view article
Figure 6
(a) Aggregated memory bandwidth and (b) single-precision computation performance for the first node of a hybrid run using ten cluster nodes with 32 threads each. The stream benchmark memory bandwidth result for 32 threads has been added as a reference (horizontal blue line). Different colours are used for each thread. Only one thread per socket reports on the memory bandwidth since these data are based on uncore events.

Journal logoJOURNAL OF
APPLIED
CRYSTALLOGRAPHY
ISSN: 1600-5767
Volume 53| Part 2| April 2020| Pages 574-586
https://doi.org/10.1107/S1600576720001776
Follow J. Appl. Cryst.
Sign up for e-alerts
E-alerts
Follow J. Appl. Cryst. on Twitter
Twitter
Follow us on facebook
Facebook
Sign up for RSS feeds
RSS