Register for the iXsystems Community to get an ad-free experience and exclusive discounts in our eBay Store.
Resource icon

Proper Power Supply Sizing Guidance

Ericloewe

Not-very-passive-but-aggressive
Moderator
Joined
Feb 15, 2014
Messages
16,048
Thanks
3,894
Generally, the default for many/most BIOS's that have it, is 3 seconds. If it is the default, it is pretty hard for it to "stop working." I have never seen it "stop working," but I am just one data point around many systems from many vendors.
You're going to need to provide evidence that most systems do staggered spinup, because that it something that I have never seen. LSI HBAs do support it, but I can't think of any BIOS that does.
most of the large computer vendors do not size that way (guessing)
There are no guesses involved here. And if (big if) other vendors don't size PSUs like this, it's because they ship full systems that are tightly integrated. That's not the case here.
To suggest empirical measurements are better than actual measurements
You do not seem to know the meaning of the word "empirical", let me help you (from Wiktionary):
Adjective
empirical (not comparable)

  1. Pertaining to or based on experience.
  2. Pertaining to, derived from, or testable by observations made using the physical senses or using instruments which extend the senses.
  3. (philosophy of science) Verifiable by means of scientific experimentation.
The 30W is not random, as is explained in the OP. The measurements are also available, linked in the OP.
To randomly say, "You should never rely on staggered spinup." just said, "Do not trust BIOS and many of the settings can randomly change on you."
Yes, I did. They're volatile and BIOSes are easily the buggiest portion of the whole system (by bugs per line of code).
No one says that and we should not say that here, unless there is a BUG in someones BIOS that has not been fixed and people need to be notified.
Are you kidding? Everyone these days uses AMI and it's buggy as all hell. It's a miracle computers boot at all and I should know because the ASRock board in my workstation literally refuses to boot any OS with the latest BIOS. Not my idea of bug-free code.
 

Ericloewe

Not-very-passive-but-aggressive
Moderator
Joined
Feb 15, 2014
Messages
16,048
Thanks
3,894
I think the term empirical is being mistaken for anecdotal.
Feel free to replicate the experiment and post your results.
 
Last edited:

Ericloewe

Not-very-passive-but-aggressive
Moderator
Joined
Feb 15, 2014
Messages
16,048
Thanks
3,894
Ok, sorry if I came across as somewhat rude, this past week has kept me a bit on edge with the whole Corral thing.
 

farmerpling2

FreeNAS Experienced
Joined
Mar 20, 2017
Messages
224
Thanks
73
Finally got around to following up to this thread.

So here are the facts, not just a bunch of guesses.

My research shows that 24 watts is more than enough per drive and even less for many models. Anything over that is a waste of money on PSU.

Enterprise drives are hard to find documentation on max startup watts. I list what I found.

Interesting tidbit of data - Seagate Enterprise (7200 RPM, 21.6 watts) drives use less max startup power than Seagate IronWolf (5900 RPM, 24 watts) and Seagate IronWolf Pro (7200 RPM, 24 watts). If you like Seagate, the Enterprise model should be looked at because of the better quality of materials, for longevity sake.

WD Red NAS Home 1TB peak usage is 14.4 peak watts. Not sure why this is so low compared to others in same series. Maybe an outlier or documentation issue possible?

I dug around a bunch of disk drive manufacturers for power usage at startup. I updated the spreadsheet with this info so people can see what the manufacturers claim. Seagate provides a strip chart listing in their documentation for the first n seconds of startup so you can see the power usage at various stages. Others just list peak power usage.

Only one manufacturer uses peak startup of 30 watts - HGST. Not commonly used in any home NAS system unless you run 7200 RPM drives from HGST.

Everyone else uses 24 watts or less. The two most commonly referenced home NAS drives are WD Red at around 21 watts or less, Seagate at around 24 watts or less.

Values are in watts. 51 drives NAS and non-NAS went into these numbers.

Code:
MIN	   14.40
AVG	   23.32
MEDIAN	23.28
MAX	   30.00


These numbers have been consistent, in my experience, for a good number of years and should not be surprising to anyone in the industry who paid any attention to them.

As I said before, using 30 watts* or higher per drive to calculate PSU size is a waste of money.

* If you use HGST drives, then do use 30 watts - They are an outlier.
 
Last edited:

farmerpling2

FreeNAS Experienced
Joined
Mar 20, 2017
Messages
224
Thanks
73
...
So here are the facts, not just a bunch of guesses.
...
As I said before, using 30 watts* or higher per drive to calculate PSU size is a waste of money.

* If you use HGST drives, then do use 30 watts - They are an outlier.
Updated on 75 drives...

Code:
MIN	 4.50
AVG	20.00
MEDIAN 21.60
MAX	31.20
 
Joined
Apr 26, 2016
Messages
37
Thanks
5
Just wanna check my calculations before buying a new PSU, I got about 924w peak (1155w when multiplied 1.25x) from this system: 24 HDD + 2 SSD, dual 80w 2011-3, dual sock motherboard, 4 DIMM's, LSI HBA & 7 fans (5 case, 2 CPU).

So if I'm right that would be a 1200w PSU and I'd need dual 8 pin CPU and 7 Molex, any recommendations? :)
 

Chris Moore

Super Moderator
Moderator
Joined
May 2, 2015
Messages
9,362
Thanks
2,997
Just wanna check my calculations before buying a new PSU, I got about 924w peak (1155w when multiplied 1.25x) from this system: 24 HDD + 2 SSD, dual 80w 2011-3, dual sock motherboard, 4 DIMM's, LSI HBA & 7 fans (5 case, 2 CPU).

So if I'm right that would be a 1200w PSU and I'd need dual 8 pin CPU and 7 Molex, any recommendations? :)
Sorry, I am just now seeing that you are in Australia. It is a difficult call, do you have anyone down under that sells computer hardware?
If you could get one of these, it is a great power supply:
https://www.newegg.com/Product/Product.aspx?Item=N82E16817151208
 
Joined
Apr 26, 2016
Messages
37
Thanks
5
Joined
Jul 15, 2018
Messages
79
Thanks
2
A friend of mine is calling me mad for having a 450w PSU planned,
according to this thread its fairly on point, but coolermasters calculator seems to confirm my friends notion.

My setup;
board: Asrock C236 Rack
cpu: Intel i3-6100
ram: Kingston 1x16Gb ECC valueram
drives: 4 x WD green (flash tweaked), 2x WD Red, 1x SSD
Three lowspeed case fans, and a big CPU cooler fan.

Can someone tell me what amount of wattage is correct for me?
 

Bidule0hm

Server Electronics Sorcerer
Joined
Aug 5, 2013
Messages
3,710
Thanks
649
A 360 W PSU will be enough but if you plan to expand later then a 450 PSU is probably a better idea.
 

Ericloewe

Not-very-passive-but-aggressive
Moderator
Joined
Feb 15, 2014
Messages
16,048
Thanks
3,894
450 W is a good value because there are several good models available that are at least semi-modular.
 
Joined
Jul 15, 2018
Messages
79
Thanks
2
I went with a gold Bitfenix 450 instead of the bronze corsair C450.
Friend said the added efficiency will pay itself back in a year,
plus its one of the rare few lowend PSUs that comes with 8 SATA power connectors,
so I don't have to screw around with molex converters.
Plus 7 year warranty, thats nice too.

Was surprised to find coolermaster recommending 238w though.
Is there something inherently different for NAS calculations?
 

Ericloewe

Not-very-passive-but-aggressive
Moderator
Joined
Feb 15, 2014
Messages
16,048
Thanks
3,894
Most calculators are crap, out of date, or both. That said, 238 W is not too far off from the minimum for your system. It'll definitely idle well below that, but the problem is disk spin up - I like to use 30 W per disk as a good target, plus whatever the rest of the system needs.
 
Joined
Jul 15, 2018
Messages
79
Thanks
2
What I've seen from real life tests is that my setup would generally be hovering around 80 to 90 watts in idle and normal workload. Peaks can occur around 125w. But like you say, that initial cold start is what does it in, probably.

I've looked at tests for this bitfenix PSU, it seems optimal efficiency occurs in 2 distinct draw zones. One at around 90% load, the other around 20%. And 90W happens to be that 20% sweet spot of the 450 supply, so it should do nicely.
 

Evi Vanoost

FreeNAS Aware
Joined
Aug 4, 2016
Messages
91
Thanks
27
If you're truly interested in energy savings, use SSD. The prices for 2TB aren't that bad any more , with some of them hitting sub-$300, about $100-150 difference from a spinning drive.

In use, 5 spinning drives cost about 100W, SSD about 5-10W and SSD will idle a lot more (since they are done faster). In the US that will save you $100-130/year and over 5-10 years you've basically saved out the cost. You're also immediately saving $50-100 on a much cheaper PSU since SSD don't have a spin up power draw.
 
Joined
Jun 29, 2015
Messages
55
Thanks
3
and when they die, all your info is GONE ! hence why my nas will likely NEVER have them, that said freenas does safegaurd the best against that and they are quicker to resliver.... however for the prices i see i would go 4 tb sdds by and far .....
 

Evi Vanoost

FreeNAS Aware
Joined
Aug 4, 2016
Messages
91
Thanks
27
@Mega Man: that's also true for spinning rust though. I manage several FreeNAS with ~200 spinning drives right now and ~50 SSD. I actually only had 3 SSD die in the last 5 years and I send back a spinning drive to its manufacturer pretty much every other month (~4-5 per year).

This is very much in line with the Google/Backblaze data, SSD's seem slightly more reliable but there is never a guarantee. Hence the need to use RAIDZ2 or Mirrors.

But when planning out the datacenter, we've noticed that for the last few years SSD save long term in both space and power, especially for database loads and it's approaching the same for large disk storages (up to ~2TB per spindle for IOPS reasons). For archival data, spinning disks that can be turned off or tape is still cheaper.
 
Joined
Jun 29, 2015
Messages
55
Thanks
3
backblazes data was very much compromised and has been a proven farce. - they shelled 4tb drives and used them in non-purpose built environments ( read: they used external 4tb drives in data centers, after shelling the " junk " external part ) and then whined when they failed ... unless i am getting the names wrong but pretty sure i am not.

i have NEVER lost a drive in 15 years,..... except the very first i lost, when i learned 5 year data life, then throw away, get new, repeat .....
only other drives that failed are ones i knew were failing but don't care ( IE i buy a lot of refurb drives for small meaningless boxes like HTPCs, that can be replaced at a whim.... ) because i r chep!!!!!
 
Top