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

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,782
Thanks
3,036
See, that's the thing about this forum. Lots of awesome geeks doing lots of cool things. Quite frankly there just aren't enough hours in the day any more for me to spend time figuring out what's a reasonable DSO...
 

Ericloewe

Not-very-passive-but-aggressive
Moderator
Joined
Feb 15, 2014
Messages
16,027
Thanks
3,881
2000-3000 bucks gets you a "pretty nice" entry-level Keysight with 2-4 channels, 70-200MHz bandwidth, nice, big screen and a a number of cool features.
Not sure how good their absolute-low-end stuff is, but the 2000 Series is pretty good and the computer interface is quite usable.

To be honest, I have no idea how they justify the ludicrous price difference to Rigol and similar, beyond "We're the OG HP".
 

Bidule0hm

Server Electronics Sorcerer
Joined
Aug 5, 2013
Messages
3,710
Thanks
648
Are the USB "hook it up to a computer" ones any good?
Usually they are total crap, don't buy them, never (think something like Realtek vs Intel... :D). If you want a cheap but high quality DSO look at the Rigol ones. I have the DS1052E hacked to be a DS1102E (exactly the same scope but with a 100 MHz BW instead of 50) and I'm happy with it, the biggest downside is that it's only a 2 channels one, and 90 % of the time you really want a 4 channel scope, trust me. It's responsive (I HATE slow interfaces...) and pretty much bug free (the only thing is that sometimes the USB stick isn't recognized so I need to unplug/replug it but I guess it's just USB doing USB crap... not a big deal). The 1054Z is actually the best bang for buck but is more bugged, I still recommend it over the 1052E though.

The price difference comes from cutting corners but not on important things (well, almost, the 1052E/1102E use an overclocked ADC but it works very well, so why not after all...). The EEVBlog youtube channel has done some great reviews and teardowns if you want to know every details, there's also some threads on the EEVBlog forum with a lot of posts of very knowledgeable peoples ;)
 

Ericloewe

Not-very-passive-but-aggressive
Moderator
Joined
Feb 15, 2014
Messages
16,027
Thanks
3,881
Aw, come on, that was funny.
It would be, were it not for the fact that it sounds credible enough to conjure up thoughts of a massively anticompetitive, stagnant hard drive market...

On second thought, we're kinda there already.
 
Joined
Jun 29, 2015
Messages
55
Thanks
3
Actually, the problem here is that the average FreeNAS user has no way to identify "a quality PSU," especially when buying one online, sight unseen, for a new build.

The maker/OEM does matter in this business. It isn't an absolute. Look at automobile reliability. I notice that Toyota isn't on this list of 10-worst-cars. The Prius C is at the top of the list, though. I can draw some inferences that Toyota tends to produce a better quality car, while Ford tends to produce a crappier car. Not absolutes, but does contribute. I'd say at least half the supplies I've seen fail over the past ten years trace back to Fortron. Fortron is the manufacturer who actually makes supplies sold under the Antec, Thermaltake, Sparkle Power, etc., labels.

The problem is that ANY manufacturer CAN make a great supply, and those manufacturers can also make crappy supplies. The reality of the PC parts marketplace is that end users usually aren't willing to pay the price premium for extreme endurance supplies built with milspec components, so the question becomes "which corners did they cut" and "are those corners safe to cut."

There's lots of discussions of the various details on PSU's, but to actually know, rather than guess, you need to rip the damn things apart. http://forums.anandtech.com/showthread.php?t=2297245 for example is an entertaining thread.

So we don't really care what MIGHT work. Anything MIGHT work. When looking at the options for something solid to suggest to users, I wanted something that came as close to "uncompromising quality" as I could find. As far as I can tell, the Seasonics suggested are that.
sorry for the delay but i wanted to respond with more then a 1 liner ~

i must respectfully disagree with you. i feel that the whole section on hardware in this forum, esp hardware recommendations are not there because you should buy this, but i feel that they are there so that we can make informed decisions -- now i was just reading on another thread in this forum, where one states, simply, that most people don't, which is true- however that does not mean that it should be advocated to not make informed decisions

you can easily find if a unit is quality imo. there are several great sites ( see below ) that do the hardware reviews of psus - a few key things - ( not speaking at you per-say ) you don't want to trust a review that uses a $20 psu tester they are trash and just make me laugh that people call it a "review" you need to check what the scope says under load, at different temps ect.

if wanted i can always point to a large collection of articles that seem to explain the basics of finding a quality psu from everything from multi~rail vs single rail to 80+ratings to ripple

lastly i wanted to also say that because we are not only end users, but that we came to freenas to make our own pc and not buy a big box brand nas is another reason that the people should have to make informed choices. for example seasonic is one of the best psu makers in the world, however they make some absolute trash psus as well, they are an oem company and will make what you want. quality or trashy, it is up to what you/the company wants to pay for we as end users need to make informed decisions, which thankfully we can with much more ease with the interwebs !

none of the psus that i recommended might work they are will work and are high quality psus


good information on psus- no it is not everything ever you need to know but great places to start

http://www.overclock.net/t/715889/phaedrus-psu-articles/0_100#post9110838

specifically the articles and myths sections,

as to where to find reviews i have never seen a bad review from johnnyguru.com they always have great data/breakdowns there are a few other site but i won't bog down any more ~
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,782
Thanks
3,036
You're free to disagree all you like, and of course you're certainly welcome to go and research your own PSU options however you want.

While there are certainly resources out there on the Internet to help you identify a quality PSU, I would like to point out that there are at least as many incompetent opinions out there.

We see people making bad hardware choices all the time. People choose cheap mainboards (no ECC, AMD APU, Realtek ethernet, Marvell ports) because "this blog said it was a good choice." People try to use RAID controllers for their pool because "this YouTube video said it was good." People choose cheap hard drives (WD Green) because "they were getting four eggs at NewEgg."

And the PC parts marketplace is plagued with cheap and/or bad. Seriously, we found Walmart had a $17 PSU when someone included that in their hardware manifest, thinking "cheap == good for wallet." It's totally possible that there's nothing seriously wrong with it, but do you really want to risk a $1K machine on it? Cheap capacitors? Undersized wiring? Crappy fan? Low end PSU's are plagued by bad.

People have been coming here having used PSU sizing web tools that were intended for average desktops with a small number of drives, and failed to include headroom for spinup. That doesn't hurt much for a one- or two-drive system, but certainly does for an 8 drive system.

So I don't trust a lot of those resources that are available out on the Internet. I know that a large number of them are worse than useless.

But your complaint doesn't even completely make sense to me; you say you want resources that allow you to make "informed choices." I'm not telling you what to buy. I *am* telling you how to figure out how big to buy it, since we see people come in here all the time with badly sized PSU's. That's why this thread is called "Proper Power Supply Sizing Guidance."

I also precalculate sizing and suggest a brand of PSU that is known to be of high quality; a large majority of the people who are looking for advice don't actually care too much to do large amounts of independent research. I've been sitting here doing this for four and a half years, now, and I have a pretty good handle on what people want and need to hear.
 

rogerh

FreeNAS Guru
Joined
Apr 18, 2014
Messages
1,069
Thanks
118
As a small-scale home user of computers, over about 20 years I have had 2 unexpected UPS battery failures (preventable by preventative maintenance or by using proprietary UPS software), about 4 hard disk failures (mitigatable (?) by using some form of RAID, except perhaps in laptops) and more than 6 PSU failures, for which there is no cheap preventative solution. Why does no-one do reasonably priced, standard form factor redundant PSUs? By reasonably priced I mean less than twice a decent non-redundant one. Am I the only person obsessive enough to buy them? For home use, with fairly reliable utility power, I would connect one half to the UPS and one direct to mains.
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,782
Thanks
3,036
As a small-scale home user of computers, over about 20 years I have had 2 unexpected UPS battery failures (preventable by preventative maintenance or by using proprietary UPS software), about 4 hard disk failures (mitigatable (?) by using some form of RAID, except perhaps in laptops) and more than 6 PSU failures, for which there is no cheap preventative solution. Why does no-one do reasonably priced, standard form factor redundant PSUs? By reasonably priced I mean less than twice a decent non-redundant one. Am I the only person obsessive enough to buy them? For home use, with fairly reliable utility power, I would connect one half to the UPS and one direct to mains.
I *love* the question. Generally, the answer is that years ago they did make standard form factor redundant PSU's (back in the AT days), but they weren't that cheap. They often went into the rack mount chassis and ended up being server-y. Once manufacturers figured out that that sort of thing primarily went into servers, then we saw an evolution towards building server redundant supplies, and then ones that fit into 2U, and then into 1U... it quickly became a highly specialized field.

One of the truths in the world of electronics is that volume matters. If you make something and manufacture 100 of them, the cost component of the final price to design and manufacture have to be N/100, but if you're manufacturing a million of them, then N/1,000,000. So any time you have a device that's in a niche market, it HAS to be more expensive or it can't be profitable. You won't find cheap redundant supplies, or, if you do, they'll be engineering disasters.

But the other thing you need to consider is that designing into a small form factor stresses electronics. Consider a typical 2U RPS.

http://www.rackmountnet.com/sparkle-520-watts-hotswap-redundant-power-supply-p-4022.html

Each of those modules is 1U high and packed with an entire power supply. This generates more heat in a small area, meaning a higher risk of failure than a non-redundant supply like a Seasonic 550, which is likely to run a lot cooler, a lot quieter, and probably at a quarter of the cost. Also, the Sparkle supply is known to fail. I'm pretty sure we've got a bin of failed modules around here, though in Sparkle's defense we do kinda run gear for something like a decade. I'm just now pulling out of the data center stuff that was deployed in 2004...

Also, when one of the modules in a redundant power supply fails, it is totally possible for that to knock the power supply offline until someone pulls the defective module. Absolutely seen that happen more than once.

My opinion is that you're probably better off getting a high quality non-redundant power supply. If you then want to defend against a failure of your deskside UPS (much more likely than failure of PSU in my experience), you can always acquire a rack auto transfer switch like the APC AP7750. This device allows you to plug your load into two different power sources, and then will automatically switch between them. Means you can power off the UPS for maintenance (battery replacement, whatever) without dumping the load, because the load just gets placed directly on utility power. Plug your NAS, your server, your network switch, and your NAT gateway into the RATS for a more high-availability computing experience. You can find these on eBay often for ~$100 so the AP7750 plus Seasonic solution is still half the cost of the redundant power supply, and probably overall more reliable.
 

rogerh

FreeNAS Guru
Joined
Apr 18, 2014
Messages
1,069
Thanks
118
I *love* the question. Generally, the answer is that years ago they did make standard form factor redundant PSU's (back in the AT days), but they weren't that cheap. They often went into the rack mount chassis and ended up being server-y. Once manufacturers figured out that that sort of thing primarily went into servers, then we saw an evolution towards building server redundant supplies, and then ones that fit into 2U, and then into 1U... it quickly became a highly specialized field.

One of the truths in the world of electronics is that volume matters. If you make something and manufacture 100 of them, the cost component of the final price to design and manufacture have to be N/100, but if you're manufacturing a million of them, then N/1,000,000. So any time you have a device that's in a niche market, it HAS to be more expensive or it can't be profitable. You won't find cheap redundant supplies, or, if you do, they'll be engineering disasters.

But the other thing you need to consider is that designing into a small form factor stresses electronics. Consider a typical 2U RPS.

http://www.rackmountnet.com/sparkle-520-watts-hotswap-redundant-power-supply-p-4022.html

Each of those modules is 1U high and packed with an entire power supply. This generates more heat in a small area, meaning a higher risk of failure than a non-redundant supply like a Seasonic 550, which is likely to run a lot cooler, a lot quieter, and probably at a quarter of the cost. Also, the Sparkle supply is known to fail. I'm pretty sure we've got a bin of failed modules around here, though in Sparkle's defense we do kinda run gear for something like a decade. I'm just now pulling out of the data center stuff that was deployed in 2004...

Also, when one of the modules in a redundant power supply fails, it is totally possible for that to knock the power supply offline until someone pulls the defective module. Absolutely seen that happen more than once.

My opinion is that you're probably better off getting a high quality non-redundant power supply. If you then want to defend against a failure of your deskside UPS (much more likely than failure of PSU in my experience), you can always acquire a rack auto transfer switch like the APC AP7750. This device allows you to plug your load into two different power sources, and then will automatically switch between them. Means you can power off the UPS for maintenance (battery replacement, whatever) without dumping the load, because the load just gets placed directly on utility power. Plug your NAS, your server, your network switch, and your NAT gateway into the RATS for a more high-availability computing experience. You can find these on eBay often for ~$100 so the AP7750 plus Seasonic solution is still half the cost of the redundant power supply, and probably overall more reliable.
Thanks for the practical advice. It is especially useful to know that redundant supplies do not always deal with module failure properly!
 
Joined
Jun 29, 2015
Messages
55
Thanks
3
You're free to disagree all you like, and of course you're certainly welcome to go and research your own PSU options however you want.

While there are certainly resources out there on the Internet to help you identify a quality PSU, I would like to point out that there are at least as many incompetent opinions out there.

We see people making bad hardware choices all the time. People choose cheap mainboards (no ECC, AMD APU, Realtek ethernet, Marvell ports) because "this blog said it was a good choice." People try to use RAID controllers for their pool because "this YouTube video said it was good." People choose cheap hard drives (WD Green) because "they were getting four eggs at NewEgg."

And the PC parts marketplace is plagued with cheap and/or bad. Seriously, we found Walmart had a $17 PSU when someone included that in their hardware manifest, thinking "cheap == good for wallet." It's totally possible that there's nothing seriously wrong with it, but do you really want to risk a $1K machine on it? Cheap capacitors? Undersized wiring? Crappy fan? Low end PSU's are plagued by bad.

People have been coming here having used PSU sizing web tools that were intended for average desktops with a small number of drives, and failed to include headroom for spinup. That doesn't hurt much for a one- or two-drive system, but certainly does for an 8 drive system.

So I don't trust a lot of those resources that are available out on the Internet. I know that a large number of them are worse than useless.

But your complaint doesn't even completely make sense to me; you say you want resources that allow you to make "informed choices." I'm not telling you what to buy. I *am* telling you how to figure out how big to buy it, since we see people come in here all the time with badly sized PSU's. That's why this thread is called "Proper Power Supply Sizing Guidance."

I also precalculate sizing and suggest a brand of PSU that is known to be of high quality; a large majority of the people who are looking for advice don't actually care too much to do large amounts of independent research. I've been sitting here doing this for four and a half years, now, and I have a pretty good handle on what people want and need to hear.
i do appreciate your response ! i was not trying to insult you in any way, i was just trying to contribute what i can! and i hope you don't take it otherwise ! i really appreciate all the help everyone here has been, esp you when i was building mine thanks so much !
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,782
Thanks
3,036
Hi, would 2x500 watt PSUs be enough power for the spin up on 12 WD RED HDDs or should I try and get higher ones?

http://www.ebay.com/itm/Supermicro-...ackplane-BPN-SAS2-826EL1-2x-PSU-/252182453681
This falls into a class of issues that gets complicated. If you are committed to making sure both supplies are functional, then, yes, I'd have a hard time picturing a system where that'd be insufficient. It's also fine for staggered spinup strategies. Both of those make assumptions about the hardware environment, assumptions I prefer to avoid. For example, if you might end up in a situation where you spin a dozen platters simultaneously with one PSU module failed, that's going to be in the "this is very stressy" badness area. But if you're disciplined enough to say "if a PSU fails then I know I cannot shut down and spin them back up unless I'm extremely careful" then - hey, fine.
 
Joined
Jul 17, 2014
Messages
11
Thanks
0
This falls into a class of issues that gets complicated. If you are committed to making sure both supplies are functional, then, yes, I'd have a hard time picturing a system where that'd be insufficient. It's also fine for staggered spinup strategies. Both of those make assumptions about the hardware environment, assumptions I prefer to avoid. For example, if you might end up in a situation where you spin a dozen platters simultaneously with one PSU module failed, that's going to be in the "this is very stressy" badness area. But if you're disciplined enough to say "if a PSU fails then I know I cannot shut down and spin them back up unless I'm extremely careful" then - hey, fine.
Maybe a better question, how many watts would I want to safely/reliably spin up all the drives with only a single PSU? I've looked at some sites that say the spin up is about ~50W per drive. So 12x that, with say an extra 100W for mobo/cpu, you're looking at a minimum of 700W required, with 80% conversion, I'm thinking a 800W would be adequate...does that sound about right?
 

Bidule0hm

Server Electronics Sorcerer
Joined
Aug 5, 2013
Messages
3,710
Thanks
648
50 W would be for 15k SAS drives, the drives we generally use for a NAS are more something like 35 W ;)

Why the 80 % conversion number?

@jgreco I just saw an error I think: in the edit you made to add my capture you've written "[...] peaking around or maybe a little over 2 amps." but actually the peak is a little over 3 amps :)
 
Last edited:

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,782
Thanks
3,036
Maybe a better question, how many watts would I want to safely/reliably spin up all the drives with only a single PSU? I've looked at some sites that say the spin up is about ~50W per drive.
You should probably not be looking at random websites for opinions of drive spin-up, unless there's some credible evidence to back it up. In my post, I use the numbers obtainable from a manufacturer's spec sheet, plus I leave headroom in the design. @Bidule0hm actually puts gear on the scope.

So 12x that, with say an extra 100W for mobo/cpu, you're looking at a minimum of 700W required, with 80% conversion, I'm thinking a 800W would be adequate...does that sound about right?
Depends on what the mainboard and CPU are likely to be chewing. The nice thing about guessing bigger is that you're not likely to end up hurting, but you also want to be careful not to go TOO big.

50 W would be for 15k SAS drives, the drives we generally use for a NAS are more something like 35 W ;)

Why the 80 % conversion number?
I'm guessing that's my conservative "leave some headroom" value (hopefully).

@jgreco I just saw an error I think: in the edit you made to add my capture you've written "[...] peaking around or maybe a little over 2 amps." but actually the peak is a little over 3 amps :)
Because I figured it was probably sucking down something massive on the 5V, so I described it as "So if we disregard about 1 amp at 5V, it seems clear that there's a massive power suck for about six seconds, peaking around or maybe a little over 2 amps."
 

Bidule0hm

Server Electronics Sorcerer
Joined
Aug 5, 2013
Messages
3,710
Thanks
648
Ah ok, no problem then.
 
Joined
Oct 20, 2015
Messages
20
Thanks
3
The nice thing about guessing bigger is that you're not likely to end up hurting, but you also want to be careful not to go TOO big.
What would be the problem(s) with going bigger? At what point does big become "too big?" Serious questions, just trying to understand if and what kind of badness that can happen, thanks.
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,782
Thanks
3,036
The two big problems with TOO big: 1) there's some overhead in a PSU and at a certain point it becomes very inefficient at supplying power when the load is very small, and 2) a PSU may not be designed to drive tiny loads and that may cause some issues such as poor power quality. So do not buy the 5000W PSU to drive your 50W load.
 

Ericloewe

Not-very-passive-but-aggressive
Moderator
Joined
Feb 15, 2014
Messages
16,027
Thanks
3,881
A modern, quality 500W (or even 1000W) ATX PSU won't have any trouble dealing with with a 50W load, though. Efficiency will probably drop noticeably if you overdo it, though, so don't just go out and buy some high-end 1.5kW PSU.

tl;dr - oversizing a bit won't hurt, overkill will hurt your wallet and undersizing can really ruin your day.
 
Top