Cores

Introduction
The 'core' is the program that performs the calculations. The same core is used by various versions of the client and is automatically updated whenever necessary. This provides an easy way for the scientific calculations to be improved without requiring you to install a new version of the client.

There are several different cores:

Note: The numbers associated with the names of the cores are arbitrary identifiers. They do not indicate a version number which can be used to tell an "old" core from a "new" one. Version 1.86 of FahCore_78 may or may not be older than Version 2.53 of FahCore_65. See also: Core MD5 Sums

Gromacs core
Gromacs (Groningen Machine for Chemical Simulations) is a single-precision scientific core that is used for the majority of F@H calculations. It is developed by the University of Groningen under the GPL and is writen in Fortran77-based GROMOS in C. Gromacs takes advantage of SIMD, including SSE and AltiVec instructions in supported processors to provide nearly a threefold increase in calculation speed. Modern processors such as the Pentium 3, Pentium 4, Athlon XP and Athlon 64 series, as well as their mobile counterparts, are SSE capable. The PowerPC G4 and G5 series are Altivec capable.

More information can be found here: http://folding.stanford.edu/gromacs.html Gromacs Home page

Double Gromacs core
The DoubleGromacs core is very similiar to the Gromacs core except it performs double-precision rather than single-precision calculations. DoubleGromacs takes advantage of SSE2 instructions in supported processors to provide about a 2x increase in calculation speed. Newer processors such as the Pentium 4 and the Athlon 64 have SSE2 capability.

GB Gromacs core
The GB Gromacs core is very similiar to the Gromacs core. The Pande Group added the &quot;Generalized Born implicit solvent&quot; model to GB Gromacs codebase; this should make it possible to drop or reduce the use of Tinker in the future.

A new version of the a4 fahcore (v2.25), based on Gromacs v4.5.1, was released in October 2010. Folding Forum: P10412 now on advanced

Double Gromacs B core
The Double Gromacs B core is very similiar to the Double Gromacs core. This new FahCore_7b has several scientific additions to the core. Initially released only to the Linux platform in August 2007, it will eventually be available for all platforms.

Double Gromacs C core
The Double Gromacs C core is very similiar to the Double Gromacs core. Initially released for Linux and Windows in April 2008.

Gromacs 33 core
The Gromacs 33 core contains features from more recent versions of Gromacs which have been ported to Folding@home, allowing a broader range of simulations to be run.

Gromacs SMP core
The Gromacs SMP core supports running on multiprocessor or multicore processor systems. This core is currently available for Mac OS X on Intel, 64-bit Linux and Windows. 32-bit Linux support is being worked on.

This core makes running multiple clients for each CPU or CPU-core obsolete. The core will make use of the two or more cores available in the system.

See: and Folding@home on multi-core/SMP boxes (SMP FAQ)

Gromacs CVS SMP core
The Gromacs CVS SMP core supports running on multiprocessor or multicore processor systems. This core is currently available for only Mac OS X on Intel and 64-bit Linux. A Windows version, and a 32-bit Linux support is being worked on.

This core makes running multiple clients for each CPU or CPU-core obsolete. The core will make use of the two or more cores available in the system.

See: and Folding@home on multi-core/SMP boxes (SMP FAQ)

This core looks almost identical to the Gromacs SMP core (a1) because it uses mostly the same code base from Gromacs. The Gromacs CVS SMP core (a2) only uses more recent code from Gromacs than the Gromacs SMP core (a1) core does, and supports some more features (like more CPUs) than the Gromacs SMP core (a1) does at current.

Specifying the -bigadv flag requests Really BIG Work Units (WU) using 8+ fast cores, 6+ GB RAM, tight deadlines and more network bandwidth. See the announcement Extra-large work units (bigadv) and the New bonus plan TRIAL (bigadv)

Gromacs SMP2 core
First SMP2 core to use a threads-based parallelization instead of MPI for its Inter-process communication

See: Folding Forum: SMP2 with passkey (core_a3) V6.29Beta Folding@home Blog: SMP2 core update

As of approximately 3/3/2011, this core a3 no longer runs bigadv WU's, allowing optimizations for non-bigadv SMP WU's. See the new Gromacs SMP2 bigadv core for running bigadv WU's.

Gromacs SMP2 bigadv core
Derivative SMP2 core using a threads-based parallelization instead of MPI for its Inter-process communication optimized for bigadv WU's.

See: Folding Forum: upcoming core changes (bigadv will switch from A3 to A5) Folding@home Blog: Introduction of a new SMP core, changes to bigadv

Specifying the -bigadv flag requests Really BIG Work Units (WU) using 8+ fast cores, 6+ GB RAM, tight deadlines and more network bandwidth. See the announcement Extra-large work units (bigadv) and the New bonus plan TRIAL (bigadv)

SHARPEN core
See: Folding@Home #N FAQ

Tinker core
The Tinker core was the second-most used core for F@H calculations (superseded by AMBER). It does not use any special instructions to accelerate floating-point operations on modern processors (such as SSE or 3dNow+). WorkUnits that use the Tinker core usually perform well on AMD processors.

QMD core
QMD uses a quantum chemical method for calculations and uses a large amount of memory (sometimes in excess of 512 MiB of RAM).

Note: Because of the licensing issues (FAH & QMD & AMD64 & SSE2) QMD WUs are currently not assigned to AMD CPUs.

Note: The QMD core is multi-thread capable.

There are no projects at the moment utilizing the QMD core:

But it may become active again, see

More information can be found here: http://folding.stanford.edu/QMD.html

AMBER core
The AMBER core (Assisted Model Building and Energy Refinement) is a molecular dynamics program originally developed by Peter Kollman's group at the University of California, San Francisco. Its main use is in force field calculations. The project is now coordinated by David A. Case at Scripps Research Institute. The AMBER core does not take advantage of SSE to provide an increase in calculation speed, but hopefully will use SSE2 in the future.

There is a project to integrate at least a portion of the AMBER force field algorithm into the Gromacs core (ffamber). It is not clear if this will negate the use of the AMBER core. http://folding.stanford.edu/ffamber/

More information can be found here: http://folding.stanford.edu/AMBER.html http://en.wikipedia.org/wiki/AMBER

GPU core
The GPU core (Graphics Processor Unit) uses the graphics chip of modern video cards to do molecular dynamics. The first GPU client supports ATI 1xxx series of GPUs. "The GPU Gromacs core isn't a port of Gromacs, but rather we've taken key elements from Gromacs we need and enhanced them based on the unique capabities of GPU's. Thus, it's really a new and different core on the inside, but with wrapped Gromacs on the outside" -- Vijay Pande, Getting close to the GPU beta launch

GPU2 core
The GPU core (Graphics Processor Unit) uses the graphics chip of modern video cards to do molecular dynamics. This second generation GPU client supports ATI 2xxx/3xxx or later, and nVidia 8xxx or later series of GPUs. This new client uses AMD's CAL or nVidia's CUDA instead of Microsoft's DX9.

GPU3 core
The GPU core (Graphics Processor Unit) uses the graphics chip of modern video cards to do molecular dynamics. This third generation GPU client/core was mentioned in the project NEWS. More details to follow.

FAHCore_17 FAH News Blog posts: http://folding.typepad.com/news/2013/05/core-17-update-video-and-reddit-live-qa.html http://folding.typepad.com/news/2013/06/welcome-to-fahcore-17.html http://folding.typepad.com/news/2013/07/a-peek-into-core-17-benchmarking.html

Gromacs SREM core
The Gromacs Serial Replica Exchange Method core, also known as GroST (Gromacs Serial replica exchange with Temperatures), uses Replica Exchange Method in its simulations also known as REMD, Replica Exchange Molecular Dynamics.

See:

See also: "Replica-exchange molecular dynamics method for protein folding"

Gromacs Simulated Tempering core
GROSimT is a new (May 2007) Gromacs core that performs simulated tempering. Simulated tempering is one of the sampling algorithms, of which the basic idea is to enhance sampling by periodically raising and lowering temperature. Using this new core, we are hoping to sample more efficiently the transitions between folded and unfolded conformations of proteins.

See also: Simulated tempering: a new Monte Carlo scheme. Authors:, Marinari, E.; Parisi, G. Publication:, Europhysics Letters, Vol. 19, p.451.

ProtoMol core
We have been working on another approach to speeding dynamics greatly, based on a new technique called Normal Mode Langevin (NML) dynamics. This method uses the same style models as normal MD (same force fields, etc) and thus should have the same accuracy, but with a pretty significant speedup due to algorithmic advances. NML is complementary to our other methods, so we're hoping to add it to everything else (in particular to the GPU core). To start, we will be testing it in a new core, based on the Protomol software. Protomol is designed to allow for rapid prototyping of molecular simulations, which is perfect for NML.

See http://protomol.sourceforge.net/

Desmond core
Place holder for a new fahcore mentioned on the project News page here. And a link to the developer D E Shaw Research

Links

 * Folding@home FAQ, you can learn about the cores
 * Folding@home Settings FAQ
 * QMD uses CPMD, this is its home page
 * GROMACS home page
 * Tinker home page
 * Downloading FAH Core files manually