Molecular dynamics large scale computing grids free for all (except for the power bill)

For those who don’t know, there are now two molecular dynamics simulations grids available for everyone to participate. Both programs install with little effort. This may give you the warm and fuzzy feeling of contributing to science AND global warming at the same time.

The more famous project is Folding@Home ( http://folding.stanford.edu ), which has only this week begun to distribute larger molecules to the GPUs that are scientifically relevant. They offer a Windows software that is available in 32 bit versions only (also working fine on 64 Bit Windows though). They do have a nice viewer that can display the molecule in 3D - the viewer is a bit buggy, but they are working out the kinks as we speak.

The other project is GPU Grid ( http://www.gpugrid.net ) which is based on a prerelease version of the BOINC client that detects the GPU as a coprocessor. They appear to be distributing test work units at the moment, as the Windows version of their client has only just recently become operational. You should download BOINC 6.3.10 alpha to get started - it is available for Windows and Linux likewise (32 and 64 bit versions). You need to attach your client to their project by manually adding this project URL when prompted for the project to join: http://www.ps3grid.net/

For starters, Folding@Home may be a bit more tricky to set up when there is more than one CUDA capable GPU in your machine. However this offering also runs on AMD’s recent GPU variants.

The BOINC client has the advantage that it offers many more interesting projects to crunch on - all with the same user interface and management console. Note that the GPU Grid project is the only BOINC project supporting CUDA capable GPUs at this time.

Folding@Home is a project based in Stanford University (California, USA), whereas GPUGrid are headquartered in Barcelona, Spain. Self presentation and public outreach seems to be managed a bit better by Stanford’s group. Stanford has had a head start, they’re crunching away at 3240 Teraflops/sec right now - led on by the GPUs and the Playstations. The GPUGrid folks do not publish any Teraflop benchmarks unfortunately.

In addition.
For those who want to do molecular dynamics simulations yourself there are:
HOOMD http://www.ameslab.gov/hoomd
and
Ascalaph Liquid GPU http://www.agilemolecule.com/Ascalaph/Packages.html

Ascalaph Liquid GPU now on www.biomolecular-modeling.com/Ascalaph/Ascalaph_Designer.html

And HOOMD has morphed into HOOMD-blue - http://codeblue.umich.edu/hoomd-blue/
There are also GPU accelerated versions of LAMMPS, AMBER, NAMD, and GROMACS (from the OpenMM group), and possibly more. NVIDIA nicely links to us all: http://www.nvidia.com/object/molecular_dynamics.html