Intel’s Aurora Supercomputer breaks the ExaFLOP barrier, but this isn’t good news

AMD’s Frontier supercomputer is making a mockery of Intel’s Aurora Supercomputer

Intel have announced that their Aurora Supercomputer has broken the exascale barrier to become the “fastest AI system in the world dedicated to AI for open science”. While this may sound like good news for Intel, I can assure you it isn’t. To prove this, all we need to do is loo at the Top 500 list of Supercomputers.

Note that Intel has not created the “Fastest Supercomputer”. That title belongs to the AMD-powered Frontier system at Oak Ridge National Laboratory. While Intel’s Aurora Supercomputer has not been fully commissioned and could become faster, it is unlikely to surpass Frontier. Beyond that, Aurora has been delivered behind schedule and consumes a lot more power than Frontier.

Currently, the Intel Aurora Supercomputer can generate 1.012 exaflops of performance using 87% of its node capacity. When 100% of this system is operational, the system should generate an Rmax value of around 1.163 exaflops. This will place Intel’s system behind Frontier, though system optimisations may boost this value significantly.

At ISC High Performance 2024, Intel announced in collaboration with Argonne National Laboratory and Hewlett Packard Enterprise (HPE) that the Aurora supercomputer has broken the exascale barrier at 1.012 exaflops and is the fastest AI system in the world dedicated to AI for open science, achieving 10.6 AI exaflops. Intel will also detail the crucial role of open ecosystems in driving AI-accelerated high -performance computing (HPC).

(Image from Top500)

The Intel-based Aurora supercompuer is located in the Argonne National Laboratory and was originally due to pass the exaflop barrier in 2021. The system contains 21,248 Intel Xeon Max series CPUs and 63,744 GPU Max series (Ponte Vecchio) graphics accelerators.

Intel’s claim to fame with Aurora is its AI performance capabilities. The Xe-based GPUs inside Aurora feature AI hardware blocks, giving the system 10.6 AI exaflops of computational performance. This makes Aurora the “fastest AI system in the world dedicated to AI for open science”.

Overall, the Aurora project has been an embarrassing one for Intel. While they have succeeded in creating an exascale supercomputer, they have done so behind AMD and far behind schedule. AMD’s Frontier system remains the world’s most powerful supercomputer, and it is clear that AMD can now do much better with their newer EPYC processors and AMD Instinct accelerators. That said, Intel could also do better if they used their latest kit.

You can join the discussion on Intel’s Aurora supercomputer on the OC3D Forums.