Discussion in 'Folding@Home - Join Team Guru3D !' started by Julepalme, Jul 2, 2008.
Are running the SMP Client also ?
Nope, I'm holding off on installing that till my computer is fully setup.
Just to report, I had to plug in a second monitor to my second card to get it running, otherwise it would just sit folding the test protein. And this is on XP, is this normal as in the OP it says you only need to do this for Vista?
Fahmon is currently reading ~9000ppd if only I could leave my machine on all day
I've heard it was a vista 64bit issues.. may also be an issue for other people..
i got my single monitor hooked up with its 2nd input (vga) on my other card to make it work. otherwise i'd get loads of errors on it.
also you may need to extend a desktop too it aswell for it to be recognized
On my jolly Holls for a week,
shut down 2 of 3 pc's
happy catching up PPD
You had better bring a treat for all of us when you come back!
hm they need to fix a workaround, so that i don't have to plug in 3 monitors, i can run it indeed @ 3 cards a very nice start
What kind of speeds are you getting from each card ? I'm running two identical cards and one is getting around 1500 ppd less than the other .....
I tried setting affinity's so they have a seperate CPU feeding them - no difference.
Still getting tons more points but a little odd to see such a difference between two cards....
They are running different BIOS versions so maybe that is the difference....
Have you overclocked them? When SLI is disabled (which obviously has to be the case if multiple GPU Clients are supposed to work) then each and every card has to be OCed separately...
Hello, i have set this up and everything seems to be running perfect. fahmon reporting 4986PPD per card and it gets/finishes WUs fine, but for somereason its not actually counting towards my score. still only getting ~5k, seems like all of the WUs are getting lost.
copied this tutorial to the T and on my side everything seems to be running perfect, its just not adding up on stanfords servers.
anyone else experiencing this problem?
Yes but both cards are reporting the same clocks in GPU-Z - also I believe if you use Ntune to overclock like it do then there is no way to clock each card seperately - you can however adjust manual fan speeds independently.
Only wierd thing is that one card is reported as having 112 unfied shaders and the other 96 - whats wierder is that it if I switch them around GPU-Z reports whichever card is in the top PCI-E slot to have 112 shaders.......
Did you edit the client.cfg of the second client and give it a seperate machine ID ?
I don't think that is mentioned in this thread but it is on the foldingforums - might be worth a try
Thanks for the reply. both the GPU clients said nomachineid=2. so i changed gpu2 to nomachineid=3
im assuming stanford set the default to 2 since the CPU is 1, this correct?
Great finaly putting sli/ dual gpu to use, hitting ~11,500 ppd at gpu stock clocks.
Hmm... really strange indeed... What does it show when SLI is enabled?
You can always try RivaTuner- there you can set them independently. And while you're at it check your Unified Shader number using it too, GPU-Z is still pretty beta...
BTW: do you have to change the position of the SLI-Switch Kit on your Abit AN8-SLI every time you turn SLI on and off? Hopefully not...
Just stock speeds, im getting about 4400 per card, but as i said i have to plug in 2 more monitors wich i dont have.
I can switch 1 monitor and see it running on each card.
the other 2 cards are running that test protein
i dont get it, if there is a work around pls post it lol
or maybe wait for an update
I noticed something weird, when i leave the settings in the NV CTRL panel @ default, like 3 frames ahead rendering and vsync off, i get about 1800 iter/sec ...
But with 0 frames ahead and vsync on i get about 2430 iter/sec ...
Anyone else noticed that? is it folding faster this way?
Using the viewer to see iter/sec is generally a very bad way to assume performance for the GPU2 client..
basically the viewer takes up a lot of the drawing power of your card.. and then it reports much lower than it actually is.. sure you can see a general difference between cards.. but if your'e like me with a lil lag you'll see it go from 50k iter/sec when ya start it down to 2000-4000 when it starts rendering... its much better to leave the viewer off and use fahmon to monitor PPD or just generally look at frametimes
And also i dont think the settings really have any major impact on the folding.. just how much performance the viewer eats while on.. don't fold with it on
Julepalme, you have written only a few posts so far but you've been a great help already, to me especially by bringing up that you have to plug a monitor into the second (third and so on) graphics card in order to be able to fold on it/them (BTW the display doesn't even have to be plugged into the power socket). :smile:
For the next month or so I won't be folding due to my (not so well deserved but anyway) vacation. :O
You can send your PC to me and I will let it fold while you are AFK.
Have a good vacation!