I tried a couple of the GPU clients (the graphical ones that goes in the "sys tray", showing Molecule icon by right-hand corner comp clock). By " command like edition" you mean the one that pops up in the DOS Window and have to type in commands and click Enter? I guess it makes sense that the it would be more efficient that way, b/c not slowed down by 3D rotating protein.
The folding core is the same for the graphical and console clients. If you keep the "Graphical one" minimized it requires very little additional resources (probably a little bit of RAM) . But if resources are critical, the "Console" client is the one to use.
In the process of using the Console client, one can acquire insight into the folding software structure that would unlikely be acquired through use of the Graphical client. This insight allows more versatility in using the clients.
...so do you run at least 1 GPU + 1 CPU simultaneously?
On systems that are capable, such as yours, GPU plus CPU (SMP client) usually yields the highest PPD.
Is the CPU Client you linked much better or can I run regular old one (1st one of the http://folding.stanf...nglish/Download page)?
Chungenhung provided a link to the DEINO SMP client.
You can run the older SMP client (the MPICH client) - it still seems to be a commonly used client for Windows.
The Windows SMP MPICH client did (in the past, maybe still) have the unfortunate property that if the LAN (network) connection was lost while folding, then the SMP client crashed. There is a workaround for this if anyone notices the problem.
If the machine's Quad Core, it means max. 4 clients right? so Running GPU + CPU 'll use little of resources
A quad core CPU can run four regular standard uniprocessor clients (usually the first client offered on the
Stanford download page).
For the highest productivity (most PPD), it is usually best not to run the uniprocessor client, instead run the SMP client as you mentioned previously (either DEINO or MPICH).
For your Nvidia GPU, it may be best to use the latest driver, it allows the GPU to be supported with very little CPU utilization. This driver frees up the CPU cores for use in folding SMP work units.
For WindowsXP download
CUDA 2.1 beta (180.60).
If you have Vista choose the Vista version.
Uninstall the old NVIDA driver and install the new one (follow special instructions if any).
Install the driver before starting the client, or stop the client then install and restart the GPU client.
Edited by dnamechanic, 04 March 2009 - 05:25 PM.