The jist of the claim made in the headline seems to come down to this passage:
> Overall, McIntosh-Smith says anything that is memory bandwidth dominated does well on ThunderX2 and at worst, is roughly similar to Skylake. However for more floating point-heavy applications, Skylake and Broadwell do better and are evenly matched because of the wider vectors, even though ThunderX2 cores strive to make up the difference.
It doesn't seem clear that they were making an apples-to-apples comparison to the Intel chips. I dunno.
Honestly I was hoping for more about just why ARM might be a good core for HPC. I'm slightly surprised that Performance/Watt were not reported, as I thought that's where ARM stood a fighting chance against chips like x86 and POWER.
From the lead on this project: "Not yet, as we've only had access to hardware for a few weeks! Our early access whiteboxes are very different from the production XC50 nodes we'll get next year, so we may leave energy measurements until then"
> Overall, McIntosh-Smith says anything that is memory bandwidth dominated does well on ThunderX2 and at worst, is roughly similar to Skylake. However for more floating point-heavy applications, Skylake and Broadwell do better and are evenly matched because of the wider vectors, even though ThunderX2 cores strive to make up the difference.
It doesn't seem clear that they were making an apples-to-apples comparison to the Intel chips. I dunno.
Honestly I was hoping for more about just why ARM might be a good core for HPC. I'm slightly surprised that Performance/Watt were not reported, as I thought that's where ARM stood a fighting chance against chips like x86 and POWER.