Python-based Hierarchical ENvironment for Integrated Xtallography

Hardware requirements for Phenix

Important: please check the requirements for other software that you plan to use alongside Phenix. As far as we know these are very similar, except for the independent choice of graphics card/chipset (see below). The choice will also be limited by the work environment and system management, which may dictate the choice of platform and/or machine type.

In general, Phenix can be run on any Intel-compatible hardware capable of running the supported operating systems. The minimum recommended RAM is 2 GB per processor core, which means at least 4 GB for most current processors. Most business-class laptops or desktops sold within the last three years should be suitable for solving the average macromolecular structure, although we recommend upgrading the memory from the default configuration. We do not recommend buying or using netbooks or systems with Intel Celeron processors for running Phenix, but any current Macintosh computer or PC with equivalent specifications is suitable, although smaller screens such as the 11-inch MacBook Air may be cramped for graphical use. Professional-grade systems up to five years old may also be adequate.

We have not extensively tested different processors; however, our experience is that AMD processors are slower but cheaper than Intel processors, and highly parallel systems can be bought relatively cheaply (less than $20,000 for a 64-core system as of mid-2012). For our own compute nodes, where we need to run large numbers of jobs, we generally buy these systems.

Several programs in Phenix are both particularly processor-intensive and highly parallel with no additional configuration, and may benefit from more powerful systems:

AutoBuild, MRage, and MR-Rosetta are also capable of using queuing systems such as Sun Grid Engine or PBS. In addition, Phaser (and by proxy AutoMR) can be compiled with OpenMP support, which can theoretically use any number of cores (although the benefit drops steeply above 4). However, we do not currently distribute binary installers with OpenMP enabled, and it does not work with the Phenix GUI.

Except for REEL and optional components of the GUI, none of the programs in Phenix support acceleration using graphics processing units (GPUs) via CUDA or OpenCL, due to the complexity of crystallographic algorithms. However, third-party software such as Coot and PyMOL will usually dictate the choice of graphics card.