Guide For GPU3 BETA Client {Windows & Linux}

If you think it might be a driver problem, see viewforum.php?f=79

Moderators: slegrand, Site Moderators, PandeGroup

Guide For GPU3 BETA Client {Windows & Linux}

Postby PantherX » Tue May 25, 2010 5:27 pm

Please note that this guide is no longer being updated as V7 has superseded GPU3.

This is the index to my GPU Guide. You will find the links to the various sections of my guide. A mini description of what that section covers is provided to ease navigation. If in doubt, please read all sections or ask in this Forum.

1 - Officially Supported Operating Systems -> The GPU3 BETA Client is Officially tested on these Operating Systems. Since the Client is still in BETA Stage, you may encounter some bugs. If you come across one, please report it in this Forum.

2 - UNOFFICIALLY Supported Operating System -> The GPU3 BETA Client is NOT Officially tested on these Operating Systems. If you encounter bugs, support is provided by 3rd party NOT Standford/Pande Group.

3 - Supported Hardware -> Commonly known GPUs that are capable of running GPU3 BETA Client.

4 - Before Using GPU3 BETA Client -> Important information that you should be aware of before using the GPU3 BETA Client.

5 - GPU3 BETA Client Download Locations -> You can download the GPU3 BETA Client from here.

6 - Installation Of Systray Version -> How to install the GPU3 BETA Systray Client.

7 - Installation Of Console Version -> How to install the GPU3 BETA Console Client.

8 - Upgrading The Console Version -> How to upgrade from an existing GPU2 Console Client.

9 - Installation Of Multiple Instances -> If you have more than 1 GPU installed in the system and would like to run the GPU3 BETA Client on all of them.

10 - Upgrading From A Non-Fermi GPU To A Fermi GPU -> When you are replacing your Non-Fermi GPU with a Fermi GPU.

11 - Using A Nvidia Fermi GPU & A Nvidia Non-Fermi GPU In One System -> If you have a Nvidia Fermi GPU and a Nvidia Non-Fermi GPU in a single system and would like to fold on them.

12 - Using Multiple Nvidia Fermi GPUs In One System -> If you have more than one Fermi GPU in your system and would like to fold on them.

13 - Changing From One GPU Brand To Another -> If you are replacing your GPU from an ATI one to Nvidia one or from a Nvidia one to an ATI one.

14 - Notes Regarding GPU3 BETA Client -> Information that is applicable to GPU3 BETA Client.

15 - Troubleshooting The GPU3 BETA Client -> If your GPU3 BETA Client is giving your errors, this is a good place to start troubleshooting your GPU.

16 - Credits -> List of Users that have helped me write this GPU Guide.

17 - Changelog -> A list of recent changes that I have done. If I have made new changes, I will be deleting the last post and making a new one. Hence it will appear in the "new posts" so that users can be informed when I have made updates to the Guide.


Suggestions for GPU3 Guide For Windows & Link To Linux -> Please use this thread for posting any corrects/suggestions/comments, etc. Thanks for your understanding and cooperation.
Last edited by PantherX on Sat Jan 29, 2011 8:56 am, edited 62 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

1 - Officially Supported Operating Systems

Postby PantherX » Thu Sep 16, 2010 3:23 am

  • Officially Supported Operating Systems
Windows XP (32 bit & 64 bit)

Windows Vista (32 bit & 64 bit)

Windows 7 (32 bit & 64 bit)
Last edited by PantherX on Sun Sep 26, 2010 1:35 pm, edited 2 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

2 - UNOFFICIALLY Supported Operating System

Postby PantherX » Thu Sep 16, 2010 3:24 am

  • UNOFFICIALLY Supported Operating System
Linux (32 bit & 64 bit)

Step-by-Step Installation: <Click>

Quick Wrapper Installation: <Click>

WIKI: <Click>

GPU3 (FahCore_15) & Linux discussion Thread: <Click>

Nvidia GPU3 Linux/Wine Headless Install Guide Thread: <Click>


THIS IS UNSUPPORTED BY STANDFORD AS OFFICIALLY THERE ISN'T A LINUX GPU CLIENT YET. SUPPORT IS PROVIDED BY 3RD PARTY, NOT STANDFORD.
Last edited by PantherX on Fri Jan 28, 2011 12:25 am, edited 3 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

3 - Supported Hardware

Postby PantherX » Thu Sep 16, 2010 3:28 am

Legend:
Red Text -> Specifically For AMD/ATI GPUs
Green Text -> Specifically For Nvidia GPUs
Blue Text -> Important Note(s) That You Must Read
Black Text -> Applicable To Both GPU Vendors


  • Supported Hardware
This is a brief list of commonly used GPUs running GPU3 BETA Client:

AMD/ATI GPUs:
Radeon HD 6000 Series
Radeon HD 5000 Series (ATI_R800)
Mobility Radeon HD 5000 Series (ATI_R800)
Radeon HD 4000 Series (ATI_R700)
Mobility Radeon HD 4000 Series (ATI_R700)
Radeon HD 3000 Series (ATI_R600)
Mobility Radeon HD 3000 Series (ATI_R600)
Radeon HD 2000 Series (ATI_R520)
Mobility Radeon HD 2000 Series (ATI_R520)


Nvidia GPUs:
GeForce 500 Series (Fermi GPU)
GeForce 400 Series (Fermi GPU)
GeForce 400M Series (Fermi GPU)
GeForce 300 Series (Non-Fermi GPU)
GeForce 300M Series (Non-Fermi GPU)
GeForce 200 Series (Non-Fermi GPU)
GeForce 200M Series (Non-Fermi GPU)
GeForce 100 Series (Non-Fermi GPU)
GeForce 100M Series (Non-Fermi GPU)
GeForce 9 Series (Non-Fermi GPU)
GeForce 9M Series (Non-Fermi GPU)
GeForce 8 Series (Non-Fermi GPU) {Read Note 5}
GeForce 8M Series (Non-Fermi GPU)



The GPU3 BETA Client identifies the GPU type and reports it in the FAHlog like this "Gpu type=X species=Y." Below is the list of what your FAHlog would show depending on the GPU Type and the Driver installed:

Gpu type=1 species=1. -> "Gpu type=1" indicates an ATI GPU; "species=1" means ATI_R520.

Gpu type=1 species=2. -> "Gpu type=1" indicates an ATI GPU; "species=2" means ATI_R600.

Gpu type=1 species=3. -> "Gpu type=1" indicates an ATI GPU; "species=3" means ATI_R700.

Gpu type=1 species=4. -> "Gpu type=1" indicates an ATI GPU; "species=4" means ATI_R800.


The GPU detection messages below is specific to v6.3X of the GPU3 Beta Client:
v6.3X wrote:The GPU3 BETA Client identifies the GPU type and reports it in the FAHlog like this "Gpu type=X species=Y." For Nvidia GPUs, the "species=Y" is based on the CUDA runtime file found. Below is the list of what your FAHlog would show depending on the GPU Type and the Driver installed:

Gpu type=2 species=0. -> "Gpu type=2" indicates a Nvidia Non-Fermi GPU; "species=0" means CUDA 2.1 or earlier. It can also mean that no CUDA runtime file found.

Gpu type=2 species=22. -> "Gpu type=2" indicates a Nvidia Non-Fermi GPU; "species=22" means CUDA 2.2 is available.

Gpu type=2 species=23. -> "Gpu type=2" indicates a Nvidia Non-Fermi GPU; "species=23" means CUDA 2.3 is available.

Gpu type=2 species=30. -> "Gpu type=2" indicates a Nvidia Non-Fermi GPU; "species=30" means CUDA 3.0 is available.

Gpu type=3 species=0. -> "Gpu type=3" indicates a Nvidia Fermi GPU; "species=0" means CUDA 2.1 or earlier. It can also mean that no CUDA runtime file found.

Gpu type=3 species=22. -> "Gpu type=3" indicates a Nvidia Fermi GPU; "species=22" means CUDA 2.2 is available.

Gpu type=3 species=23. -> "Gpu type=3" indicates a Nvidia Fermi GPU; "species=23" means CUDA 2.3 is available.

Gpu type=3 species=30. -> "Gpu type=3" indicates a Nvidia Fermi GPU; "species=30" means CUDA 3.0 is available.

The GPU detection messages below is specific to v6.4X of the GPU3 Beta Client:
v6.4X wrote:The GPU3 BETA Client identifies the GPU type and reports it in the FAHlog like this "Gpu type=X species=Y." For Nvidia GPUs, the "species=Y" is based on the compute capability of the installed GPU. Below is the list of what your FAHlog would show depending on the GPU Type and the Driver installed:

Gpu type=2 species=0. -> "Gpu type=2" indicates a Non-Fermi GPU; "species=0" means unknown compute capability

Gpu type=2 species=10. -> "Gpu type=2" indicates a Non-Fermi GPU; "species=10" means compute capability 1.0 is available {Read Note 5}

Gpu type=2 species=11. -> "Gpu type=2" indicates a Non-Fermi GPU; "species=11" means compute capability 1.1 is available.

Gpu type=2 species=12. -> "Gpu type=2" indicates a Non-Fermi GPU; "species=12" means compute capability 1.2 is available.

Gpu type=2 species=13. -> "Gpu type=2" indicates a Non-Fermi GPU; "species=13" means compute capability 1.3 is available.

Gpu type=3 species=0. -> "Gpu type=3" indicates a Fermi GPU; "species=0" means unknown compute capability

Gpu type=3 species=20. -> "Gpu type=3" indicates a Fermi GPU; "species=20" means compute capability 2.0 is available.

Gpu type=3 species=21. -> "Gpu type=3" indicates a Fermi GPU; "species=21" means compute capability 2.1 is available.


Under some conditions, your GPU will not be recognized so it will display this message "Gpu species not recognized." The common reasons are:

1) You haven't installed the correct Driver. Check what Driver is recommended for Folding or try the latest ones by your GPU Vendor.

2) You may have stubborn files from the previous driver that is causing this error. Please use Driver Sweeper and try again.

3) You might be using a recently launched GPU Model that isn't recognized by the GPU3 BETA Client. If this happens, by using this method, an appropriate flag may be added which would allow your GPU to fold. The flag will only work if the GPU architecture is similar to the previous one. It is advised to search the Forum for the GPU in question to see which flag is required for it to work:

-forcegpu ati_r600 -> It bypasses the Clients' auto detect feature and instructs it to treat the GPU as a 3000 Series one

-forcegpu ati_r700 -> It bypasses the Clients' auto detect feature and instructs it to treat the GPU as a 4000 Series one

-forcegpu ati_r800 -> It bypasses the Clients' auto detect feature and instructs it to treat the GPU as a 5000 Series one


The meanings of the -forcegpu flags are specific to v6.3X of the GPU3 Beta Client:
v6.3X wrote:-forcegpu nvidia_fermi -> It bypasses the Clients' auto detect feature and instructs the Client to treat the GPU as a Fermi architecture, i.e. the GPU is a 400 Series one

-forcegpu nvidia_g80 -> It bypasses the Clients' auto detect feature and instructs the Client to treat the GPU as a G80 architecture, i.e. the GPU is a 8/9/200 Series one {Read Note 5}

The meanings of the -forcegpu flags are specific to v6.4X of the GPU3 Beta Client:
v6.4X wrote:-forcegpu nvidia_fermi -> It bypasses the Clients' auto detect feature and instructs the Client to treat the GPU having compute capability 2.0

-forcegpu nvidia_g80 -> It bypasses the Clients' auto detect feature and instructs the Client to treat the GPU having compute capability 1.1 {Read Note 5}

-forcegpu nvidia_g80_1.0 -> It bypasses the Clients' auto detect feature and instructs the Client to treat the GPU having compute capability 1.0 {Read Note 5}

4) You may be trying to run the GPU3 BETA Client on the GPU Model that is too old hence isn't supported by F@h.


Note 1: Using the -forcegpu flag with the wrong GPU will mess things up. It may invalidate the scientific results produced. DO NOT MISUSE THESE FLAGS. (Details)

Note 2: Please remember that officially, Overclocking is discouraged. (Details)

Note 3: If you are using 8800 Series GPU or 9800 series GPU, you may encounter bugs. The Development Team is aware of it and are working on a fix: <Click>

Note 4: On some GPUs, when it is folding, a high pitched whine noise is emitted from the GPU. This is called Capacitor Whine and doesn't pose any serious threat to the GPU in use.

Note 5: Please note that the following GPUs MUST NOT USE the -forcegpu nvidia_g80 since they will be assigned FahCore_15 WUs which they cannot process. This is due to the compute capability limitation hence they have to use the -forcegpu nvidia_g80_1.0 flag:
GeForce 300 Series -> GeForce GT 340, GeForce GT 330, GeForce GT 320, GeForce 315, GeForce 310
GeForce 9000 Series -> GeForce 9800 GT, GeForce 9600 GT, GeForce 9400GT
GeForce 8000 Series -> Nvidia GeForce 8800 Ultra, Nvidia GeForce 8800 GTX, Nvidia GeForce 8800 GTS
Last edited by PantherX on Fri Jan 28, 2011 12:27 am, edited 9 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

4 - Before Using GPU3 BETA Client

Postby PantherX » Thu Sep 16, 2010 3:29 am

Legend:
Blue Text -> Important Note(s) That You Must Read


  • Before Using GPU3 BETA Client
Users of GPU2 Client, wanting to check out GPU3 BETA Client, are advised to use the -oneunit flag. After the Client exits, proceed with the installation of GPU3 BETA Client. It is recommended to upgrade the Client after the WU has been completed to avoid any potential problems that you may experience from incompatibilities between the FahCore and the WU. Also it won't cause loss of points if any problems does occur that causes the WU to be lost.

Please ensure that you are meeting the minimum requirements for running the GPU3 BETA Client. Usage of Nvidia 258.96 WHQL Driver is recommended. If you use drivers prior to this, it may or may not work. Note that GPU2 Client will not use the FahCore_15 to process the WU. As the Client is in BETA Stage users can expect instability so if you have such an experience, please report it.

Some users complain that after they installed the recommended driver, they are having problem(s) when running F@h GPU3 BETA Client. Most of this can be avoided by making a clean installation of the driver. However, you can use this freeware application Driver Sweeper to remove any stubborn file(s) which may be causing problem(s).


Note 1: Please refrain form folding on non-recommended Drivers. The reason is that you may be producing scientifically invalid results which may have a negative effect on the Project. (Details)

Note 2: Please refrain form folding on BETA Drivers and reporting any issues with them. The reason is that if you are using BETA Drivers, and a problem appears, F@h developers can't differentiate if it is the Drivers fault (GPU Vendors' bug) or something in their software (F@h bug). Hence please wait for the WHQL Driver release which will be tested by the F@h developers and once it is verified to produce scientifically valid results, you can update your Drivers if necessary. (Details)
Last edited by PantherX on Mon Nov 01, 2010 7:46 am, edited 5 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

5 - GPU3 BETA Client Download Locations

Postby PantherX » Thu Sep 16, 2010 3:31 am

Legend:
Blue Text -> Important Note(s) That You Must Read
Black Text -> Applicable To Both GPU Vendors


  • GPU3 BETA Client Download Locations
Latest Versions:

[Installer] Systray 6.41 For All Windows (32 bit & 64 bit): <Click>*
MD5Sum -> 690E91725987500817AA61F3B631E824

[Drop-in] Console 6.41 For All Windows (32 bit & 64 bit): <Click>*
MD5Sum -> 95C4DDA773F09C85C466DF5FDBEE2377

Previous Versions:

[Installer] Systray 6.32 For All Windows (32 bit & 64 bit): <Click>
MD5Sum -> EFFD87BA12C96BE28E252BCCBE776FF9

[Full] Console 6.31 For Windows XP (32 bit & 64 bit): <Click>
MD5Sum -> 885E36A477D247487F8009335BD4E3CC

[Full] Console 6.31 For Windows Vista/7 (32 bit & 64 bit): <Click>
MD5Sum -> B41301886881958C64C1907B3ED6ACAE


Note 1: Please make sure that you rename the Drop-in file for the latest Drop-in Console version (6.41) from "Folding@home.exe_console_6.41" to "Folding@home.exe" (remove "_console_6.41")

Note 2: Please note the the latest Drop-in Console version (6.41) comes in an executable file replacement form. That means you should already have GPU3 Beta Client Console version 6.31 on your system so you just replace the .exe file. It is preferred to use the -oneunit flag before upgrade to minimize WU loss.


*Now available on High Performance Windows Clients download page
Last edited by PantherX on Sat Jan 29, 2011 8:55 am, edited 8 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

6 - Installation Of Systray Version

Postby PantherX » Thu Sep 16, 2010 3:33 am

Legend:
Blue Text -> Important Note(s) That You Must Read
Black Text -> Applicable To Both GPU Vendors


  • Installation Of Systray Version
The Client by default is installed in C:\Program Files\Folding@Home and you may have to start the application from there. This will cause a screen to appear where you can configure the Client. Restart the Client and if everything is properly configured, it will then download an appropriate WU along with the required FahCore.

Official Installation Guide: <Click>


Note 1: Make sure that you have the correct paths in the shortcut. By default, everything should work without any issue:
Windows All Versions (32 bit) -> Target: C:\Program Files\Folding@home\Folding@home-gpu\Folding@home.exe
Windows All Versions (64 bit) -> Target: C:\Program Files (x86)\Folding@home\Folding@home-gpu\Folding@home.exe

Windows XP (32 bit & 64 bit) -> C:\Documents and Settings\[username]\Application Data\Folding@Home-gpu
Windows Vista (32 bit & 64 bit) -> C:\Users\[username]\AppData\Roaming\Folding@Home-gpu
Windows 7 (32 bit & 64 bit) -> C:\Users\[username]\AppData\Roaming\Folding@Home-gpu
Last edited by PantherX on Tue Nov 30, 2010 4:57 pm, edited 3 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

7 - Installation Of Console Version

Postby PantherX » Thu Sep 16, 2010 3:34 am

Legend:
Blue Text -> Important Note(s) That You Must Read
Black Text -> Applicable To Both GPU Vendors


  • Installation Of Console Version
The installation requires you to simply unzip the file in the folder of your liking and run "Folding@home-Win32-GPU" OR "Folding@home-Win32-GPU.exe" (depending on your Windows settings) which will allow you to configure the Client. Restart the Client (by using "Ctrl + C" to exit) and if everything is properly configured, it will then download an appropriate WU along with the required FahCore.

Official Installation Guide: <Click>


Note 1: To a layman, using "Ctrl + C" OR "X Button" to exit the Client may appear to be the same but there is a technical difference between them. Using "Ctrl + C" informs the console application to send a wm_close message to all child threads, while pressing the "X Button" on the console triggers the onclose function of the console application and if this doesn't handle closing all child threads, you may get into trouble (not necessarily, but the possibility does exist). Thus we prefer to use "Ctrl + C" to minimize the possibility of any error.

Note 2: Make sure that the F@h Folder isn't located in the Program Files. If it is, it is advised to change it to a simpler location like C:\F@H due to the UAC implemented in Windows Vista and Windows 7 which may conflict with the Client and may cause unexpected results.
Last edited by PantherX on Tue Nov 30, 2010 4:58 pm, edited 5 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

8 - Upgrading The Console Version

Postby PantherX » Thu Sep 16, 2010 3:36 am

Legend:
Blue Text -> Important Note(s) That You Must Read
Black Text -> Applicable To Both GPU Vendors


  • Upgrading The Console Version
Download the correct GPU3 BETA Client version and after you unzip the file, copy/paste the files into the pre-existing GPU2 Client folder. If asked to overwrite the file(s), answer "Yes". The already created shortcuts should work fine.


Note 1: Users of GPU2 Client, upgrading to GPU3 BETA Client, are advised to use the -oneunit flag before upgrading. After the Client exits, proceed with the upgrade.
Last edited by PantherX on Sun Sep 26, 2010 1:39 pm, edited 5 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

9 - Installation Of Multiple Instances

Postby PantherX » Thu Sep 16, 2010 3:38 am

Legend:
Blue Text -> Important Note(s) That You Must Read
Black Text -> Applicable To Both GPU Vendors


  • Installation Of Multiple Instances
Initially install the first Client and make sure that it is running smoothly before installing multiple instances.

Detailed Installation Guide: <Click>

Quick Installation Guide: <Click>


Note 1: To find information about the GPUs installed in your system along with the driver information, you can use GPU-Z. For obtaining the X value in -gpu X, you should use CUDA-Z.
Last edited by PantherX on Tue Oct 26, 2010 11:40 pm, edited 6 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

10 - Upgrading From A Non-Fermi GPU To A Nvidia Fermi GPU

Postby PantherX » Sun Sep 26, 2010 1:33 pm

Legend:
Blue Text -> Important Note(s) That You Must Read
Black Text -> Applicable To Both GPU Vendors


  • Upgrading From A Non-Fermi GPU To A Nvidia Fermi GPU
If you are removing a Non-Fermi GPU and replacing it with a Nvidia Fermi GPU, there are some steps that you have to take to ensure a smooth transition and continue folding without much hassle. Please note that this isn't for running a Nvidia Fermi and Non-Fermi GPU in one system, that is covered in another section.

Step 1: Use the -oneunit flag on the GPU Client
Step 2: Once the Client exit, uninstall the F@h GPU Client and delete anything that is related to it*
Step 3: Power down the system and swap the Non-Fermi GPU with the Nvidia Fermi GPU
Step 4: Install the GPU3 BETA Client and run it
*Depending on whether you installed the Systray Version or the Console Version, you may have to do the following:

For Console Version:
Step A: Locate the folder in which you have installed the GPU Client
Step B: Delete it and any shortcuts you made to it

For Systray Version:
Step A: Press Windows Key + R to open Run window
Step B: Type this %APPDATA% (with the % sign) and press enter
Step C: A window will appear (C:\Users\[username]\AppData\Roaming\ OR C:\Documents and Settings\[username]\Application Data\) and there you have to delete a folder called "Folding@Home-gpu"
Step D: Delete any shortcuts you have made


Note 1: Make sure that you delete the FahCore_xx, queue.dat and work folder as incorrect version\type will cause problems and won't allow your F@h GPU3 BETA Client to function normally
Last edited by PantherX on Fri Jan 28, 2011 12:29 am, edited 2 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

11 - Using A Nvidia Fermi GPU & A Nvidia Non-Fermi GPU In On

Postby PantherX » Sun Sep 26, 2010 1:47 pm

Legend:
Blue Text -> Important Note(s) That You Must Read


  • Using A Nvidia Fermi GPU & A Nvidia Non-Fermi GPU In One System
When combining a Nvidia Fermi GPU with a Nvidia Non-Fermi GPU, the Nvidia Fermi GPU will fold FahCore_15 WUs while the Nvidia Non-Fermi GPU will fold FahCore_11 WUs. However, if the Nvidia Non-Fermi GPU attempts to fold a FahCore_15 WU, it will result in an error if the Clients aren't properly configured. In certain situations, this setup may not work. Nonetheless, you have the following options:

Option 1: Make sure that there are separate folders for each GPU. After that, use these flags with the correct GPU:
Nvidia Fermi GPU: -forcegpu nvidia_fermi
Nvidia Non-Fermi GPU: -forcegpu nvidia_g80
You may use one or both flags to see if your GPUs can fold FahCore_15 WUs without any issue.

Option 2: Run one instance of GPU2 Client for the Nvidia Non-Fermi one and one instance of the GPU3 BETA Client for the Nvidia Fermi one. IMO, you may want to use the Console Version of both Clients as all the files are included in it and it is a no-nonsense version which makes it easier to set-up and reduces the possibility of errors.

Option 3: Takeout the Nvidia Non-Fermi GPU from your current system and place it in another system (should also contain Nvidia Non-Fermi GPUs only) so that it can fold FahCore_11 WUs and FahCore_15 WUs without any issue.


Note 1: Using the -forcegpu flag with the wrong GPU will mess things up. It may invalidate the scientific results produced. DO NOT MISUSE THESE FLAGS. (Details)

Note 2: Since there are different versions of FahCore_15 for the Nvidia Fermi GPUs and Nvidia Non-Fermi GPUs, plesae make sure that you have the correct FahCore_15 version by checking your FAHlog. Haven't the latest version isn't required since the Client will automatically upgrade FahCore_15 if the WU needs it:

FahCore_15 For Nvidia Fermi GPUs:
Version 2.15 (Tue Nov 16 09:05:18 PST 2010)
Version 2.14 (Thu Nov 11 10:05:53 PST 2010)
Version 2.13 (Mon Nov 8 13:19:45 PST 2010)
Version 2.11 (Fri Nov 5 12:50:40 PDT 2010)
Version 2.09 (Thu May 20 11:58:42 PDT 2010)

FahCore_15 For Nvidia Non-Fermi GPUs:
Version 2.15 (Tue Nov 16 08:44:57 PST 2010)
Version 2.13 (Mon Nov 8 13:20:33 PST 2010)
Version 2.12 (Mon Nov 8 11:24:19 PST 2010)
Version 2.11 (Fri Nov 5 12:37:29 PDT 2010)
Version 2.09 (Thu May 20 11:51:02 PDT 2010)
Last edited by PantherX on Fri Jan 28, 2011 12:30 am, edited 2 times in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

12 - Using Multiple Nvidia Fermi GPUs In One System

Postby PantherX » Sun Sep 26, 2010 1:49 pm

  • Using Multiple Fermi GPUs In One System
If the F@h GPU3 BETA Client isn't getting enough CPU Cycles, the drop in PPD is significant especially if you have multiple Nvidia Fermi GPUs installed in one system. The reason is that the GPU is waiting for the CPU to give it data to process thus it causes a delay in the processing of the WU which is reflected negatively in your PPD.

A rough estimate is 1 Nvidia Fermi GPU = 1/2 CPU free. Thus systems with Dual Nvidia Fermi GPUs would need 1 free CPU. Triple and Quad Nvidia Fermi GPU systems would need ~2 CPUs free (Results taken from Q9550 @ 2.83) It will vary depending on CPU Model, CPU Speed and other factors, hence you must experiment with your system to find the highest PPD possible. You may also want to experiment with the F@h Core Priority to optimize your F@h Clients. Do not use Task Manager to change this, use the F@h Clients' settings only.
Last edited by PantherX on Fri Jan 28, 2011 12:32 am, edited 1 time in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

13 - Changing From One GPU Brand To Another

Postby PantherX » Sun Sep 26, 2010 1:50 pm

Legend:
Red Text -> Specifically For AMD/ATI GPUs
Green Text -> Specifically For Nvidia GPUs
Blue Text -> Important Note(s) That You Must Read
Black Text -> Applicable To Both GPU Vendors


  • Changing From One GPU Brand To Another
When you replace your existing GPU from an ATI/AMD one to a Nvidia one OR from Nvidia one to an ATI/AMD one, you should prepare your system for the new GPU to ensure that you won't have problems when running the F@h Client. Please note that this isn't for running an ATI/AMD and Nvidia GPU in one box/rig/system. Below is the procedure that you must follow to ensure a smooth transition from one GPU brand to another GPU brand:

Replacing an ATI/AMD GPU with Nvidia GPU:
Step 1: Use the -oneunit flag and wait for the Client to upload the wuresult and exit
Step 2: Uninstall the F@h GPU Client and delete anything that is related to it* (you may want to remove the Environment Variables if you have set them up)
Step 3: Uninstall ATI/AMD Drivers using its uninstaller
Step 4: Install Driver Sweeper (Don't run it)
Step 5: Switch off system and swap the ATI GPU for Nvidia one
Step 6: Boot into Safe Mode and run Driver Sweeper to remove any remnants of ATI Drivers if they exist
Step 7: Boot into Windows (normal) and install the appropriate Nvidia Driver
Step 8: Reboot the system and download the F@h GPU Client for Nvidia
Step 9: Install the Nvidia F@h GPU Client and run it

*Please read Note 1

Replacing a Nvidia GPU with an ATI/AMD GPU:
Step 1: Use the -oneunit flag and wait for the Client to upload the wuresult and exit
Step 2: Uninstall the F@h GPU Client and delete anything that is related to it*
Step 3: Uninstall Nvidia Drivers using its uninstaller
Step 4: Install Driver Sweeper (Don't run it)
Step 5: Switch off system and swap the Nvidia GPU for an ATI/AMD one
Step 6: Boot into Safe Mode and run Driver Sweeper to remove any remnants of Nvidia Drivers if they exist
Step 7: Boot into Windows (normal) and install the appropriate ATI/AMD Driver
Step 8: Reboot the system and download the F@h GPU Client for ATI/AMD
Step 9: Install the ATI/AMD F@h GPU Client and run it
Note: If you want to make folding more efficient, please use Environment Variables; Guide 1, Guide 2.

*Please read Note 1

Please remember that the above method is for a single GPU brand in a single system. While it is possible to fold on an ATI and Nvidia GPUs in one system, it is complex and doesn't necessarily work for everyone.


Note 1: Depending on whether you installed the Systray Version or the Console Version, you may have to do the following:

For Console Version:
Step A: Locate the folder in which you have installed the GPU3 BETA Client
Step B: Delete it and any shortcuts you made to it
Note: Make sure that you delete the FahCore_xx, queue.dat and work folder as incorrect version\type will cause problems and won't allow your F@h GPU Client to function normally

For Systray Version:
Step A: Press Windows Key + R to open Run window
Step B: Type this %APPDATA% (with the % sign) and press enter
Step C: A window will appear (C:\Users\[username]\AppData\Roaming\ OR C:\Documents and Settings\[username]\Application Data\) and there you have to delete a folder called "Folding@Home-gpu"
Step D: Delete any shortcuts you have made
Note: Make sure that you delete the FahCore_xx, queue.dat and work folder as incorrect version\type will cause problems and won't allow your F@h GPU Client to function normally
Last edited by PantherX on Tue Nov 30, 2010 5:00 pm, edited 1 time in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

14 - Notes Regarding GPU3 BETA Client

Postby PantherX » Sun Sep 26, 2010 1:51 pm

Legend:
Green Text -> Specifically For Nvidia GPUs
Black Text -> Applicable To Both GPU Vendors


  • Notes Regarding GPU3 BETA Client

1) Distribution of WUs are like this:
GTX 400 Series GPU, GTX 500 Series (Fermi) = Uses only FahCore_15
8000 Series GPU, 9000 Series GPU, GTX 200 Series GPU (Non-Fermi) = Uses FahCore_11 & FahCore_15


2) The GPU3 Client uses new Project and won't be porting any of the GPU2 Project.

3) The Viewer that is present in the Systray Version is known to be buggy/resource hog. You don't have to run the viewer for the Client to fold and it isn't used to monitor the Client. To monitor the Client, please visit the tools list and download an appropriate application.

4) Please note that when using the passkey with the GPU3 BETA Client (also for GPU2 Client), some of the WUs (GPU ONLY) will not be credited under the passkey/username combination BUT will be credited under the username only. However, in the future, this will change so that the WUs will be credited under the passkey/username combination (Details)

5) The GPU3 is in BETA stage so the Credits may change: <Click>

6) In the FAHlog file, the version is represented as 6.30r2 and it is the same as 6.32. It's just a different way of representing the same information. Here are some more examples:
A) 6.30 R2
B) 6.3 R2
C) 6.3R2
D) 6.32
Last edited by PantherX on Tue Nov 30, 2010 5:01 pm, edited 1 time in total.
User avatar
PantherX
Super Moderator
 
Posts: 6614
Joined: Wed Dec 23, 2009 9:33 am

Next

Return to V6 GPU3 beta (including Fermi) OpenMM

Who is online

Users browsing this forum: No registered users and 1 guest

cron