Chipzilla told the world+dog at the ISC 2024 supercomputing conference that it has bagged the world’s top performance for AI at 10.61 “AI exaflops.”

Intel’s been shouting about it on their blog, saying Aurora’s now officially the fastest supercomputer for AI on the planet. It’s a joint effort with Argonne National Laboratory and Hewlett Packard Enterprise (HPE), who both built and house the system. Intel says it was at eighty-seven per cent functionality for the recent tests.

In the all-important Linpack (HPL) test, the Aurora computer hit 1.012 exaflops, meaning it has nearly doubled the performance since its initial “partial run” in late 2023, where it only managed 585.34 petaflops. The company had said it expected to cross the exascale barrier with Aurora eventually, and now it’s gone and done it.

For the ISC 2024 tests, Intel says Aurora was running with 9,234 nodes. It came second overall in LINPACK, so it has still not managed to knock AMD’s Frontier system off the top spot, which is an HPE supercomputer.

AMD’s Frontier was the first supercomputer to break the exascale barrier in June 2022. Frontier’s sitting pretty at around 1.2 exaflops in Linpack, so Intel’s hot on its heels but still has a bit of a way to go before it can take the crown. However, Intel says Aurora came in first in the Linpack-mixed benchmark, which shows off its unmatched AI performance.

Intel’s Aurora supercomputer uses the company’s latest CPU and GPU hardware, with 21,248 Sapphire Rapids Xeon CPUs and 63,744 Ponte Vecchio GPUs. When it is fully up and running later this year, Intel reckons the system will be capable of crossing the 2-exaflop barrier