Exascale Computing and Energy-Efficient Supercomputing

Leggi questo articolo ad alta voce

Attiva il microfono e leggi con il tuo ritmo. Le parole si illuminano mentre procedi, e nulla ti ferma quando una viene pronunciata male.

Il tuo audio non arriva mai a MovaReader: a riconoscerlo è il tuo browser. Informativa sulla privacy

Accedi per conservare questa sessione nella cronologia delle letture.

Exascale Computing and Energy-Efficient Supercomputing

Noor had not come for a tour of the machine room. A storm in her model was still a smear: the cloud was not a cloud, only a number tuned so the rest of the atmosphere would behave. She wanted a grid fine enough that the cloud would have edges. The class of machine that can attempt that is called exascale, able to carry out at least one quintillion floating-point operations in a second. Frontier, at Oak Ridge National Laboratory, was the first of them to appear on the public lists past that line. The system under this floor was built on the same bargain.

Ellis, who held the schedule, did not quarrel with the cloud. He quarreled with the socket. The useful arithmetic on these machines sits mostly on accelerators, large fields of smaller processors that earn their speed by repeating one kind of sum.