Easy Guide to Multi-GPU

1. Download and install the latest nvidia drivers for your system.
   http://www.nvidia.com/Download/index.aspx?lang=en-us
2. Download and install the Windows GPU systray client.
   http://www.stanford.edu/group/pandegroup/folding/release/[email protected]
   If you have a Fermi Card use this client instead.
3. Download and install my Multi-GPU Installer for your version of windows.
   [H] Multi-GPU Installer x64
   [H] Multi-GPU Installer x86
4. Now either attach a monitor or Dummy Plug to each card and extend the desktop to each one.
5. Open your newly created C:\Folding\ folder and run each shortcut one by one depending on how many GPUs you have, and setup the client for each GPU according to the screenshots and directions below.
   
   
   
   
   
   In this panel, if you want to fold under your own username, enter it here along with the team number if you are part of the team. If all you want is to contribute while not caring about a username or team, leave them as is. If you have a passkey associated to your username, enter it (if you want to know what is a passkey and how to obtain one, check this page). When you are done, click on the Connection tab on the top to get to the next panel :
   
   
   
   
   
   The first checkbox should be checked if you want the client to ask before downloading or uploading a work unit. This is useful if you are on a non-permanent connection like dialup or to control when to download/upload. If you are connected behind a proxy, you'll need to configure the client for it (if you need the information about configuring a proxy, ask your administrator). The last checkbox will let you decide if you would like to receive large work units (they have bigger bandwidth and memory requirements). After doing this, click on the Advanced tab to get to the next panel :
   
   
   
   
   
   These are the advanced configuration of the client so if you are unsure about them, you can leave them alone since the defaults are suitable for the majority of users. Here is what each element does :
   
   • Core priority : If you are running multiple clients or have another program running continuously in the background, you may need to adjust the priority so it won't sit idle while waiting for free cycles.
   • CPU Usage Percent : If you don't want the CPU to allocate all the cycles to F@H, adjust the percentage to suit your needs.
   • Disable highly optimized assembly code : If you are having problems running the client due to a processor incompatibility or any weird error, check this to disable the SSE optimizations. Doing this will cause the client to run slower depending on the workunit.
   • Checkpointing frequency : Increase or decrease the frequency of Work Unit progress backups. This won't prevent the client from writing the data before the checkpoint if the next frame is completed before.
   • Do NOT lock cores to specific CPU : If you have more than one core, checking this will allow the client to use any free CPU/core. If it is left unchecked, it will only use one core no matter if others is free or not.
   • Core Network Address (Advanced) : For future use
   • Pause work while battery power is being used : This is useful for laptops to stop the client from running when it's on battery power, to avoid draining power.
   • Additional Client Parameters : This is where you can save the frequently used flags like -verbosity 9 or -advmethods.
   • Machine ID : This is the client ID and it needs to be different if you are running more than one client under the same OS. An example would be if you are running the Systray client and the GPU2 client at the same time, in which case each client will use Machine ID 1 and 2 respectively.
   
   When you are done, click on Ok. If you entered any Additional Client Parameters during configuration, you may need to restart the client to use them. It will now run in the background with a icon in the Taskbar.

Click to expand...