Supercomputer

• TAIWANIA 1
• TAIWANIA 2
• TAIWANIA 3

 


High performance computing (HPC) is the foundation of scientific research. As the only national laboratory that provides HPC services for interdisciplinary scientific research projects, the NCHC plans to establish a world leading computing and data storage environment and high performance PC cluster, as well as various engineering and scientific applications and databases, which will be provided to industries, academia, and research institutes. This will help users contribute with their high performance academic output.The NCHC built the largest open supercomputer "TAIWANIA 1" in Taiwan in 2018. The system has a total of 27,600 computing cores, a parallel file system with a total capacity of 3.4PB, OPA 100Gbps high speed network, and overall performance reaching 1.3 PFLOPS. The system is a general purpose supercomputer with a large number of compute nodes and an independent file system, high security mechanisms, and supports multiple fields of application, including physics, chemistry, mathematics, atmospheric, engineering and life sciences. In additional to commercial package software, the system also offers several compilers and MPI libraries for researchers to develop their own programs.




 

---

TAIWANIA 2 supercomputer consists of 252 compute nodes, each containing 2 CPUs and 8 state-of-the-art GPUs, and the host's architecture design and specifications are synchronized with international trends and levels. In the use of big data for deep learning, allowing scientists to demonstrate higher performance, however, in terms of energy conservation, TAIWANIA 2's energy efficiency is up to 11.285 GF/W, computing power is up to 9 PFLOPS, but electricity usage is only 798 KW, which has become the most energy-efficient high-performancecomputing mainframe in Taiwan's history.

TAIWANIA provides fast computing power, large storage space and secure network, and it also becomes the largest domestic data market and model market in Taiwan through the establishment of AI cloud computing platform, providing more real-time and more convenient computing services to the industry, university and institute circles. In the future, 50% of computing resources will meet the needs of intelligent robots led by the government, in addition to prospective projects as well as university and institute projects, such as the self-driving experimental fields, AI innovation and research centers. Another 50% of computing resources will provide innovative industrial use in order to accelerate the development of artificial intelligence related technologies and services in Taiwan, so as to promote industrial innovation in the use of artificial intelligence, such as the applications of financial technology, intelligent manufacturing, intelligent medicine/health, and smart cities.

System environmental specifications:

• Hardware - whole system:

          252 nodes / 9072 CPU cores / 2016 GPUs
          193.5 TB memory
          10 PB storage
          EDR InfiniBand 100 Gbps
          1.2 PUE (Warm Water Cooling)
 

• Hardware - single node:

            Intel Xeon Gold CPU x 2
            Nvidia Tesla V100 w/32GB x 8
            768 GB memory
            240 GB SSD + 4TB NVMe
 

• Software Environment:

            Slurm / Kubernetes
            Nvidia NGC Docker
            Ceph
            Spectrum Scale (GPFS)
            CentOS
 

• AI Framework:

            Tensorflow
            Caffé / Caffé 2
            PyTorch / Torch
            ……and more


 NCHC Headlines   ➤  Domestic self-developed and self-made AI supercomputer "TAIWANIA 2" creates a new record
 Watch the video   ➤  TAIWANIA 2 Next Generation AI Computing Platform (EN)
 For more information   ➤  TAIWANIA 2 & Taiwan Computing Cloud







 

---

  【A new generation of Peta-level high-speed computing host】

With 900 computing nodes, Taiwania 3 is a high-speed computing host with the largest CPU in Taiwan (2021) that is open for the public to apply for services. Taiwania 3 features 50,400 computing cores, and the total capacity of the parallel file system shared with Taiwania 2 is about 9.4PB. With InfiniBand HDR100 100Gbps high-speed network, the overall performance can reach 2.7 peta floating point operations. (PFLOPS).

Taiwania 3 has a large number of computing nodes and independent file system space. Considering the needs of multiple operations when it is first designed, Taiwania 3 can allocate computing tasks to appropriate physical nodes through the scheduling system. To effectively improve system security and avoid leaking sensitive information, a one-time password (OTP) is required to log in to this system.

This system is an all-purpose computing host that can provide computing services in a variety of fields, including physics, chemistry, mathematics, atmospheric sciences, engineering applications, and life sciences. In addition to commercial software packages, the system also provides several compilers and MPI library for researchers to develop their own computing programs.

System environmental specifications:

• Compute node : 2 Intel® Xeon® Platinum 8280 2.4GHz CPU (28 Cores/CPU) and 192GB main memory
• Network interface : 2 Intel® Xeon® Platinum 8280 2.4GHz CPU (28 Cores/CPU) and 192GB main memory
• Operating system : CentOS x86_64 7.8
• Scheduling system : Slurm Workload Manager
• Connection information : t3-c1.nchc.org.tw

 

---

iService 
- Provide the services for computing resources, including TAIWANIA 1, TWCC(Taiwan Computing Cloud), and TAIWANIA 3...etc.

Account Support：iservice@narlabs.org.tw

Tech Support：isupport@narlabs.org.tw