Dual GPU Folding Question

[xConVirus]

Hey guys, so I recently installed a 750 Ti I had laying around into my PC alongside my GTX 980 for folding. For some reason the 750Ti is reporting a higher PPD, and faster completion times than my 980. This seems ridiculous to me. They're folding on the same F@H Core, 0x18 yet the 980 takes way longer. Any thoughts to why this is? I have a 5820K CPU. Also, the 750 Ti fails alot of work units with the error: Error downloading array interaction count, clEnqueueReadBuffer(-5) on 0x21. How can I fix this? I'm not sure what other information to give. Thanks for your help.

 

EDIT: It seems like the 750 Ti keeps trying 0x21 but then falls back to 0x18 and it works? Perhaps 0x18 is not designed to exploit the parallelization of the 980 and for some reason they are both syncing to that core?

 

 

 

EDIT2: Playing around with the settings more make it seem like my system or the software is prioritizing the "top" GPU in the folding slots. Here is a screen shot of the timings reversed. I haven't checked to see how much slower the 750 Ti should be compared to the 980, but does my troubleshooting help point to a certain issue? I'm not folding with any CPU cores, so why does it seem like it's prioritizing one GPU over another?

[Glenwing]

11 minutes ago, xConVirus said:

Hey guys, so I recently installed a 750 Ti I had laying around into my PC alongside my GTX 980 for folding. For some reason the 750Ti is reporting a higher PPD, and faster completion times than my 980. This seems ridiculous to me. They're folding on the same F@H Core, 0x18 yet the 980 takes way longer. Any thoughts to why this is? I have a 5820K CPU. Also, the 750 Ti fails alot of work units with the error: Error downloading array interaction count, clEnqueueReadBuffer(-5) on 0x21. How can I fix this? I'm not sure what other information to give. Thanks for your help.

How long have you been observing them?

[xConVirus]

2 minutes ago, Glenwing said:

How long have you been observing them?

I installed it just today, so it's been about 6 hours. They've each gone through 2 work units, and now the 750 Ti is in it's "Failed" state. Also, I saw a post earlier about the 373.xx NVIDIA driver being the most recent working one so I am on that driver revision.

[Imakuni]

For the misterious "750ti > 980", the answer is actually pretty simple: the F@H sucks at determining which GPU is which. Most of the time, it swaps out the order of cards; that's how I had a Gtx 970 doing tasks in 8 days and a Radeon HD 6650 running in a couple of hours. Don't worry about that, it's just a visual bug that will never be corrected (cuz y would it?).

 

As for the errors, that's beyond my knowledge.

[xConVirus]

20 hours ago, Imakuni said:

For the misterious "750ti > 980", the answer is actually pretty simple: the F@H sucks at determining which GPU is which. Most of the time, it swaps out the order of cards; that's how I had a Gtx 970 doing tasks in 8 days and a Radeon HD 6650 running in a couple of hours. Don't worry about that, it's just a visual bug that will never be corrected (cuz y would it?).

 

As for the errors, that's beyond my knowledge.

That honestly seems to be it. I checked out average PPD data for both GPUs and my estimated PPD is within margins of the correct PPD from the graphs, so it must just be displaying the progress incorrectly. Also, the reason it was failing was because Windows or GeForce Experience seems to have REINSTALLED NVIDIA driver 376.xx...so I used DDU and once again set it back to 373.xx (unless I never uninstalled it correctly in the first place *shrug*). Thanks you both for your assistance.