellanox 4X QDR (40Gb/s) InfiniBand
Storage

22 TB


Formosa 3 Cluster – Taiwan's First Render Farm & IaaS Service Platform

Formosa 3 is the name given to the NCHC’s self-built cluster cloud computing and 3D rendering system. Combining the NCHC’s solid strength in cluster computing and cloud technology, it provides complete and reliable cloud computing and 3D rendering services.
Currently, Formosa 3 provides Infrastructure as a Service (IaaS), thus, offering its users virtual machine management, complete customization and control, and easy installation and integration.
Render Farm is the name given to the NCHC’s self-built 3D rendering machine. In order to provide complete services and exceptional rendering performance, the Render Farm cluster utilizes a Linux operating system as well as a professional rendering queuing system. Users are able to dispatch rendering jobs and monitor their computing status via a graphical user interface (GUI), whether using Linux, Windows, or Mac OS.
Formosa 3 Cluster
Performance 9 TFLOPS
Compute Node
76 computing nodes; each computing node consists of:
CPU – Intel Xeon x5660 six cores 2.8GHz x2
Memory – 48GB DDR3 ECC
Interconnection 4x QDR Infiniband
Storage
IaaS:18 TB
Render Farm:28 TB




Home | Top | Back