AMD and Nvidia provide computing power for Singapore’s new supercomputing system
Whether it is AMD’s EPYC series processors or NVIDIA’s A100 computing card, they are currently very popular products in supercomputer systems. In the past period of time, the two companies have relied on such flagship products to occupy the market in the professional field, and these products have used by many customers.
According to The Register, Singapore has selected HP to build a new 40 million Singapore dollar (approximately US$30 million) supercomputer for its National Supercomputing Center (NSCC), which will be deployed in phases from now until 2025 to expand and upgrade Singapore’s high-performance computing capabilities. NSCC provides high-performance computing services for scientific research institutions, universities, government agencies, and enterprises. The supercomputer project has received US$200 million in investment from the Singapore government in March 2019.
The new supercomputing system is based on the AMD EPYC 7003 series processor and NVIDIA A100 computing card architecture system, with a total of 900 nodes. It is understood that the total number of cores of its processors exceeds 100,000, and there are also 352 A100 computing cards, which can provide a total of 10 PFLOPS of computing power. Among them, the A100 computing card will provide 6.8 PFLOP computing power, and AMD EPYC 7003 series processors will provide 3.2 PFLOP computing power. This supercomputer will also have multiple 100 Gbps links, which can be used with the SingAREN-Lightwave Internet Exchange from Singapore’s Advanced Research and Education Network.
NSCC issued an announcement stating that the new supercomputing system uses a green solution based on warm water cooling and is the first such system to be deployed in tropical regions. At the same time, the computing power of the new supercomputing system is eight times that of the current ASPIRE 1 supercomputing system. ASPIRE 1 supercomputing system used in 2016 and is currently operating at close to full capacity to support research projects that require high-performance computing resources.