Topic: Power supply question 240volt 3phase power vs 1 phase (Read 4869 times)

philipma very nicely done on the install.  Looks great.

Thanks for help aswell on 240, hopefully I will finnaly get one in next week or two.

You will want the meter to see what is going on at the level of each phase.  If you are going to push what is available at the location, you will want to see what each of the phases are pulling at the service entrance.  You know, if you have a 100 amp service, make sure you keep it at 80 amps for each phase, that kind of stuff. 

Personally, I have a lot of machines and it would be impossible to see the big picture at the service entrance unless you measure it there.  If you just have a few machines, no need for the meter. 

 the pdu's have built in amp meters so I will attempt to have each one at 4  

I believe using 1 of the 3 phases is fine (along with a neutral and a ground), but using 2 of the 3 phases may cause bad things to happen. In a normal 2 phase configuration, the 2 hots are orientated 180 degrees from each other. But if you use 2 phases from a 3 phase line, the orientation is different, being in a 120 degree / 240 degree rotation. This causes the net difference in voltage between the 2 phases to be much lower, and I have concerns that the circuitry inside the power supply only expects a certain waveform of net voltage.

Ninja edit: After some shenanigans with wolfram, turns out the waveform of 2 phase power coming from 3 phase is exactly the same as a normal 2 phase coming from 2 phase with just different net voltages. Assuming the peak voltage of 240v 3 phase is 240v between all 3 phases, each phase is 120v. Putting 2 of the phases as hots in an ATX power supply, you have 208v @ 60hz. So what you are proposing is perfectly fine. Just make sure your PSU input looks something like 100v-264v. If its 115/230v, bad things might happen.

I'm electrically stupid. But not that stupid.

Also grab one of those amp meters you can put around each of the phases (to see how many amps are being used).  They will allow you to answer weird questions like say you are using 10 amps on each phase, and you add another 5 amp psu (using say 1200 watts) to 2 of the phases or 1 phase and neutral.  Does that increase each leg by 5 amps, 2.89 amps, or 2.5 amps?  5 minutes of testing and it will all make sense, have fun, 3 phase is good stuff!

that is the plan

Generally, a 240V 3 phase system will be configured as a delta, while 120V/208V systems are configured as a wye. Either way, you just want to balance the load across the phases. Run the PDUs in multiples of 3, and try to keep the draw on each even.

Forgot to add there are two main forms of 3 phase called 'delta' or 'wye'.

Here is a place where they get geeky about the sematics: https://www.physicsforums.com/threads/delta-vs-wye-pros-and-cons.682676/

TL;dr: wye will have 4 wires instead of 3, the fourth being a neutral, so if you tie any hot to a neutral you get 120v, and there is no 'wild leg' (all hots carry 208v).

All the wye stuff I have seen is 120/208/277/480v, take a look at this chart to see whats out there: http://www.ccontrolsys.com/w/Electrical_Service_Types_and_Voltages

Yes, they are loud. Not for the faint of heart. Sloopy claims he sleeps with them, but I'm not convinced he actually sleeps. LOL!  Yes, J4bber has already been successful with a fan design that keeps noise lower than an S3 running.  I just received a couple PWM fans this weekend that I'm going to rig up and see both how much lower I can get the noise while still supporting full 2880W. I'll let J4bber comment further on his design, but I am excited for allowing more people to take advantage of them for still quite reasonably.  If only they could work on 120V, we'd be in great shape.

Sloopy, it's all good.  Again, I've always known 240V is superior for PSU's in almost every way, but I always thought the DC output voltage was stable regardless of input.  If it does have a positive effect on the hardware, that's an even bigger reason for people to make the switch.

Phillip, keep us updated.

Isn't it pretty loud over there on the dark side?   

I haven't seen a server power supply that even pretends to be quiet. The ATX EVGA supplies seem pretty quiet to me. That's only based on a sample size of one. My bet is that by now, all the server PSU designers are hearing impaired from the noise.

Do not be so fast to assume Phillip's gear did not have a gain in performance.
If he says it did be more optimistic.
I have seen this first hand and there are examples posted not only on this forum but all over the internet.
It is easily explainable by confirming the unit's were not receiving "enough" voltage and amperage prior to the move
Or
The units have specific design qualities which utilize the new scenario.

Phillip has shown his knowledge of the miner side. Just because he asks a few questions on the 3 phase does not mean he presents data without factual basis.

For a real world example imagine you under power a lot of S3s with a power supply which regulates the output evenly (I say evenly, but it regulates somewhat without bursting into flames across all S3s), even when "overdrawn", IE the 2880. You can load that baby up with significantly more S3's than you should and she happily hums along lowering the output to each. Most of the S3's may run at 400 a couple at 440, one or two at 390, but they run. Performance is impacted.
Remove enough S3's and alacazam, you get 450 and 500 out of your poor little beasts.


Edit for safety:
Do not overload any power supply, it simply isn't safe and most go pop. (Same for any power source, calculate your supply VS your draw, add 20% safety factor)
Power cycle after any significant change to any electronic device
Keep your hands away from anything with live voltage, IE do not connect live wires to miners.

Your Mileage Will Vary, depending on several things.

Don't waste your money on the EVGA's Phil, come to the dark side. 2880 Virgin Watts await you in the kingdom of 240 volts... 


Sloopy,

I'm not doubting his claim, I was asking if he had come up with any reasoning, since AC input voltage theoretically should not greatly influence DC output voltage, which is why I asked if he measured the output to see if there was a difference.