2 GPU3 clients, same WU different PPD all the time GPU#1 vs GPU#2?

[Camaroz06]

Hi Guys,
I cant figure this one out. I have two GTX480's folding at the same time. Both are clocked at the same speeds, but one will fold faster than the other resulting in better PPD. I dont get it, there is no other program running besides the GPU clients. Shouldnt both put out the same ppd if the same WUs are downloaded?

 

[Nathan_P]

Hi Guys,
I cant figure this one out. I have two GTX480's folding at the same time. Both are clocked at the same speeds, but one will fold faster than the other resulting in better PPD. I dont get it, there is no other program running besides the GPU clients. Shouldnt both put out the same ppd if the same WUs are downloaded?

There is a lot of this happening at the moment. No one is exactly sure what is going on either. I do know that the card running your moitor will suffer a PPD loss. BUt that varies from rig to rig.

 

[Patonb]

My 260s have always done this.

I chaulk it up to differences in silicon... With the speed of gpus, 1 or 2 seconds in tpf can seem huge.

 

[TheMTZ]

I can't give a definite answer either, but i heard several times that people have differences in PPD even with 100% the same card + specs in one box. Maybe this has something to do how the traffic on the PCI lanes gets handled by the mainboard.

 

[Camaroz06]

I am pretty sure that my monitor is attached to the less ppd card. Could be the PCI lanes as well. I had always thought the X58 3x SLI had 2 16x slots, I guess it has one 16x and two 8x. I was looking at GPUid and it labels one card running at 8x.

 

[Patonb]

I have screens on both my gpus.... Saves on dummy plugs.

 

[jfb9301]

Dummy Plugs - no longer required......

 

[nitrobass24]

Hi Guys,
I cant figure this one out. I have two GTX480's folding at the same time. Both are clocked at the same speeds, but one will fold faster than the other resulting in better PPD. I dont get it, there is no other program running besides the GPU clients. Shouldnt both put out the same ppd if the same WUs are downloaded?

This is normal. No two parts are exactly alike.

I had two 2x E5530 systems.
Exact Same everything minus the case. Same mobo, ram, psu, hdd, hsf and same windows 2008 image. one of them would run about 1min faster on all WUs. In BigADV that is a huge PPD difference. Could never figure it out.

 

[sirmonkey1985]

Hi Guys,
I cant figure this one out. I have two GTX480's folding at the same time. Both are clocked at the same speeds, but one will fold faster than the other resulting in better PPD. I dont get it, there is no other program running besides the GPU clients. Shouldnt both put out the same ppd if the same WUs are downloaded?

its because windows requires one of the cards to render your desktop thus taking performance away from the card. theres no way around it. even setting windows to classic mode and aero turned off wont fix it. its something at driver level when having displays connected to the card. i lose about 450-500 PPD on my 8800GS's with 2 displays connected to it. same PPD loss with 1 or 2 displays connected so my assumption is that the drivers allocate a specific amount of gpu usage for video display. i know its not my cards since i've swapped both displays between cards and the performance drop follows the swap..

primary(1 or 2 displays) 5100 PPD secondary(no displays) 5600 PPD
primary (no displays) 5600 PPD secondary (1 or 2 displays) 5100 PPD

its just something you have to deal with.

 

[Patonb]

I have screens on both my cards and still get issues like this....

 

[sirmonkey1985]

I have screens on both my cards and still get issues like this....

well you don't need to anymore because nvidia fixed that in the 258.96 drivers. then re-added the sli cuda support in the 260.xx and later drivers.

but like i was saying i think the problem is at the driver level. why, i don't know...