Anton is a massively parallel supercomputer designed and built by D. E. Shaw Research in New York, first running in 2008. It is a special-purpose system for molecular dynamics (MD) simulations of proteins and other biological macromolecules. An Anton machine consists of a substantial number of application-specific integrated circuits (ASICs), interconnected by a specialized high-speed, three-dimensional torus network.[1]
Unlike earlier special-purpose systems for MD simulations, such as MDGRAPE-3 developed by RIKEN in Japan, Anton runs its computations entirely on specialized ASICs, instead of dividing the computation between specialized ASICs and general-purpose host processors.
Each Anton ASIC contains two computational subsystems. Most of the calculation of electrostatic and van der Waals forces is performed by the high-throughput interaction subsystem (HTIS).[2] This subsystem contains 32 deeply pipelined modules running at 800 MHz arranged much like a systolic array. The remaining calculations, including the bond forces and the fast Fourier transforms (used for long-range electrostatics), are performed by the flexible subsystem. This subsystem contains four general-purpose Tensilica cores (each with cache and scratchpad memory) and eight specialized but programmable SIMD cores called geometry cores. The flexible subsystem runs at 400 MHz.[3]
Anton's network is a 3D torus and thus each chip has 6 inter-node links with a total in+out bandwidth of 607.2 Gbit/s. An inter-node link is composed of two equal one-way links (one traveling in each direction), with each one-way link having 50.6 Gbit/s of bandwidth. Each one-way link is composed of 11 lanes, where a lane is a differential pair of wires signaling at 4.6 Gbit/s. The per-hop latency in Anton's network is 50 ns. Each ASIC is also attached to its own DRAM bank, enabling large simulations.[4]
The performance of a 512-node Anton machine is over 17,000 nanoseconds of simulated time per day for a protein-water system consisting of 23,558 atoms.[5] In comparison, MD codes running on general-purpose parallel computers with hundreds or thousands of processor cores achieve simulation rates of up to a few hundred nanoseconds per day on the same chemical system. The first 512-node Anton machine became operational in October 2008.[6] The multiple petaFLOP,[7] distributed-computing project Folding@home has achieved similar aggregate ensemble simulation timescales, comparable to the total time of a single continuous simulation on Anton, specifically achieving the 1.5-millisecond range in January 2010.[8]
The Anton supercomputer is named after Anton van Leeuwenhoek,[9] who is often referred to as "the father of microscopy" because he built high-precision optical instruments and used them to visualize a wide variety of organisms and cell types for the first time.
The ANTON 2 machine with four 512 nodes and its substantially increased speed and problem size has been described.[10]
The National Institutes of Health have supported an ANTON for the biomedical research community at the Pittsburgh Supercomputing Center, Carnegie Mellon University, and currently (8/20) continues with an ANTON 2 system.
{{cite book}}
: |journal=
ignored (help) (Related paper published in Proceedings of the 34th Annual International Symposium on Computer Architecture (ISCA '07), San Diego, California, June 9–13, 2007).
{{cite book}}
: |journal=
ignored (help)
{{cite book}}
: |journal=
ignored (help)