Aurora Supercomputer With Intel Ponte Vecchio Fails To Beat All-AMD Frontier In Exascale Race

Yesterday, Top500.org released the list of the fastest supercomputers on the planet and it looks like Intel's Aurora supercomputer failed to beat AMD's Frontier, the only exaflop machine that's fully operational.

Intel's Ponte Vecchio accelerators powering the Aurora Supercomputer at the Argonne Leadership Computing Facility in Illinois, USA are now operational but the promised compute performance is far from achievable as the system is operating at a quarter of its expected compute output. Both in terms of raw performance and raw efficiency, the AMD Frontier system beats Intel's Aurora system.

Following are the top three supercomputers on the planet:

Frontier remains the No. 1 system in the TOP500. This HPE Cray EX system is the first US system with a performance exceeding one Exaflop/s. It is installed at the Oak Ridge National Laboratory (ORNL) in Tennessee, USA, where it is operated by the Department of Energy (DOE). It currently has achieved 1.194 Exaflop/s using 8,699,904 cores. The HPE Cray EX architecture combines 3rd Gen AMD EPYC CPUs optimized for HPC and AI, with AMD Instinct 250X accelerators and a Slingshot-11 interconnect.

Aurora achieved the No. 2 spot by submitting an HPL score of 585 Pflop/s measured on half of the full system. It is installed at the Argonne Leadership Computing Facility, Illinois, USA, where it is also operated by the Department of Energy (DOE). This new Intel system is based on HPE Cray EX - Intel Exascale Compute Blades. It uses Intel Xeon CPU Max Series processors, Intel Data Center GPU Max Series accelerators, and a Slingshot-11 interconnect.

Eagle the new No. 3 system is installed by Microsoft in its Azure cloud. This Microsoft NDv5 system is based on Xeon Platinum 8480C processors and NVIDIA H100 accelerators and achieved an HPL score of 561 Pflop/s.

So starting with the details, the Aurora Supercomputer makes use of a mix of Intel Xeon Max "4th Gen Sapphire Rapids" and Data Center Max GPUs "Ponte Vecchio". The system features a total of 4.742 Million cores and that's way less than the final figures that were expected in the system. The system outputted a peak FP64 compute performance of 585.34 PFLOPs which is less than half the peak performance (Rmax) of the AMD-based Frontier which is the only exaflop-capable machine as it hits 1,194 PFLOPs. The system has been operational since 2022.

In terms of efficiency, the Aurora supercomputer peaks out with a Rpeak of 1059.33 PFLOP/s while the Frontier supercomputer peaks out at 1,679.82 PFLOP/s. The Aurora supercomputer has a peak power draw of 24.6 MW but that is the figure for the entire system & not the current configuration which is running with half of the specs/hardware. The Frontier supercomputer has a power draw of 22.7 MW.

Rpeak values are calculated using the advertised clock rate of the CPU. For the efficiency of the systems you should take into account the Turbo CPU clock rate where it applies.

via Top500

There's no surprise that Aurora is in the state in which it is right now. Although the system is finally operational, it is nowhere close to the 2 Exaflops that Intel had promised. This current situation has been partly due to the major delays with Intel's Sapphire Rapids and Ponte Vecchio chips which led AMD to take the first spot in the exascale race and retain it for a whole year. Intel Aurora was first unveiled all the way back in 2019 which was four years ago. But if you want to go really back, then the original design for the Aurora was laid out all the way back in 2015 with a scheduled arrival in 2018. Back then, it was a 180 PFLOPs system and based on the now discontinued Xeon Phi "Knights Hill" platform.

It is not known when we'll see Aurora finally dish out its 2 Exaflops of compute but AMD is already planning a 2 Exaflop+ system known as El-Capitan which should also become operational in the coming year. That will be a further dent to Intel in the HPC and supercomputing space.

Intel did announce two additional supercomputers, Dawn (University of Cambridge, UK) and SuperMuc-NG (LRZ, Germany). The company also shared some performance numbers of Aurora nodes versus Frontier and Polaris but I believe those don't matter much when Frontier leads the overall performance figures on the Top500 list.

AMD already leads by powering over 140 supercomputers across the globe and the list will continue to expand with its strong EPYC and the forthcoming Instinct offerings which look mighty powerful.