Anton, a special-purpose machine for molecular dynamics simulation

Abstract

The ability to perform long, accurate molecular dynamics (MD) simulations involving proteins and other biological macromolecules could in principle provide answers to some of the most important currently outstanding questions in the fields of biology, chemistry and medicine. A wide range of biologically interesting phenomena, however, occur over time scales on the order of a millisecond-about three orders of magnitude beyond the duration of the longest current MD simulations.In this paper, we describe a massively parallel machine called Anton, which should be capable of executing millisecondscale classical MD simulations of such biomolecular systems. The machine, which is scheduled for completion by the end of 2008, is based on 512 identical MD-specific ASICs that interact in a tightly coupled manner using a specialized high-speed communication network. Anton has been designed to use both novel parallel algorithms and special-purpose logic to dramatically accelerate those calculations that dominate the time required for a typical MD simulation. The remainder of the simulation algorithm is executed by a programmable portion of each chip that achieves a substantial degree of parallelism while preserving the flexibility necessary to accommodate anticipated advances in physical models and simulation methods.

Is there an en route folding intermediate for cold shock proteins?

“…However, the folding time of Csp is on the order of several milliseconds, 3 which is far beyond the time scale that most simulations using all-atom force fields today can reach. Although significant progress has been made in designing specific platforms that can accelerate molecular dynamics by two or three orders of magnitude and run milliseconds simulations with explicit solvents, 29 more powerful systems are highly desired as many independent trajectories are needed to gain good statistics on folding pathways. Other than the lack of powerful computational resources, the existing force fields might not be accurate enough for ab initio folding simulations, especially for b-proteins.…”

Compressing molecular dynamics trajectories: Breaking the one-bit-per-sample barrier

“…For a state‐of‐the‐art MD application the integration of HRTC in its inner loop poses additional challenges. High performance MD simulators rely on specialized hardware—from GPUs to custom ASICs . The performance characteristic of these platforms differs from a typical desktop CPU.…”

Reactive Molecular Dynamics: Numerical Methods and Algorithmic Techniques

“…While this simplified model reduces the number of bonded interactions in GROMACS by about a factor of four, it has more significant impact on the number of non-bonded pairs, decreasing the number of such pairs by more than an order of magnitude. Considering that non-bonded force computations takes about 90% of the total time in a typical classical MD simulations [25] and that most of the allocated space is used for storing non-bonded pairs, the run-time per time-step and memory footprint of GROMACS simulations would approach those of sPuReMD, had H atoms been modeled explicitly. There is the additional burden of computing bonds, 3-body, and 4-body structures dynamically and equilibrating charges at every step of a ReaxFF simulation, and this explains the larger running time per time-step and memory footprint figures for sPuReMD.…”