Should I upgrade my power supply?

[Evi Vanoost]

You don't want to size your power supply based on idle power usage, you'll never get to the point of idle if it doesn't boot up.

But given your total peak power consumption is ~250W, a "good" 450W should suffice, the biggest consumers will be GPU and CPU, something like an nVidia Titan will kill that poor box though. Drives are ~10W, 1-5W for an SSD, swapping out one CPU for another is usually zero sum.

Very small boxes containing low end CPUs like the one the OP posted fit up to 2 or 4 drives with 250W power supplies.

 

[Stux]

The safe sizing for HDs is to allow 36W per HD for spin up plus all the other components.

 

[joeschmuck]

allow 36W per HD

Seems a bit high to me but it can be a rough guide.

A WD Red 8TB drive (picked a heavy drive) runs at (21.6 Watts + 6.4 Watts =) 28 Watts maximum at any given point in time. The 21.6 Watts is for the motor and this surge will not be there for long, but it is a significant surge. I don't believe the 6.4 Watts would be present during the spinup but it would not exceed that regardless. The 10TB drive actually pull 6.2 Watts, must be better read/write electronics.

While this was an example of the WD Red lineup, other drive manufacturers will have different values. It would be important to look at those specs while sizing a power supply, and also factor in future expansion. It's okay to have extra capacity and a sin to not have enough. If you attach too much load to a power supply, you will get dirty power and the power supply will hopefully fail without harming any other components.

 

[danb35]

A WD Red 8TB drive (picked a heavy drive) runs at (21.6 Watts + 6.4 Watts =) 28 Watts maximum at any given point in time.

I seem to recall @Bidule0hm coming up with some higher results for a number of drives, though my own results didn't agree with his.

 

[joeschmuck]

I'm thinking that your values are likely more accurate. I'd need to see a schematic that @BiduleOhm used to measure current but a DC Amp Clamp is a very basic tool to use, making one yourself can lead to inaccuracies. I have I think 5 of then for low currents and another 4 for very high currents (10,000 A). And a variety of AC Amp Clamps. All of which are calibrated every year. If I could bring an O'scope home and a few probes, I'd take measurements myself. I can't bring anything in the building due to security concerns.

The results I provided above were from the WD Spec Sheets. I would hope that they didn't make them look better than they actually were.

 

[Dice]

I would hope that they didn't make them look better than they actually were.

...Well wouldn't we say the same about car manufacturers regarding emissions? :P

OOOOOOOPPS they did it again *insert Britney Spears whine*

 

[danb35]

I'd need to see a schematic that @BiduleOhm used to measure current

https://forums.freenas.org/index.ph...drive-spin-up-peak-current.38885/#post-237511

 

[danb35]

...Well wouldn't we say the same about car manufacturers regarding emissions? :p

As the current or former owner of seven VW/Audi diesels, I know for certain that automakers are scrupulously honest in such matters.

 

[joeschmuck]

https://forums.freenas.org/index.ph...drive-spin-up-peak-current.38885/#post-237511

Doh, just scroll up some.

A 20 milliohm shunt resistor, that's not a common part to have laying around in the workshop, well I've never had one. Also the measurement was taken on the ground leg. While I'd normally agree with that, I think taking two separate measurements for each voltage line would have been better and would show both legs. This is not a calibrated circuit in the way it was connected.

I Feel like we got a bit off topic :D

 

[danb35]

A 20 milliohm shunt resistor, that's not a common part to have laying around in the workshop, well I've never had one.

No, but it's easy enough to fabricate if you have about a 1' length of Cat5 cable.

Also the measurement was taken on the ground leg.

Necessary in order to be able to measure with a scope (Edit: unless you have a battery-powered scope). You could do it inline with either of the power lines using a DMM, but of course that wouldn't give you waveforms (and if you wanted to use a DMM, you wouldn't need to build this circuit to begin with, though it might reduce the burden voltage).

I Feel like we got a bit off topic :D

Maybe a little--but it does go to the key question of how much power is actually needed. @Bidule0hm and I never resolved our differing results, though, so I'm not sure how much value is there.

 

[Stux]

No, but it's easy enough to fabricate if you have about a 1' length of Cat5 cable.


Necessary in order to be able to measure with a scope (Edit: unless you have a battery-powered scope). You could do it inline with either of the power lines using a DMM, but of course that wouldn't give you waveforms (and if you wanted to use a DMM, you wouldn't need to build this circuit to begin with, though it might reduce the burden voltage).


Maybe a little--but it does go to the key question of how much power is actually needed. @Bidule0hm and I never resolved our differing results, though, so I'm not sure how much value is there.

Which is why the "safe" value to use is 36W. No one has ever measured an HD exceeding that.

If you want to be conservative, multiply your drives by 36W add all the other components and buy about that size of PSU, done.

Ie HBA, CPU TDP, motherboard, ram and fans.

 

[joeschmuck]

I may try the DMM method this weekend. If I can, I'll see what I can do with a true DC Current Probe and O'Scope but I'd like to use a common NAS hard drive for the testing. We got some new hard drives in at work, but are they common NAS drives or more datacenter fast drives is the question. If I ever get this done, I'll provide data, just don't hold your breath.