Taipei, Saturday, Jul 20, 2019, 06:46


Taiwan Commences Operation of First Peta-Level Supercomputer “Taiwania”

By Korbin Lan
Published: May 09,2018

TAIPEI, Taiwan - Taiwan’s National Applied Research Laboratories (NARLabs) on May 8 announced that a new-generation Peta-level super high-speed computing mainframe entitled “Taiwania” had begun operations and will provide Taiwan with important high speed computing for use in advanced research.

More on This

NARLabs Taiwan Semiconductor Research Institute Officially Open

TAIPEI, Taiwan - NARLabs Taiwan Semiconductor Research Institute on January 30 was formally unveiled as the world’s on...

NARlabs Launches the World’s First Multi-sensor Integrated MCU

TAIPEI, Taiwan — The National Applied Research Laboratories (NARLabs) lauched the world’s first multi-sensor integra...

In order to solve Taiwan’s problem of insufficient high-speed computing resources, NARLabs’ National Center for High-Performance Computing constructed this new-generation supercomputer, which is entitled “Taiwania” after an endemic species of coniferous trees in Taiwan. Similarly to the “Taiwania” species of tree, the Peta-level supercomputer is expected to function as an important cornerstone for stabilizing scientific and technological research in Taiwan.

Taiwania is equipped with 630 pure CPU computing nodes along with 64 CPU GPU accelerator computing nodes, 25,200 Intel CPU cores, 256 NVIDIA P100 GPU and a gigantic storage capacity of 3.4 Pbytes (1 Pbyte is the equivalent of 1,024 Tbytes).

In terms of CPU computational capacity, it can perform up to 1.33 PFLOPS of calculation measurements; however, with the addition of the GPU calculation capacity, it can exceed 1.7 PFLOPS of calculations. “Taiwania” is the first public computing host with Peta-level computational capacity.

In addition to featuring faster computational speeds, “Taiwania” is also more energy efficient and takes up less space, giving it the benefits of better usability additional safety. In terms of energy usage, if features energy efficiency of up to 4 GFLOPS/W and offers as many as 10 times more computations per watt that the previous generation of supercomputers. Meanwhile, it covers an area of approximately 10 pings, which is only one third of the space required for the previous generation of supercomputers.

(TR/ Phil Sweeney)

CTIMES loves to interact with the global technology related companies and individuals, you can deliver your products information or share industrial intelligence. Please email us to

1340 viewed

comments powered by Disqus