All-AMD Powered & World’s First Exascale Supercomputer, Frontier, Has Been Running Into Issues Ever Since It Powered On - Wccftech

1 year ago 33

Oak Ridge National Laboratory oregon ORNL is the location to the Frontier supercomputer. Frontier is marked arsenic the archetypal exascale-level strategy created utilizing AMD's EPYC Trento CPUs & Instinct MI250X compute Accelerators. The full strategy makes usage of HPE's Slingshot interconnects. It is besides slated arsenic the world's fastest supercomputer disposable and is the world's lone operational Exascale design.

The Cray EX architecture by HPE was created for large-scale applications that researchers would beryllium capable to entree to assistance successful technological probe starting successful 2023. However, the supercomputer cannot tally an full time without several failures located wrong the hardware.

The ORNL Frontier boots up but tin lone nutrient a maximum of 1 FP64 ExaFLOPS, whereas the strategy was designed to present 1.685 FP64 ExaFLOPS. While nary connection has been fixed regarding the circumstantial issues, a fewer rumors are coming to light.

All-AMD Powered & World's First Exascale Supercomputer, Frontier, Has Been Running Into Issues Ever Since It Powered On 2

First, the Slingshot interconnects, the web created for HPE Cray supercomputers, conflicts with the HPE clusters. Unfortunately, the specificity of the nonstop contented is unknown. Secondly, the AMD Instinct MI250X compute GPUs and the EPYC Trento CPUs are rumored to struggle with the Slingshot interconnects. Again, nary authoritative connection has travel from the task leads oregon researchers of the ORNL Frontier supercomputer.

Mike Bernhardt, the Department of Energy's (DOE) Exascale Computing Project, states that the afloat integration of ORNL Frontier volition beryllium disposable to researchers starting adjacent twelvemonth but is not quoted arsenic having immoderate concerns oregon issues with the afloat motorboat of the Frontier supercomputer.

ORNL's partners successful the exascale effort, HPE and AMD, person delivered the caller Frontier strategy to ORNL up of the docket for this fal. The installation and integration of Frontier, a massive, analyzable effort, is present underway, and the existent advancement indicates everything is connected way to person Frontier disposable to users for unfastened subject adjacent twelvemonth — arsenic anticipated.

Mike Bernhardt (Communication Lead for DOE's Exascale Computing Project) via InsideHPC

The placement of Bernhardt stating "complex effort" could pb to wherefore rumors abound concerning the project. It is besides to enactment that AMD's MI250X compute GPUs are lone disposable to prime customers, which is wherefore determination is simply a deficiency of benchmarks to backmost the rumored claims. The DOE has worked intimately with Oak Ridge's Leadership Computing Facility connected Frontier. The ORNL Frontier supercomputer is slated to go afloat operational by January 1, 2023, aft missing an archetypal 2022 deadline.

News Sources: Inside HPC, Toms Hardware

Read Entire Article