PPD by GPU - Advice for Dedicated F@H Machine

[Aurum]

Hmm, guess I'll have to learn how to overclock. I just run them at stock settings assuming they'll last longer that way.

[ryoungblood]

Fixed my misread, but that sheet is incorrectly applied by most everyone that ever sees it.
Everything on that sheet is estimated PPD based on an entire day of folding a specific WU (which will never happen in real world usage).

Okay. Well, that being said, I still get pretty high PPD on one of my 1070. Is it unrealistic to assume the other 4 are able to output that much?

[ComputerGenie]

...Turing has 2x ASUS 1070 Strix, Zuse has 2x EVGA 1070 SC Black, CMD has 1x 1080 FTW Hybrid and 1x EVGA SC...

OK, so if I'm reading this right, this time...

That's 5 1070s and 1 1080.
Now, assuming you're hitting ~800k for the 1080...
You're at an average of ~591k for the 1070s, which is just a touch under the 600k mark that most folks say they average.

[ComputerGenie]

...I still get pretty high PPD on one of my 1070. Is it unrealistic to assume the other 4 are able to output that much?

Is that card always significantly higher than the others? If so, is it still higher if placed in one of the other rigs?
Are you comparing actual daily totals or just looking at a specific PRCG (like in the image you posted)?

[ryoungblood]

Yes, it is consistently higher. I will swap it out with the card with the lowest PPD output and report back.

The highest performing card is CMD Slot 2, it is the EVGA 1070 SC.

[ComputerGenie]

Again, to avoid any confusion, are your PPD numbers actual points received or are they theoretical numbers?

[ryoungblood]

Also, to note, Turing is the new machine and is on an AMD FX-8320E @ 3.2GHz. Zuse is the oldest machine and is on an i5-2500k @ 3.3GHz. Could a CPU (or any other hardware) act as a bottleneck, even if not fully utilized? All machines have roughly the same CPU utilization (40-60%) while folding.

[ryoungblood]

Again, to avoid any confusion, are your PPD numbers actual points received or are they theoretical numbers?

I'm using data output from HFM.NET. I'm not sure if the PPD output is estimated or actual points received.

[ryoungblood]

For actual numbers, here is the data on credit received.

CMD Slot 2 still appears to be an outlier.

[ryoungblood]

Actual PPD.

[Aurum]

All cards are on a x16 slot.

Being in a x16 slot does not mean it's fast. Use GPU-Z and see what bandwidth they're running at in the "Bus Interface" box:
https://www.techpowerup.com/gpuz/

What MB models are you using?

BTW, if you place img inside url HTML code it doesn't link. If seperate we can see your screenshots enlarged. Thx.

[ComputerGenie]

All cards are on a x16 slot.

Being in a x16 slot does not mean it's fast....

That was in reference to my question about slots (insuring that it wasn't an issue of comparing x16 values to x1 values, or something like that)

[ComputerGenie]

Actual PPD.
img removed for space

Looking at those numbers (and going but what others have posted about 1070 averages), I'm more inclined to think that the 1 rig/slot/card is just "lucky" (for lack of better words) and the others are about on par. Unless I'm missing something....

[ryoungblood]

CMD - PCI x16 3.0 @ x8 3.0 - Bandwidth @ 131.8 GB/s
Turing - PCIe x16 2.0 @ x8 2.0 - Bandwidth @ 263.7 GB/s
Zuse - PCIe x16 2.0 @ x8 2.0 - Bandwidth @ 263.7 GB/s

[ryoungblood]

Actual PPD.
img removed for space

Looking at those numbers (and going but what others have posted about 1070 averages), I'm more inclined to think that the 1 rig/slot/card is just "lucky" (for lack of better words) and the others are about on par. Unless I'm missing something....

Well, I appreciate your help, I feel better knowing my numbers are in line with the reported average. I guess I just got lucky with the one out-performing 1070.