The 6th HPCG Performance List was presented at SC16 on Tuesday,
November 15 at 5:15 pm, during the TOP500 BOF and on Wednesday,
November 16 at 5:15 pm (room 355-E) during the HPCG BOF,
where additional details from vendors was presented.
The High Performance Conjugate Gradients (HPCG) Benchmark list was
announced at SC16. This is the sixth list produced for a benchmark
designed to complement the traditional
High Performance LINPACK
(HPL) benchmark used as the official metric for ranking the
TOP500
systems. The first HPCG list was announced at ISC 2014 two and a half
years ago, containing only 15 entries, the SC15 list had 60 and
ISC’16 had 80. The current list contains more than 100 entries
as HPCG continues to gain traction in the HPC community.
The SC16 list contains entries for more than half of the top 50
systems from TOP500, but exhibits a significant shuffling of the HPL
rankings, indicating that HPCG features are exposing different and
complementary system characteristics.
The announcement was made during a session entitled
HPCG Benchmark Update
took place on Wednesday, November 16, 2016 in room 355-E
in Salt Lake Palace Convention Center.
The full list of November 2016 results
is now available. Final set of slides presented at the session
is also available.
Highlights from the 6th HPCG Performance List:
- HPCG list of supercomputing sites now features over 100 entries,
most of them from the very top of the TOP500 list.
- More effective optimization of HPCG to K Computer put it at the top
of the list.
- Strong showing from Japan with number 1 and number 3 entries.
- HPCG shows significant difference in rankings compared to HPL,
especially for top systems.
- Vendor optimizations illustrate adaptability of HPCG and they
inform similar adaptations for important applications.
BOF Presenters:
- Mike Heroux, HPCG Update: SC'16
- Massimiliano Fatica, Update on the CUDA Implementation of the HPCG Benchmark
- Jongsoo Park, HPCG on Intel Xeon Phi 2nd Generation, Knights Landing
- Kiyoshi Kumahata, HPCG Performance Improvement on the K computer ~short introduction~
- Bob Ciotti, HPCG Pleiades
- Peter Kogge, HPCG Architectures and Performance
