PSU expert anyone?


#1

Hi :smile:
I have my Titan running on 2 x EVGA 750W G2 80+ gold PSUs.

According to the miners own readings, it is using a total of 1050W.
That’s split between 4 cubes, not exactly equally but close - between 255W and 268W each.

That’s fine. Being 80+ gold PSUs I would expect the power usage at the wall to be 12 to 13 % higher than this.

I’ll cheat with the figures here because it makes it easier, but the principle should still hold.

Assuming a total power draw of 1050, split between 2 PSUs

1050/2 = 525W per PSU

525 x 1.12 = 588W per PSU at the wall

This should give a total draw (2 PSUs) of 1176W

My killawatt is showing 1550W to 1630W at the wall.

What might cause this ? Are the PSUs on their way out ?

Possible errors in my calculation are that I don’t know whether the controller power is included, but a RPi doesn’t pull 300W. I also have a large fan moving the hot air away, but this pulls about 30W and is not powered by the EVGAs anyway. I’ve tested its draw by unplugging it and power at the wall only drops by 30 - 31 W.

Any clues would be appreciated. I have some new cables on the way so that I can use another 2 PSUs giving each cube its own PSU. Hopefully this will bring the EVGAs usage down, but they should only be using about 70% capacity anyway.

Confused :stuck_out_tongue_winking_eye:


#2

Inefficiency of the PSU? It will dissipate some heat so you have to take that into account too.


#3

I thought that heat would come under the 80+ spec ?


#4

So the miners are cheating on the readout. Best way to see if that is true is to measure yourself between PSU and miner.


#5

Agreed. This wouldn’t be the first time a miner claimed to draw less power than it actually does. Notable examples would be…pretty much every miner ever.


#6

Yes, that appears to be it. I just used a crude induction ammeter and it looks like each cube is drawing 30A. Which gives 360W @ 12v per cube - 1440W for the miner. Add the PF in and that gives 1612W at the wall

Lying bar stewards. Still this is KnC we’re talking about …

I’m still going to put on 1 PSU per cube, that should bring the PF factor into the 0.9 range and save a bit of power.


#7

I haven’t run the calculations but a quick glance at your numbers suggests that investing in good Platinum PSU(s) would be worth it if you can get them at a decent price. i.e. the differential between PSUs in efficiency should more than make up for the increase in cost if you can get a good deal on them and intend to run the miner for an extended time frame.


#8

Ok, I’ve got my new leads and gave each cube it’s own PSU. After roughly an hour, it appears that power usage is down by around 90W in total. Not a huge amount, but significant.

I removed the large air moving fan from the killawatt, so that accounts for 30W, leaving a saving of 60W from the PSUs.

I also didn’t previously factor in that I am overclocking the cubes, but this only appears to add 60W in total to the power usage. So if they were not overclocked and running a PSU each then I would save maybe 120W.

I’m getting an increase of around 30 to 60 MH/s by overclocking, so I feel it’s worth the extra there. I also have solar power, and it appears that 3 hours of decent solar a day covers the overclocking usage anyway.

One immediate advantage is that I can turn all the PSUs on to eco mode so the fans rarely run. In fact the AX760 hasn’t spun up at all yet. This may account for the drop in power somewhat.

The PSUs efficiencies now should be right around 48% which on 230V puts them right in the sweet spot :smile:

Don’t think I can do much more without underclocking, so this appears to be it for KnC Titan power usage. 1560W when running at 325MHz per die.