TR Forums

just brew it! wrote:

Waco wrote:

just brew it! wrote:

That said, if we assume 12 cents per kilowatt hour (the US national average) and 250W for the idle power draw of the server (wild-ass guess), that works out to 72 cents/day, or $21.60 per month. So yeah $50 seems kind of high, but not outrageously so.

250 watts idle? Jeebus, it's not 2003!

He did say "older"... and that it looked like it was using $50 worth of power (I assume he meant per month). If he meant per year, then never mind.

For all I know it's got a RAID array of 10K SCSI drives and a half dozen leafblower-level fans, which would push the power usage way up. My dually Athlon MP idled at around 150W, and that was without any storage devices. (I measured it because I was thinking of firing it back up and re-purposing it as a server; decided not to when I saw the power numbers.)

just brew it! wrote:

I doubt he'd be complaining if it was only $50/year though...

Ding ding! One would hope $50 a year wouldn't be a big issue....But we're looking at just over $50 a month per my meter https://www.amazon.com/P3-International ... ater+meter with the rates here in the midwest. I'm currently paying just under 14¢ for all the surcharges, but the test was done before I moved with different service, which I believe was a little cheaper.

It's a Poweredge T110 I believe, with leafblower level fans. It has a PERC Card, and 2 128GB drives I believe. I picked it up free to tinker with and haven't run it in several months as I moved my office space (with pretty LED trim lights I might add), and been getting resettled and trying to figure out what I want to do. Main rig is a monstrous Lian-Li (older version of this http://www.lian-li.com/en/dt_portfolio/pc-a76/) I think I want to gut and put in some cheaper gear, raid, and drives to act as a server for all my goodies. Probably waiting for CPU refresh later this year to build myself a new PC (also 2008 era main rig, see other post somewhere).

I wouldn't trust your meter then. A PowerEdge T110 has 305W rated power supply. I don't think it's possible for that system to burn $50 of power in a month, at the prices you state. I would be surprised if it was over 150W at idle.

--SS

SecretSquirrel wrote:

I wouldn't trust your meter then. A PowerEdge T110 has 305W rated power supply. I don't think it's possible for that system to burn $50 of power in a month, at the prices you state. I would be surprised if it was over 150W at idle.

Yeah, even if you assume the thing is running at the max capacity of the PSU (and that the PSU is only 70% efficient at full load), that still gets you to only $44/month. Something doesn't add up.

SecretSquirrel wrote:

I wouldn't trust your meter then. A PowerEdge T110 has 305W rated power supply. I don't think it's possible for that system to burn $50 of power in a month, at the prices you state. I would be surprised if it was over 150W at idle.

--SS

Yeah, something is off there. Even with it running flat out, it couldn't use that much power.

My guess is that it's a relatively cheap power supply with poor or no power factor correction and the meter is a "cheap" power meter that is interacting very poorly with the power waveform generated by the supply. If the meter simply measured peak voltage during the voltage cycle, peak current during the current cycle, calculated the phase between the two based on the timing of the peaks, and then did some math to calculate real power, then it would way over estimate real power if the current waveform was "distorted" due to poor PFC.

--SS

P3 Kill-A-Watt meters do have the ability to compensate for power factor. That said, I don't know how accurate that compensation is.

just brew it! wrote:

P3 Kill-A-Watt meters do have the ability to compensate for power factor. That said, I don't know how accurate that compensation is.

Yeah, my thought is that it's not the power factor that's the issue, it's that they are assuming the current waveform is a sine wave like the voltage waveform. In other words, they are doing linear power factor calculations rather than non-linear. Just accounting for the phase difference between voltage and current and not integrating the actual current waveform. Way less math so much less stress on a small micro. Just WAG though.

--SS

SecretSquirrel wrote:

Yeah, my thought is that it's not the power factor that's the issue, it's that they are assuming the current waveform is a sine wave like the voltage waveform. In other words, they are doing linear power factor calculations rather than non-linear. Just accounting for the phase difference between voltage and current and not integrating the actual current waveform. Way less math so much less stress on a small micro. Just WAG though.

Want to try and pin it down further? See if you can decode this (a P3 metering board):

Metering board:


Source (wall) side:


Load (front) side:


Approximate Schematic:


I got as far as noting that current is apparently being derived by I²R on the neutral resistor R2 (which, by the look of it, must be in the fractional-ohm range), and I'm not sure if an RC time constant is being provided as a second voltage reference with a known phase delay, or if that's just a cheap way of providing isolation to the ASIC board's power supply without using a transformer.

ludi wrote:

... I'm not sure if an RC time constant is being provided as a second voltage reference with a known phase delay, or if that's just a cheap way of providing isolation to the ASIC board's power supply without using a transformer.

Looks like just a cheap-ass way of providing a DC power source for the rest of the unit to me (the diodes are a big clue, as well as the fact that they seem to have set it up to yield a positive DC voltage referenced to the neutral lead), but I could be wrong.

Aside from that... all of the details are buried in whatever's at the other end of the cable connected to J1. J1 has the raw current and voltage waveforms; who knows what they do with them. The analysis could be reasonably sophisticated, or it could be cheap-n-nasty.

Yeah, what JBI said.

Pins 5 and 6 will provide a voltage that represents the current being pulled. The voltage drop across R2 will be fed to an op amp and amplified before being sampled. Pins 1 and 2 will give a voltage that can be divided down and sampled to get the line voltage. Pins 3 and 4 give DC to the rest of the circuit, probably filtered and regulated to some degree on to other board.

No way to know how the current and voltage are sampled.

With a decent scope, monitoring the waveform on the current and voltage sense lines, you could make a good guess though. Otherwise you need to borrow some high end measurement gear.

--SS

That said... with as cheap as microcontrollers with integrated A/D converters are, I really feel like it should be possible to "do it right" with less than $5 in parts (possibly a LOT less, given the volume of Kill-A-Watt meters sold). Taking 10-bit samples of the waveforms at a few kilohertz should be enough to get pretty good accuracy, and this is well within the capabilities of a cheap microcontroller. So it really all comes down to whether the person who wrote the firmware for the micro had a clue...

just brew it! wrote:

So it really all comes down to whether the person who wrote the firmware for the micro had a clue...

Wasn't the supplier supposed to do that ...?

Captain Ned wrote:

just brew it! wrote:

So it really all comes down to whether the person who wrote the firmware for the micro had a clue...

Wasn't the supplier supposed to do that ...?

Err... I'm confused. Supplier of the microcontroller? No. It's typically up to the designer of the device it goes into to come up with all of the firmware.

just brew it! wrote:

Err... I'm confused. Supplier of the microcontroller? No. It's typically up to the designer of the device it goes into to come up with all of the firmware.

Oh dear, I have trodden in monsieur's sarcasm tag bucket.

I should know better.