GPU2 configuration guide

[Xilikon]

You are in the right place if you have a NVIDIA or ATI video card and plan to fold with it. The guide will outline the installation process in a very simple and clear manner to help even those who doesn't know a lot on computers. This will be divided in 4 parts (4 posts) :

-Basic GPU2 installation and configuration.
-Advanced configuration and tweaking.
-Troubleshooting issues.
-Multi-GPU configuration.


Now, on the guide...


Part 1 : Preparation

The first step is to make sure your GPU can run the client. If you have a ATI GPU, the requirement is to be compatible with CAL and this mean that any card from X2xxx (R6xx) series and above can run any CAL application. This also means that unfortunately, X1xxx and older is not compatible because the shaders architecture is not unified. If you own a NVIDIA card, it must be compatible with CUDA, so the 8 series and above is compatible (a list is available to check if your card is compatible, including the GeForce, Tesla and Quadro lines). The GeForce series 7 and older is not compatible for the same reason as ATI.

After you checked that your GPU is compatible, you should update the driver to be able to use the client. NVIDIA official CUDA 177.35 driver can be downloaded here and you should use this version unless there is a compelling reason to use a different one. ATI driver supported for folding is currently the Catalyst 8.7 that you can find it here. To update the driver, it's suggested that you uninstall the old driver, reboot then install the new driver.

When it's done, it's time to download the current Folding@Home GPU2 client, which is currently 6.20 beta and found here. When you are done downloading, it's time to go to the next part...

Part 2 : GPU2 Systray installation

When you are done downloading the client, it's time to install so click on it to run. You should see this :

Click on next then accept the license. Click next and you should see this :

Click on next and it will install the necessary files. When it's done, close the installer and look at the new entry in the Start menu. The short cut is called Folding@home under the Folding@home-gpu folder so click on it. You will be presented with this configuration window :

After entering the necessary informations, click on the tab to get the next panel :

After configuring, click the tab again to get to the last panel :

For now, if you are unsure about the settings in the last panel, you can leave them alone and come back later when you feel more comfortable or after you asked questions and gotten the desired answers.

When it's done, click on Ok and it should run fine.

Part 3 : How can I monitor the client ?

Warning : For now, the graphical viewer is buggy and can cause issues if you open it to monitor the progress. For this reason, we strongly suggest you avoid using the viewer and rely on alternate methods below.

The simplest way is to check the log file generated by the client. For this, you can right-click on the systray icon to bring this menu :

Click on the item and you will see a notepad with the contents of fahlog.txt, which should look similar to this :

Just scroll to the bottom to see the percent done and check if uploads and downloads work fine. It's also a great way to check if you have problems like UNSTABLE_MACHINE or any other error.

A quick way is also possible by hovering the mouse cursor over the systray icon to have a balloon showing what the client is doing currently like that :

Finally, you can also use a third-party monitoring software like Fahmon or Fahspy. The fahlog.txt file can be found in the APPDATA folder, which can be found by opening the client shortcut properties and looking at the "Start in" field.

If you followed the guide carefully, the client should be running now and producing lots and lots of points for the good of the science If you have questions about the GPU2 client, a good FAQ has been written by the F@H staff and it can be found here or here.

If you have comments or suggestions about the guide, please PM me. If possible, make a new thread if you have a question or a problem and you need help.

 

[Xilikon]

Setting the core affinity with NVIDIA GPU2 client (original by Sunin)

If you run a NVIDIA client with core version 1.07 or above (this can be found by looking at fahlog.txt, before a workunit start), you can configure the affinity precisely. This can often boost the performance, especially when you are trying to run multiple clients or you want to avoid using core 0 to reduce the graphics slowdown.

With core 1.07 or later, they introduced a new environment variable called NV_FAH_CPU_AFFINITY and it take a binary mask. A binary mask is just a programming trick to set the bits but for the majority, it may be a bit confusing so here is a full list to help you get the correct value :

0  = all cores
1  = core 0
2  = core 1
3  = core 0 + 1
4  = core 2
5  = core 0 + 2
6  = core 1 + 2
7  = core 0 + 1 + 2
8  = core 3
9  = core 0 + 3
10 = core 1 + 3
11 = core 0 + 1 + 3
12 = core 2 + 3
13 = core 0 + 2 + 3
14 = core 1 + 2 + 3
15 = core 0 + 1 + 2 + 3

Alternatively, there is also a simple way to determine the value :

Core 0 = 1
Core 1 = 2
Core 2 = 4
Core 3 = 8

If you want core 1 and 3 only, add 2 and 8 to get 10.
If you want core 0, 2 and 3, add 1, 4 and 8 to get 13.

For your information, if you have a quad-core, setting a value or 0 or 15 set the same affinity to all cores. If you have more than 4 cores, each additional core always take a 2^n value where n is the core number, starting from 0.

Now that you know how to determine the value to enter in the environment variable, we need to create that variable. Here is the simple steps (OS independent) :

-Go to start and run type in: sysdm.cpl then a system config panel should pop up.
-Under the Advanced tab click the environment variables button.
-Click new and under Variable name put NV_FAH_CPU_AFFINITY
-Under Variable value put the corresponding # you desire... Look above to determine the desired value.
-Walla, you have now setup the environmental variable hit Ok and close the window.
-Stop and restart the GPU2 client to make use of the new environment variable.

Your GPU should now utilize whatever cores you specified! I generally use 15 to allow it to use all cores. Then just check the box as indicated above in Xil's guide to not lock the GPU to a single core and your ready to roll!

Overclocking your video card

Soon

 

[Xilikon]

Most common issues and solutions

- If you get EARLY_UNIT_END or UNSTABLE_MACHINE, you have a setup that MIGHT require you to trick your system into thinking it has a monitor plugged into the 2nd video card or you Overclocked it too much! close the GPU client now by right clicking the tray icon and selecting Quit. If you overclocked the card, return the card to stock. If you still EUE then you will have to try one of two things:

A. Add the flag -forcegpu nvidia_g80 (or ati_r600 or ati_r700 if your running ATI) :

Note: All nvidia cards work with the nvidia_g80 flag

Run the GPU client again and see if it still errors, if so you are probably running Vista and need to do step B below.

B. Right click your desktop and click personalize, select Display settings. You should see a tab that says monitor with one that is dark black while another is grey. Select the 2nd display and click Extend the desktop onto this monitor.

Restart your client and you should be good to go now.

-Slow Performance 2-3k with 8800GTs for example. You will need to remove the drivers and reinstall them. Go to your start menu and under control panel find add/remove programs. Uninstall the driver. Download a driver cleaner. I use driver cleaner from guru3d.com, and install it. Reboot into safe mode by hitting F8 before the windows screen comes up and select safe mode. Run the driver cleaner program and clean the nvidia or ati graphics drivers out. Then reboot. As windows boots up it may ask you to search for a driver hit cancel and then run the driver you downloaded in step 2 above.

-Video card not recognized. Perform step 1 A. above. It means your system didn't properly detect your video card and you need to provide the -forcegpu command.

I didn't find my problem in the list above, what more can I do ?

The first step is to check the logs to see what kind of issue you are having then do a search on the DC forum here to see if someone else had this issue and if a solution is found. If no problem is found or no solution is known, the next step would be to first uninstall the client and delete the user data before reinstalling the latest client.

If uninstalling and reinstalling the client doesn't work and it cannot fold due to a driver issue, uninstall the current driver, run DriverCleaner in safe mode then reinstall the CUDA 177.35 driver before reinstalling the client.

If the issue is related to networking (inability to upload or download), the first thing you should do is to disable the firewall to see if it works. If not, look at your router if the configuration is preventing workunits from downloading or uploading. Alternatively, try to connect the pc directly to the internet, bypassing the router and network to determine. If it still doesn't work, do a tracert to the IP listed in the log file to see if it's a ISP having issues or a dead router along the path to destination.

If you still have issues, chances are that the F@H server is having a issue. Just wait for a few hours (up to 24 hours if it is in the weekend) to see if it will end up connecting and download/upload correctly. If not, make a new post and see if others have issues.

 

[Xilikon]

After you install your GPUs:

When windows ask to look for a driver hit CANCEL. Everytime I've let windows install the driver I end up with sub-par performance.

Download and install the necessary Files:

-Download the most recent Nvidia Cuda driver here : http://www.nvidia.com/object/cuda_get.html click the operating system you need and search. As of writing this, 177.35 are the most recent drivers. Install them by click on the file you downloaded.

-Download the current GPU 2 client found above. As of writing, 6.20 is the most recent GPU 2 program. Install the GPU Client by clicking the file you just downloaded from Stanford.

Preparing, Installing and Configuring the GPU Client:

1. 6.20 currently does not copy a shortcut into the startup folder. This step is here in case a future release starts doing that nonsense again! Double check it anyways... to make sure since earlier versions did. If it exists delete it or face headaches!

2. Next, open up windows explorer and find the directory it installed the GPU client under your user name. Below is a picture of the location depending on the OS :

XP :

Vista:

3. Once you’ve located your directory, you will want to cut and paste a copy into the same location and cut the “Copy of” off and add a 2 after the directory name. See example under User directory:

Before:

After:

4. Now go into GPU 2 client in the Program directory and make a shortcut of the executable :

5. Cut and paste that shortcut out to your desktop. Rename by right clicking it and selecting rename and call it something easy, I use GPU1

6. Right click the shortcut and select properties :

7. It should bring up the following screen:

8. Now open up explorer again and navigate to the GPU Directory that is under your user name. Once you are in the directory, go to the navigation bar and cut the location out :

9. Paste it in the Start in section :

10. Make sure to surround it with Quotes!

11. Under Target after the Quotes add – GPU 0 and make sure you put a space between GPU and the 0 as shown here :

12. Once your done hit Complete it should look like this before you click OK :

13. Now copy and past that shortcut once and rename that to GPU 2 or whatever you enjoy:

14. Right click the newly created shortcut and select properties. This time it is much easier! Select the start in directory and add a 2 after the directory name as shown in this before and after shot :

Before and after:

15. Change the target line from –GPU 0 to –GPU 1 and hit OK :

16. You may want to set the environmental variable as shown in Xilikon’s guide above. If you don't set a environmental string it defaults to NULL and uses all 4 cores.

17. Now click on the GPU1 and configure your client as is demonstrated in Xilikon’s guide above. The key things to make sure you do are (Well besides folding for team 33):
-On the 2nd tab Click yes to accept WUs larger than 10 megabytes.
-On the 3rd tab Check the Core Priority for Slightly Higher
-On the 3rd tab Click do not lock cores to specific CPU
-On the 3rd tab make sure the Machine ID is different then any cpug or gpu client you have running. Note default is 2.
I generally setup my smp's to be 1 and 4 and let gpu be 2 and 3. Here is what it should look :

18. Now hit accept and repeat step # 17 for the GPU2 shortcut and make sure you pick a unique machine ID.

19. Congrats, you are up and folding !

 

[Sunin]

Not all dual gpus need to be seen by windows. Both of my Abit-IP35 Pros have teh 2nd 88GT not connected to anything and it is producing just fine. How to tell if you need to attach the card is if it EUE's and goes unstable at stock speeds. If it chugs away with no issue you'll be ok without plug'n it in to a monitor or dummy vga adapter.

 

[Dan-515]

This is good

 

[Sunin]

See it added up top removed for now...

 

[Sunin]

We need a GPU OCing Guide or links to guides already out there:

G80 Guide:
Guide to G80 OC'n
I followed that guide for my 8800GTX's it works flawlessly.

G92 Guide:
Gide to G92 OC'n and Fan Setup!
Everything works except making the usb bootable flash drive... if that worked I'd have about 10 G92 cards perma OC'd....

Priority for OCn

1. Shader
2. CPU
3. Memory

My experience is 90% of the production is shader based. For example:

88GT at 700/1850/1800 = 4500-5000 PPD depending on WU etc.
88GT at 600/1500/1500 = 2500-3000 PPD depending on WU etc.

It is amazing the effect it has to do the shaders higher:

Another example:
88GTX at 575/1350/900 = 3300 PPD
88GTX at 620/1750/1050 = 5500 PPD

So you can see that even though I'm only OC'n the shader 23-26% the PPD goes up 67-80%... so a great way to not add a ton of heat and boost your PpDpW!!!

 

[Xilikon]

Everyone can contribute to the guides. I'll pick the good ones and integrate with the guide (I reserve the right to edit to make it much simpler to follow or to clarify certain points).

Sunin, good job and I added your environment variable setting guide above (I expanded with my own experience).

 

[collegeboy69us]

So -- I just got my GTX 280 OC - got it installed and folding just fine.

I put my 8800GTS into my mediabox and got that folding just fine. I woke up this morning and both of them finished their work units but then just stopped. I restarted the machines and now anytime it goes to start up the GPU core i get the message saying "could not start, no compatible video card could be found, please check your drivers etc etc"

Why would both machines fold fine thru the night and then just finish and crap out like that and not be able to start back up? I'm using a fresh install of Vista 64 and the 177.35 drivers on both.

Any suggestions?

Thanks!
Jeff

 

[Xilikon]

Did you try restarting the computer and see if it works fine ?

This error is often tied to a badly installed driver (or it's not installed at all and still using a older driver like 175.19).

 

[Avrithor]

Another example:
88GTX at 575/1350/900 = 3300 PPD
88GTX at 620/1750/1050 = 5500 PPD

I wish mine would run at 1750 shaders
ATI Tool starts reporting errors at 1500 and system crashes come shortly thereafter. By contrast I have my 8800GS at 1500 shaders 100% stable and I think it will go faster. WTF!! Sorry, I guess this isn't really an overclocking thread, but if anyone has any insight...

Great guide Xilikon! I've been running the GPU2 console clients without the glitchy 3D display or tray icon, downloaded from foldingforum.org. I like just keeping console windows open on my other boxen to keep an eye on their progress while I do work or game on my main rig, and FahMon doesn't work well for me, it constantly reports half my clients as hung when they're not.

 

[collegeboy69us]

Did you try restarting the computer and see if it works fine ?

This error is often tied to a badly installed driver (or it's not installed at all and still using a older driver like 175.19).

Yeah I did a restart - I will run driver sweeper and try to reinstall the drivers, but it doenst make sense if it worked fine and then a few hours later just stopped.

Jeff

 

[Xilikon]

Yeah I did a restart - I will run driver sweeper and try to reinstall the drivers, but it doenst make sense if it worked fine and then a few hours later just stopped.

Jeff

Try this and if it still fail, please paste the fahlog.txt (50 lines around where it fails) and we should be able to tell what went wrong.

 

[MrBooMY]

Got my shizz all hooked up. Now I have GPU client running on 1 core, and the regular running on the other. I'm guessing the GPU client works in tandem with your processor because it uses up CPU cycles. It seems to work MUCH faster than the standard core by itself.

 

[collegeboy69us]

with Vista64, a Core2Quad @ 3.0Ghz, the 280GTX would use between 15% and 30% of a single core and be able to crank out 6500-7500 ppd from the GPU.

with Vista 64 a Core2Duo @ 2.5 feeding a 880GTS, the CPU usage is only about 10% on a single core, cranking out about 4000 ppd.

Later tonight if i can't figure it out or get it to work I'll past the fahlog.txt file for both computers and see if you guys can help.

 

[tkam]

Thanks for the multi-gpu guide, just added a 2nd 9800gtx to my main box today and followed the guide step by step. Worked like a charm!

 

[spincut]

I'd still like to know how all this folding actually helps.

I mean, they've been at it for a while right? and it sounds like ALOT of data has been processed with this movement (not just hardocp's massive team but everyone with their cpu's, gpu's, and PS3's cell processors spare cycles all working in tandem).

so with all this raw processing power put to good use for so many years, what is the update?

I actually had always though that the find a cure for cancer or aids (which is why i thought folding was processing data for) there has to be many many tests done on a vast combination of proteins to find the one that reacts favorably to counteracting the disease. Hence it's almostl ike hitting the right one (not necesarily processing every known combination out there, likely to find one that works before all have been tried).

Am i off base on how it works though? is that right at all?

I guess I was just curious at what kind of progress has been made due directly to folding after all this time and work done with it so far.

 

[Xilikon]

I'd still like to know how all this folding actually helps.

I mean, they've been at it for a while right? and it sounds like ALOT of data has been processed with this movement (not just hardocp's massive team but everyone with their cpu's, gpu's, and PS3's cell processors spare cycles all working in tandem).

so with all this raw processing power put to good use for so many years, what is the update?

I actually had always though that the find a cure for cancer or aids (which is why i thought folding was processing data for) there has to be many many tests done on a vast combination of proteins to find the one that reacts favorably to counteracting the disease. Hence it's almostl ike hitting the right one (not necesarily processing every known combination out there, likely to find one that works before all have been tried).

Am i off base on how it works though? is that right at all?

I guess I was just curious at what kind of progress has been made due directly to folding after all this time and work done with it so far.

Actually, the processing power of F@H is not for applied science (where you research for a cure) but for pure science, to understand how proteins misfolding affect the normal development of proteins and cause diseases. With a better understanding of this event, we can then research for a cure. What you described is not what F@H do.

To better threat a disease, we need to first understand how the disease appear in the first place.

For now, they released a number of papers about that with the findings directly related to F@H and which received peer reviews from fellow scientists : http://folding.stanford.edu/English/Papers

 

[spincut]

Actually, the processing power of F@H is not for applied science (where you research for a cure) but for pure science, to understand how proteins misfolding affect the normal development of proteins and cause diseases. With a better understanding of this event, we can then research for a cure. What you described is not what F@H do.

To better threat a disease, we need to first understand how the disease appear in the first place.

For now, they released a number of papers about that with the findings directly related to F@H and which received peer reviews from fellow scientists : http://folding.stanford.edu/English/Papers

so the folding is constant and ongoing processing of new experiments and noty a finite (but very large) number of combinatory tests that need to be processed?

I guess i had always thought it was working towards a cure to cancer and that each protein fold was part of a processing block of one of many combinations that needed to be tried, and the faster the process was undertaken the fast a protein reaction that could combat it could be discovered.

so am i getting folding mixed up with something similar?

 

[MasterOfPupets]

Folding is more about sully understanding the cause of diseases than directly curing them. That being said though, once the cause is fully understood, the cure should be much easier to discover.

 

[APOLLO]

I was forced to reinstall the GPU clients on one of my systems because they expired early, and DL the new console version. Both GPU console clients are installed exactly like the CPU clients in their own folders with separate machine IDs. Is there anything else required to get them set up working properly? They are producing exactly half the PPD of the expired systray client as if they're accessing only one video card. I didn't configure any targets in the shortcut since Stanford stated that wouldn't be supported. However, I did include the -GPUx and -local flags. With or without the flags makes no difference, so, why is the production halved?

 

[nomad8u]

I was forced to reinstall the GPU clients on one of my systems because they expired early, and DL the new console version. Both GPU console clients are installed exactly like the CPU clients in their own folders with separate machine IDs. Is there anything else required to get them set up working properly? They are producing exactly half the PPD of the expired systray client as if they're accessing only one video card. I didn't configure any targets in the shortcut since Stanford stated that wouldn't be supported. However, I did include the -GPUx and -local flags. With or without the flags makes no difference, so, why is the production halved?

I stayed with the systray client and despite what Stanford said about targets not being supported, I used my old shortcuts, one for each client with the -gpu 0/-gpu 1 switch and left my work and config files in their original locations. Mine are producing normally.

Did you use -GPUx (caps) for emphasis or is that how you entered the switch? Far as I know it's small letters only. i'm guessing you did it for emphasis.

Look at both the logs and see if they're running different WUs and if both cards were recognized properly (seperat ID's. Check to see that they're each running on their own core (gpu 0 should start on core 3 and gpu 1 on core 2) for a quad. Not sure of the order on a dual core but I think gpu 0 gets core 1. It sounds like either they're running the same WU at the same time? Or both folding cores are getting starved? Do they look right for CPU usage in Task Mangler? did you run -configonly and set them up (not sure if -configonly works on GPU client as I run systray)? XP or Vista?

 

[APOLLO]

I stayed with the systray client and despite what Stanford said about targets not being supported, I used my old shortcuts, one for each client with the -gpu 0/-gpu 1 switch and left my work and config files in their original locations. Mine are producing normally.

That is good to know, but then why did Stanford specify a lack of target support in their changes log? More importantly, why did previous clients expire one day early?

Did you use -GPUx (caps) for emphasis or is that how you entered the switch? Far as I know it's small letters only. i'm guessing you did it for emphasis.

Emphasis, I'm so used to writing it that way on the forum that I did it out of habit. I actually have it lower case in my client

Look at both the logs and see if they're running different WUs and if both cards were recognized properly (seperat ID's. Check to see that they're each running on their own core (gpu 0 should start on core 3 and gpu 1 on core 2) for a quad. Not sure of the order on a dual core but I think gpu 0 gets core 1. It sounds like either they're running the same WU at the same time? Or both folding cores are getting starved? Do they look right for CPU usage in Task Mangler? did you run -configonly and set them up (not sure if -configonly works on GPU client as I run systray)? XP or Vista?

XP and the WUs are different. The system is a dual quad and I tried running on all core combinations, so that isn't the problem. I didn't try -configonly since that flag was never required by a previous console client. I just replaced the old files in the original folders with the new ones from the DL archive after uninstalling the 'old' client. I know something is wrong with my setup, but I'm too tired to figure it out with all the other folding issues I had to deal with today. Having 5 clients expire prematurely was NOT something I expected to deal with.

 

[Sunin]

**For single gpu installs just uninstall the old client and re-install the new.

**For multi gpu installs uninstall your first installation of the gpu just like single gpu installs.

For your 2nd, 3rd or 4th installations

1. Delete everything under your User name (related to the GPU client) for my vista install it was this:
C:\Users\David\AppData\Roaming\Folding@home-gpu2
EXCEPT your Work directory with all of its content, queue.dat, unitinfo, Fahlog and MyFolding.
Make sure to delete the following
the FahCore_11
amdcalcl.dll
amdcalrt.dll
cudart.dll
client.cfg

2. Find your GPU Client directory under your program directory for my vista install it was
C:\Program Files (x86)\Folding@home\Folding@home-gpu2
and delete the entire directory.

Now intsall the GPU client as you normally would

3. Copy from the new install directory for vista mine was:
C:\Users\David\AppData\Roaming\Folding@home-gpu
Copy the following files: amdcalcl.dll, amdcalrt.dll, cudart.dll, and Fahcore_11 (if it exists, generally does not) into:
C:\Users\David\AppData\Roaming\Folding@home-gpu2

4. If you kept your shortcuts on your desktop, which you should have since I never said delete those you should now be able to fire up both clients. (gpu1 and gpu2, etc) You will have to make one change:
Change the Target directory back to the original install. I you had it under:

C:\Program Files (x86)\Folding@home\Folding@home-gpu2

REMOVE the 2 or your client will never get WUs! These directories should be the same


*** So to recap. Same Program Directory and Different Directories under your user name!****

6. It should go into a config screen for each client and make sure you assign them different machine IDs, etc just like you did when you originally set them up. Note you no longer have to edit the config file to change that. You now have the optionon the last tab at the very bottom to set Machine ID in the config screens!! I generally set >10M yes, Core Priority to Slightly Higher, Saves to 30minutes and check the box to not lock to specific core.

7. After you fill in the needed info in and check the correct boxes your set to go.

8. Troubleshooting: Note when I did this the servers were having issues and only one of my clients was able to download the new 1.08 FahCore_11.exe. So I closed both gpu clients down and manually copied the client to the 2nd directory.

 

[Sunin]

Targets .....
C:\Program Files (x86)\Folding@home\Folding@home-gpu -gpu 0
C:\Program Files (x86)\Folding@home\Folding@home-gpu2 -gpu 1


With 6.20 this part of Tiger's guide is in error. It should be:

C:\Program Files (x86)\Folding@home\Folding@home-gpu -gpu 0
C:\Program Files (x86)\Folding@home\Folding@home-gpu -gpu 1

 

[Sunin]

Another wonderful tip!

Say you want to run without a monitor, this applies for both single and multi gpu's

add this command after your executable or after -gpu x flag

-forcegpu nvidia_g80

This now solves the issue with windows having to have a monitor attached... In fact you could have a box setup to run without any monitor at all now!

WOOT!

 

[Xilikon]

Another wonderful tip!

Say you want to run without a monitor, this applies for both single and multi gpu's

add this command after your executable or after -gpu x flag

-forcegpu nvidia_g80

This now solves the issue with windows having to have a monitor attached... In fact you could have a box setup to run without any monitor at all now!

WOOT!

If I understand you correct, if I add this flag in the client, it will fold on the card even if nothing is plugged ? If yes, that's so awesome !

 

[Sunin]

Bingo Xil! and it is awesome!!!

 

[Dan-515]

Another wonderful tip!

Say you want to run without a monitor, this applies for both single and multi gpu's

add this command after your executable or after -gpu x flag

-forcegpu nvidia_g80

This now solves the issue with windows having to have a monitor attached... In fact you could have a box setup to run without any monitor at all now!

WOOT!

Brilliant, does it work on any nvidia gpu using the 'nvidia_g80' ? Or do you have to use the code for your gpu?

 

[Sunin]

Brilliant, does it work on any nvidia gpu using the 'nvidia_g80' ? Or do you have to use the code for your gpu?

Yes any nvidia gpu just use the nvidia_g80 code.

 

[Dan-515]

Yes any nvidia gpu just use the nvidia_g80 code.

Excellent news. I wasted £10 on little resistors and VGA plugs

 

[APOLLO]

Another wonderful tip!

Say you want to run without a monitor, this applies for both single and multi gpu's

add this command after your executable or after -gpu x flag

-forcegpu nvidia_g80

This now solves the issue with windows having to have a monitor attached... In fact you could have a box setup to run without any monitor at all now!

I've had a dual GPU system set up without any monitor or terminator for a while now, but it's running XP. I think it might have been different with Vista systems previously.

 

[APOLLO]

I was curious about something and I think it's probably best to ask here. What are the minimum CPU requirements for GPU folding? Can any CPU be effective as long as it is able to feed the GPU, or are there specific generations that demarcate which CPUs can and cannot fold with the GPU client?

 

[nomad8u]

Another wonderful tip!

Say you want to run without a monitor, this applies for both single and multi gpu's

add this command after your executable or after -gpu x flag

-forcegpu nvidia_g80

This now solves the issue with windows having to have a monitor attached... In fact you could have a box setup to run without any monitor at all now!

WOOT!

Just to add a little background on this and give up a little lovin' for our ATI Brothas...

there are only three flags: ati_r600 , ati_r700 and nvidia_g80 . It's mostly for use with things like the tesla range of cards (on nvidia's side), where you aren't going to be able to use a screen to start the program. I would guess that a change was made between the r600 and r700 series which meant that they need to be handled slightly differently, whereas G80, its descendants and successors are all similar enough to be handled the same way.

So for you ATI folders with second and third cards that are in the r600/700 family you can use this as well.

 

[smileyscout]

This might be a silly question but how do you set the GPU client priority to be slightly higher in the service client instead of the graphical client?

 

[Sunin]

I've had a dual GPU system set up without any monitor or terminator for a while now, but it's running XP. I think it might have been different with Vista systems previously.

In 6.20 they broke xp to mirror vista now you will get a gpu not supported message if it is not plugged in or EUE's...

I can confirm that Vista still needs a monitor plugged in..


Smiley - Same way you did when you configured it the first time. Run the exe with -smp -configonly and specify the priority... its on tab 3 if memory serves

 

[Sunin]

Hmm double post?

 

[Xilikon]

Added Sunin's exhaustive multi-gpu guide with pictures and added some troubleshooting tips from Sunin as well.

If you see errors or issues with the guide, PM me and I'll fix asap.

 

[APOLLO]

In 6.20 they broke xp to mirror vista now you will get a gpu not supported message if it is not plugged in or EUE's...

I can confirm that Vista still needs a monitor plugged in..

I haven't changed my setup, though. Everything is the same as it was before the client expiration. No monitors at all on the multi-GPU system. No VGA terminators either.