Convey ranked at #26 in Graph 500 results
Some exciting news came out of Germany at ISC’12, when the 2012 Graph 500 results were announced.
The results for Convey were superb, ranking #26 out of 81 entries. This is a David and Goliath story.
David v. Goliath
Placing 26th may not sound like much, but when you consider the “who’s who” that are on this list, it is quite a feat.
The top 25 entries have an average of over 8,000 nodes and over 77,000 cores. The Convey hybrid-core system ranked at #26 has 1 node with 12 cores. Convey is of course, “David” in this analogy.
Fewer nodes means lower operating costs (cooling and power) and reduced space requirements, not to mention the purchase price. In fact, many of the systems on the list are research in nature and can’t be purchased. The Convey HC-2 ex is commercially available and began shipping last month. Eurotech are proud to work with Convey, please contact us to discover how this system can work for you.
Why is the Graph 500 important for Big Data Analytics?
Data intensive supercomputer applications are increasingly important for HPC workloads, but are ill-suited for platforms designed for 3D physics simulations. The Graph500 was created because current benchmarks and metrics do NOT provide useful information on the suitability of supercomputing systems for data intensive applications. Backed by a committee of over 50 international HPC experts from academia, industry, and national laboratories, Graph 500 established a set of large-scale benchmarks for these applications. The Graph 500 steering committee is in the process of developing comprehensive benchmarks to address three application kernels: concurrent search, optimization (single source shortest path), and edge-oriented (maximal independent set).
Convey hybrid-core systems use personalities—customized instruction set architectures that increase performance of specific portions of an application, in this case the BFS algorithm. Using the new personality, the single-node Convey HC-1 ex increased performance from 1.7 GTEPS to 5.9 GTEPS (billion edges per second), an increase of 3.5x on problem size 27. On the new HC-2ex, the performance was even more dramatic, clocking in at 7.8 GTEPS for an increase of 4.5x over the earlier HC-1ex. This kind of performance in a single-node system reinforces Convey as a performance/power leader in executing graph type applications.
Convey’s latest version of the BFS is a highly threaded algorithm that utilizes over three thousand independent threads of computation. This capability allows for massive parallelism that speeds all graph type applications. In addition, the unique hybrid nature of the architecture allows portions of the benchmark to simultaneously execute across the multiple compute resources of the system.
For more information please call 01932 260470 or email email@example.com
Graph 500- http://www.graph500.org/