windows 7 and sli

wadec22

2[H]4U
Joined
Jul 26, 2005
Messages
3,948
okay, no idea seting up folding was going to be so difficult. :(

okay, system is in sig. i have finally decided to join the [H]orde. downloaded the gpu client, tried setting things up, my client just sits there. the display still says user-123 and team -456. i entered a username, team 33 and even requested a passkey.

is win 7 making things difficult for me?

now i have noticed if i mouse over the client in the system tray, it says download error.
 
You have to disable SLI to fold using the GPU client. I'm not sure if that is what's causing your issues, but it's worth a shot.
 
ugh.... just installed latest nvidia drivers today, followed shortly by the gpu client. :(
 
for some reason it will not run on my third card though.....
Mixing 8800/9800 and GTX200 GPUs doesn't work out too well anyway. If you want to run the 9800GT, you'll have to stick it in another PC.
 
I the other thread linked above, I reported the new 190.xx drivers allow CUDA to access all GPUs with SLI enabled. I had only tested with BOINC, which works fine.

Today I tried to set up Folding@home, and at first it would only let me use the GPUs that were associated with monitors/extended desktops. I added the "-forcegpu nvidia_g80" flag to the other shortcuts and I was able to get all five GPUs folding. At that point I promptly overloaded my UPS. :D Anyway, the new drivers work. You will still need dummy plugs on any additional cards that are not included in SLI.
 
That's too bad :(.

You might be able to solve this with some of the weirdo registry hacking of the NVIDIA drivers. I think this is what the "LimitVideoPresentSources" parameter is supposed to solve. I never got very far messing around with the registry, and it seems to reset itself every time you change your hardware configuration or upgrade your drivers, so it's easier for me to just use a dummy plug.
 
i actually have a monitor hooked up to the third gpu but still cannot force it into action.
 
i actually have a monitor hooked up to the third gpu but still cannot force it into action.
Having a monitor hooked up isn't the problem. GT200 and G8x/9x GPUs are not compatible with each other for folding. There's no way to get around that as of yet.
 
so will i get more ppd by using both gtx 280's or my i7 920?
Your 920 folding by itself using VMs and the Linux SMP client will get around the same amount of points as a single GTX280. A pair of 280s will outperform the CPU by quite a lot.

Of course, the ideal thing to do would be to fold on both your CPU and GPUs.
 
I''m folding on both intel 9650 and my pair of 260s, the 9650 running 2xVMs cranks out about the same PPD as each of the 260s (if they are a2 WUs), the whole rig is cranking about 21,000 PPD.

Go big if you have the memory for the VMs, run both VMs and GPU2 clients. You will be well rewarded with showers of PPD.

But I am well versed in the "do not mix" a G92 cored 9xxx or 8xxx with a GT200 cored 2xx. It just won't work. My 8800 GS rapidly got replaced with a second 260 for my main rig. and now that I balanced the shader processors to max PPD on my main rig, my lowly 192 shader 260 is in the SWMBO's machine and the 8800GS is collecting dust.

 
looks like i have some more reading research to do. i thought i read the gpu2 client required the cpu to do some work, so you couldn't fold on the cpu at the same time. i have no idea what a vm is but i will work on it this weekend. thanks guys.
 
VM is a virtual machine. It will alow you to run linux inside the vm so you can run the linux verion of the SMP clinet (much better ppd and rock solid)

NV GPUs need just a few CPU cycles even at peek.
 
The GPU client does require a tiny bit of CPU power, but as long as you use a program like WinAFC to set their priority above the VMs, they won't take any performance hit and you'll be able to fold on both at the same time.
 
I want to get this strait......

On a NV based mobo (s775) we can now properly operate 2x, 200 series cards in sli without the system crippling the 2nd card?

Or is this only on i7 boards?
 
I want to get this strait......

On a NV based mobo (s775) we can now properly operate 2x, 200 series cards in sli without the system crippling the 2nd card?

Or is this only on i7 boards?
As far as I can tell, this has nothing to do with the motherboard. The new drivers allow individual GPUs to be accessed by CUDA while SLI is enabled, so you can have one instance of F@H for each individual GPU regardless of whether or not they are in SLI.

The practical implication of this is that you can now fold on a multi-GPU setup without disabling SLI. Gamers rejoice ;).
 
As far as I can tell, this has nothing to do with the motherboard. The new drivers allow individual GPUs to be accessed by CUDA while SLI is enabled, so you can have one instance of F@H for each individual GPU regardless of whether or not they are in SLI.

The practical implication of this is that you can now fold on a multi-GPU setup without disabling SLI. Gamers rejoice ;).

I get that part of it.

It used to be that if (and I'm trying to find out if it still is) if you put 2, 200 series cards into a NV based mobo the mobo would limit the 2nd card to 2d only mode thus giving only 1/3 the points and allowing no OC options.

I'm trying to find out if that is no longer the case..... I have an upgrade coming my way if that has changed.
 
I get that part of it.

It used to be that if (and I'm trying to find out if it still is) if you put 2, 200 series cards into a NV based mobo the mobo would limit the 2nd card to 2d only mode thus giving only 1/3 the points and allowing no OC options.

I'm trying to find out if that is no longer the case..... I have an upgrade coming my way if that has changed.

Well I've got 2x 260s now running in sli and folding, and both overclocked via rivatuner on an evga 790i board, so I'd say it's working :D I find your comment odd though, as before today I would disable sli, and overclock and fold on both cards just fine, the second card usually producing a few more pdd than the first (since the first had the windows desktop i suppose). I'm running vista ultimate sp2 x64. Let me know if you have any more questions regarding the issue.

 
Well I've got 2x 260s now running in sli and folding, and both overclocked via rivatuner on an evga 790i board, so I'd say it's working :D I find your comment odd though, as before today I would disable sli, and overclock and fold on both cards just fine, the second card usually producing a few more pdd than the first (since the first had the windows desktop i suppose). I'm running vista ultimate sp2 x64. Let me know if you have any more questions regarding the issue.


How long have you been doing this...? A few months over over a year?
 
Built the system in early Novemeber of '08, so not quite a year. I've been folding on it since then, if you go to my stats page on eoc, and look at monthly production, it should be pretty clear lol. I was using 181.22 drivers last. Never had any issues, EUE, etc (unless I tried beta drivers, but I've tried to stick to good solid performers)

*edit* as a side note, fahmon has finally started working with the new drivers. I'm getting 7500 & 7400ppd on the 260s, running 1888 point projects 5912 & 5915.
 
Last edited:
As an update - I have found folding with sli to make the desktop and general computer use impossible. It bogs down even the start menu popping up, watching a movie or browsing websites is painful. This behavior did not occur previously, and it goes away when I switch back to independent rendering mode. I have also had a few of the famous display driver recovering popup notifications when switching modes, something that I had previously not seen for myself. I think we will have to wait a few cycles for this to mature, but as of now, I cannot leave sli on. I suppose those of you with gpu rigs that do nothing but fold it is ok, but for gamers, or people who actually use the computer, it is unacceptable right now.
 
As an update - I have found folding with sli to make the desktop and general computer use impossible.

I've noticed some sluggishness on my machine too. It goes away when I pause the F@H clients. My box is kind of a hybrid workstation/server, and desktop performance is still good enough that I don't mind. Running Aqua@home on CUDA with the old drivers and no SLI causes even more sluggishness--I see it as a consequence of GPU crunching. I think these SLI CUDA support is still in the early stages, although the 190.38 drivers are official (WHQL) now.
 
Back
Top