Quantum Chemistry on Graphical Processing Units. 3. Analytical Energy Gradients, Geometry Optimization, and First...

Area of Science:

• Computational Chemistry
• High-Performance Computing
• Quantum Chemistry

Background:

• Traditional high-performance computing for quantum chemistry relies on numerous CPU cores.
• Large-scale Hartree-Fock and Density Functional Theory calculations are computationally intensive.
• Consumer-grade hardware is typically not considered for demanding scientific simulations.

Purpose of the Study:

• To evaluate the performance of consumer-grade graphics processing units (GPUs) for quantum chemistry calculations.
• To demonstrate the feasibility of running large-scale electronic structure calculations on a single workstation.
• To assess the speedup achievable with GPU acceleration in ab initio molecular dynamics.

Main Methods:

• Utilized a video gaming machine equipped with two consumer graphics cards.
• Performed Hartree-Fock energy and gradient calculations.
• Conducted benchmark Born-Oppenheimer molecular dynamics simulations on a hydronium ion and an aspartic acid molecule.

Main Results:

• Achieved a performance increase of over 180× compared to a state-of-the-art quad-core processor workstation for Hartree-Fock calculations.
• Demonstrated the capability to perform ab initio molecular dynamics simulations at rates of 27 ps/day and 0.7 ps/day for different molecular systems.
• Showcased that complex calculations, typically needing hundreds of CPU cores, can be run on a single workstation.

Conclusions:

• Consumer graphics cards offer a significant performance advantage for quantum chemistry.
• GPU acceleration democratizes access to large-scale computational chemistry simulations.
• This approach enables efficient ab initio molecular dynamics on accessible hardware.