The High Performance Conjugate Gradients (HPCG) Benchmark list was announced at SC’15. This is the fourth list produced for the emerging 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’14 a year and a half ago, containing only 15 entries and the SC’14 list had 25. The current list contains more than 60 entries as HPCG continues to gain traction in the HPC community.
The 4th list contains entries from many of the TOP500’s highest ranked systems but exhibits a significant rearrangement of the HPL rankings, indicating that HPCG features are exposing different and complementary system characteristics.
Highlights from the 4th HPCG list:
- HPCG list of supercomputing sites now features over 60 entries, majority of them from the very top of the TOP500 list.
- The upgrade to the TSUBAME KFC super from K20x to K80 cards nearly doubled its HPCG score.
- Strong showing from Japan and NEC SX vector computers continues with new systems and over 10% of peak performance of HPCG.
- New IBM BlueGene machines continue appearing on the list.
- A number of submissions from Russia significantly increased the total count of systems from that country.
- An HP system makes a strong showing in the first half of the list with the Prometheus super from Poland.
HPCG 3.0 Announced
The third major release of HPCG is available in its reference version. Highlights are:
- More diverse computations: Problem setup time is included as part of the overhead. This addition provides some instruction and data access diversity and provides another indicator of system balance.
- Quick Path execution mode: HPCG 3.0 provides a Quick Path execution mode, available for production computing systems where large blocks of dedicated time are difficult to schedule. Quick Path mode enables execution of the entire benchmark in just a few minutes on existing production computing systems.
- Dual rating report: HPCG 3.0 reports both the 2.4 and 3.0 rating results, for easier historical transition.
- Vendor-‐optimized versions: Will be available in the near future, and announced on the HPCG Benchmark website.
The High Performance Conjugate Gradients (HPCG) Benchmark Project started over two years ago as an effort to probe important characteristics of a computer system, highlighting and rewarding investment in system features, such as high performance interconnects, memory systems, and fine grain cooperative threading, that are important to a broad set of applications. HPCG has emerged as a complementary benchmark to the traditional High Performance LINPACK (HPL) benchmark used to rank the TOP500 systems. The pair of numbers provided by HPCG and HPL act as bookends on the performance spectrum of a given system.