ASUS-Built Supercomputer With NVIDIA HGX H100 Ranked in Global TOP500 and Green500

Delivers 1.23x higher effectivity and greater than 45 PFLOPS

ATLANTA, GA & TAIPEI, TAIWAN / ACCESSWIRE / November 20, 2024 / ASUS and Ubilink at the moment introduced the profitable institution of a supercomputing facility that has positioned Taiwan within the top 31 on the global TOP500 list of supercomputers and secured a prestigious one 44th place on the Green500 listwhich yields 45.82 PFLOPS. This bold challenge, outfitted with 128 ASUS ESC N8-E11 servers, every with NVIDIA HGX H100 servers for a complete of 1,024 GPU playing cards, will redefine the requirements of AI computing. As the most important knowledge middle in Taiwan, the achievement of Ubilink – an organization based by Foxlink, Ubitus and Shinfox Vitality – and assisted by ASUS illustrates the ability of strategic collaboration and permits extraordinary efficiency milestones in synthetic intelligence. This world-class knowledge middle can be notable for giving prospects the liberty to decide on packages that use solely renewable power.

“The strategic partnership with ASUS has turned our imaginative and prescient into actuality, permitting us to construct a world-class supercomputing facility that can allow the following wave of AI innovation in Taiwan and past,” mentioned Michael Chen, Common Supervisor of Ubilink. “The corporate’s superior software program options and complete providers have enabled extraordinary efficiency in a outstanding time-frame of simply three months.”

Highlights of this groundbreaking supercomputing challenge embody:

• World Class Facility: The brand new inexperienced power AI computing middle can have a velocity of 45.82 PFLOP/s, making it the most important of its form in Taiwan.

• Modern options: The information middle will provide public cloud providers, AI computing energy rental, and subscription-based providers for coaching giant language fashions, with an emphasis on good manufacturing and dwell streaming.

• Superior infrastructure: Geared up with 128 NVIDIA HGX H100 servers and 1,024 GPU playing cards, this facility will redefine the requirements of AI computing.

• Quick implementation: The development of the information middle concerned the ASUS Infrastructure Deployment Heart (AIDC) instrument for shut administration of infrastructure, cooling programs and superior automated instruments, guaranteeing reliability and effectivity.

• Superior Computing Effectivity: The ASUS-Ubilink challenge makes use of NVIDIA HGX H100 SXM5 80GB InfiniBand NDR400 servers with a complete of 128 nodes of ESC N8-E11. The ASUS Resolution Efficiency Staff measured the efficiency and achieved an Rmax of 45.82 PFLOPS, with an effectivity (Rmax / Rpeak) of 66.08%. As proven within the desk under, below the identical GPU and high-speed community card situations, the ASUS infrastructure answer is extra environment friendly and delivers 1.23x the general efficiency of rivals.[i]

ASUS, because the infrastructure supplier for this challenge and related large-scale efforts worldwide, delivers unparalleled computing energy and stability, reinforcing its dedication to driving AI innovation.

Learn extra about this success story: https://servers.asus.com/stories/Ubilink-supercomputing-center.

Availability and costs

ASUS server options can be found worldwide. Please contact your native ASUS consultant for extra info.

Press contact:
Olivia Liu
Product marketer
[email protected]

Notes for editors:

ASUS Server web site: https://server.asus.com

ASUS AI server: https://www.asus.com/event/ai-server-solutions

ASUS Enterprise web site: https://www.asus.com/business

ASUS LinkedIn: https://www.linkedin.com/company/asus/posts

ASUS Enterprise LinkedIn: https://www.linkedin.com/showcase/asus-business

ASUS Press Room: https://press.asus.com

ASUS International Fb: https://www.facebook.com/asus

ASUS International https://www.x.com/asus

[i] Primarily based on knowledge from Top500.org. All knowledge used on this evaluation comes from the revealed info of the TOP500 checklist, November 2024. We assess the efficiency of every knowledge middle by evaluating its computing effectivity, represented because the ratio of Rmax to Rpeak. For extra particulars please go to: ASUS HGX H100 AI servers deliver 1.23x performance efficiency | ASUS servers

[ii] Model A makes use of NVIDIA HGX H100 SXM5 80GB, InfiniBand NDR400 with a complete of 190 nodes

[iii] Model B makes use of NVIDIA HGX H100 SXM580 GB, InfiniBand NDR400 with a complete of 64 nodes

[iv] Effectivity is measured by Rmax/Rpeak, the place larger effectivity signifies higher pc efficiency.

Contact particulars

Brian Ngo
Senior product advertising supervisor
[email protected]

SOURCE: ASUS

View the unique press release at newswire.com.