am i losing out?

Se7ven

Se7ven

Se7ven

Associate
Joined
7 Nov 2009
Posts
1,472
hey guys im relatively new to f@h and i have only just realised you can fold using your GPU(s)?

can someone give me a rough idea how to get this going and do you get a lot more points for doing it? my machine is running GTX 295, will this be able to fold?

thanks and sorry for the nubness :P
 
yes you are
get that 295 working hard
all you need to do is download the gpu client from Here
Create 2 folders called gpu1 and gpu2
and install it in each folder
Place a shortcut on your desktop and edit the target by adding -gpu0 for the first client and -gpu1 for the second client.
run them and away you go
 
I fold on mine, nets 16k-19k ppd by itself if i leave it running. (it is overclocked quite a bit tho)

It is a little more complicated than just running the client.

You may still need to disable the SLi in nvidia control panel (this may not still be the case, as it was meant to be sorted in the recent driver releases, but i just havent tried it yet!) so both cores are accessible to CUDA.

You will need to extend you desktop across two monitors in the windows display options - if running vista or 7 then this requires you to have connected two monitor connections to the card - or a real one and a dummy plug.

you may need to use several flags to get the cards to operate correctly. I use "-gpu 1 -advmethods -forcegpu nvidia_g80" and "-gpu 0 -advmethods -forcegpu nvidia_g80". The ..g80 flag forces the drivers to accept it is possible to use the card as sometimes they decide otherwise!

When you install the client, it will go to 2 locations. In vista these are C:\Program Files (x86)\Folding@Home\Folding@home-gpu
and
C:\Users\USERNAME\AppData\Roaming\Folding@home-gpu
"AppData" is a hidden folder btw.

To run the client succesfully you need to replicate each of these folders with a 2 on the end (C:\...\Folding@home-gpu2)

You now need to create a shortcut to run each client.
In the 'Target' box put the "program files" path
In the "run in" box put the "appdata" path

If you want ot use soemthing like FahMon to monitor the clients, then the target folder it wants is the appdata one.
 
Se7ven - are you running just the one 295?
if so you only need the one client (i think) The rest of the info above is if you have two gpu cores. (does the 295 have two core?)

Just looked at some 295 waterblocks - yep two cores - sorry ignore me
 
Last edited:
yeah with one single core card it is as simple as just installing and running the client.
Multi gpu setups get more complicated.
 
Is your rig going to be on 24/7? Thats a nice little 16k-19kppd bonus! You'll be flying up the ranks :)
 
vertica said:
You will need to extend you desktop across two monitors in the windows display options - if running vista or 7 then this requires you to have connected two monitor connections to the card - or a real one and a dummy plug..
Click to expand...

this is what im confused about since i only have 1 monitor and only 1 DVI connection :s

and yes its on 24/7 so i want to get the most out of it ;)
 
sorry no, i meant does you gtx295 have two circuit boards or one?

the first gen ones had two circuit boards with the cooler sanwiched in the middle. on the rear panle there are two DVI and one HDMI socket, they look like this:

gtx-295-12-17-08.jpg


The second gen 'single pcb' ones, as the name suggests, only have one board, with the cooler on top. these only have 2 dvi ports.

GeForce_GTX_295_single-PCB_03.jpg



Sorry to ask but there are some subtle differences between them when setting up folding (despite being identical in terms of performance).
 
Last edited:
ok, needed to know as both cards have their own quirks - there was no point in me telling you about the dual pcb card.

Unfortunately i've got a dual pcb card, so i can now only advise on what i've read and not from experience.

Are you running xp or vista/7 ? (again the setup varies slightly)
 
Right from what i can gather you should be alright leaving the card in SLi mode (may be called 'multi-gpu mode' in NV control panel now - i cant remember).

First thing you need to get sorted is for windows to 'see' two monitors. This can be done in several ways:

  • use another monitor and hook it up to the second DVI connection
  • If your monitor has multiple inputs, connect a second lead between the gpu and monitor. (I do this, DVI->VGA convertor on the gpu and then a vga cable to my monitor's second input)
  • Add a dummy plug (guide here, pinout here, any resistors in the 60-100ohm range seem to work)

Once this second 'monitor' is attached, you should be able to extend your desktop to it in the 'display settings' screen in windows.

Now it should just be a case of installing the client once, then copying the two folders, creating shortcuts and setting up the additional paraments as above.
 
I'd suggest you take care of that BSOD you've got first (if you haven't already), because it's a pretty sure bet that taxing the card with CUDA stuff will make the problem worse.

Of course, installing CUDA-compatible drivers might be just the fix you were looking for.
 
You must log in or register to reply here.
Back
Top Bottom