Aurora doesn't reach Exascale levels

Looks like AMD and Frontier are ahead… something of a black eye for Intel once again. of course this is a result determined by choices and actions years ago, I don’t know how it factors into the current attempt to right the ship.


It also features the AMD M1250 while the Aurora features the H100 so I wonder how Nvidia feels about this as well…doc

To be clear, for this submission they only had half of Aurora running so there should be better data for the June submission. Also, Aurora uses the Intel Ponte Vecchio GPU data accelerator, not the Nvidia H100.

The bigger concern is the Rmax was about 50% of Rpeak, while for typical top500 systems this should be between 70% and 80%. This indicates there is still a lot of work for the Aurora team to get done.