COMPUTE ATLAS Supercomputer supply-chain graph
1540 systems 545 sites Sourced data

Parts/Accelerators/NVIDIA A100 SXM4 64GB

Accelerators · Ampere · 2020

NVIDIA A100 SXM4 64GB

Custom 64 GB HBM2e SXM variant built for the Leonardo procurement.

Why this part

A capacity variant used in Leonardo. Memory per accelerator is frequently the binding constraint rather than arithmetic: a model or a domain that does not fit has to be split across devices, and every split costs communication. Buying the larger memory version of the same chip is one of the least glamorous and most consequential decisions in a procurement.

Systems using this part

Systems
2
Rmax underneath
244.7 PFlop/s
Units recorded
13,824
1 of 2 edges disclose a quantity
System Role Country Units System Rmax Supplied by Rank
Leonardo
2022-11
Accelerator Italy 13,824 241.2 PFlop/s NVIDIA #18 →
THX.A.B
2022-06
Accelerator France - 3.50 PFlop/s NVIDIA #5 →

“Units recorded” sums only the edges where a public source states a quantity. It is a floor, not a total, and should never be read as installed base.

Further reading