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
11,776
Thanks
3,018
#1
I've seen about 1,000 threads like this one where people decide that they can power a dozen hard drives off a 360 watt supply. DO NOT DO THIS.

I've seen another 1,000 threads where people decide to buy the cheapest power supply that they can find. DO NOT DO THIS.

Your NAS lives or dies by its power supply. In building your NAS, you will be putting together a server platform that costs at a minimum many hundreds of dollars, and then putting hundreds or even thousands of dollars of hard drives into it. If your power supply blows, it could potentially ruin some or all of your expensive gear. Do you really want to trust a $16 power supply ?


Meta-Issues To Consider

You've probably come to FreeNAS for its awesome data protection and storage resiliency features. You've hopefully learned that you want server-grade gear, and that you want ECC memory, and that you want redundancy in your storage system. But another question you should ask yourself, how long do you want your storage system to last? Most users are looking to create a storage platform that won't be obsolete next year, and in fact usually want it to last as long as it can.

Your power supply ends up being one of the most complicated bits of electrical engineering in your system. You want to pick a power supply that has high quality components, because a failure of a component could mean anything from power loss to voltage sag to high voltage being fed through to your low voltage computer parts. These are very bad things!

Further, because we want the power supply to work as well in five (or even ten) years as it does today, we have to consider that as components age, their ability to perform slowly degrades. This degradation is made worse if a component is stressed out to near (or past) its specification. In the world of electronics, we typically cope with this using a principle known as derating. This simply means that, for example, if you needed a supply that can deliver 300 watts, you get a 400 watt supply. A typical high quality modern power supply delivers a fairly consistent level of efficiency when loaded between 20%-80% of its rated capacity, so you're not "saving lots of power" by getting a 300 watt supply for a 300 watt load. The rule of thumb in the shop here is that a power supply should never be pushed beyond 80% of its rated capacity.


The Importance of Clean Power

It is important to have a clean source of power, such as that supplied by a UPS. Under normal circumstances, your utility probably provides a fairly clean power, but under certain circumstances, such as a tree branch falling on the feeder, you may experience a brownout, followed by an autorecloser shutting off and then restoring the power to the line (often several times). Each time power returns, you get a significant amount of cruft on the line, from the inrush brownout caused by motors starting (and possibly stalling), etc. Now, you might have "figured" that you can start 12 drives and a Xeon on a 500W PSU, but that's on a normal utility supply. When that power is fluctuating rapidly in an adverse situation, can the PSU still prevail? You definitely want to avoid answering that question. Get a quality UPS as an integral part of your power supply solution.


Calculating How Much Power You Need

The big thing that most people miss when figuring power consumption is drive spinup current. Most drives take up to about 2.1 amps of additional current on the 12 volt rail to start the platters spinning - that's 25 watts per drive. This is in addition to the eight watts that the drive electronics may be consuming. Do note that some drives use less power, and some use more.

<Edit - 10/27/2015 ->
Frequent forum contributor and electronics sorcerer @Bidule0hm did the hard work of actually showing us what the combined 5V+12V power consumption on drives looks like.

5803873c7afcf2b7fca58b83d0241c0c.png


Each vertical division is 0.5A. 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.
I do not particularly like to undercalculate power requirements, because that may lead to voltage sag, which then leads to equipment failure. Some current drives require a lower amount of spinup current (~1.7A) but I encourage you to contemplate that drives might be replaced or upgraded. Running things right on the edge is a bad idea. If you are building a system with more than four drives, I encourage you to look at the specifications for your drives, but still suggest that you want to reserve about 35 watts for each drive.

When spinning up a large fleet of drives (12+), the use of a chassis with a redundant power supply gives you a large extra safety margin. I will note, however, that such an enclosure should still be sized so that it can operate on a single supply.

So there are three things that should ideally be calculated to size your power supply:

1) Total potential watts
2) Total potential 12V amps
3) Average idle watts

The total potential watts is the sum total of all your drives (~35 watts each), ~25 watts for a mainboard, look at the TDP of your CPU for a poor estimation of the peak watts there (~80?), ~6 watts per stick of memory, ~10 watts per LSI 8 port HBA, and ~15-30 watts per fan. Honestly it can be a fair amount of work to come up with a good number, take some time and look at data sheets.

The total potential 12V amps is important because you want to make sure the supply can provide it, and you want to be aware of any other issues, such as if the supply has dual rails.

But also important is the average idle watts. Unlike the calculations above, instead you take the measured or estimated idle watts. For a hard drive, that's often about 6 watts. A modern CPU, might only be 10 watts. Components like the mainboard, memory, and HBA tend to be nearly fixed in their consumption. Fans are a pain in the rear. Just try to see what you can figure out.

Now, look at your total potential watts. It is probably a big number. Multiply that by 1.25, and that's your target size for a power supply.

Multiply that number by 0.20. If the result is less than your average idle watts, you have a power supply size that's going to be in the efficiency sweet spot for a high quality PSU.

Force fitting that into an available PSU is of course a little fun. You are encouraged to go down a shade or up a tier as needed in order to get the right PSU size.


Awww, Do I Have To?

No, you don't. But you really should. However, there are ways to control the amount of load thrown at a PSU. For a NAS, since the thing that normally dominates the calculation is the spinup current for the drives, there are two things:

1) Use a disk with a lower spinup current. I don't like this method, because invariably someone will replace a drive with a higher current unit.

2) Carefully design your systems to use staggered spinup. This is fraught with peril but if you are sufficiently disciplined, can work. On the other hand, you wouldn't be reading this sticky in that case.

The third option:

3) Random chance suggests that drives are not actually LIKELY to all experience a surge inrush at the exact same moment of time. You can tempt fate.

It isn't clear that there's much value to be had in relying on any of these strategies, unless you are driving so many disks that you can't get a sufficiently large PSU.


Power Supply Efficiency/Cost

There's a school of thought that wants to minimize the size of a PSU and instead rely on the manufacturer's competence and build quality. You CAN do this. I think it a fool's errand, but, yes, you can.

The last time I debated this with a supporter of this strategy, I ran some numbers. It turns out that if you run a 60W load on a SeaSonic G-360, efficiency is 86%, and watts consumed at the wall are 70W. If you run the same 60W load on a SeaSonic G-550, efficiency is 83%, and watts consumed at the wall are 72W. That's a two watt difference to move up to a power supply that supplies 50% more power. At a price of 14c/kWh, the additional electrical cost is about $12... over the next five years.

I am totally willing to pay $2 per year more in electricity for a larger supply that is more suitably sized and less likely to ruin my $1000+ NAS boxes. Trying to microsize your PSU is a game for chumps.


Forum Favorite PSU

We don't have much call for non-rackmount gear here in the shop, and I normally don't care to suggest things we don't have a lot of experience with. However, the SeaSonic G-series (360, 450, 550, 650, 750) is a highly respected PSU line that is well-loved by PC builders all around, with high quality components, tight voltage regulation, and a five year warranty. It is also the #1 choice here on the forums.

The G360 is not modular. The others are.

2017-11: Apparently Seasonic has discontinued the G in favor of the FOCUS (thanks @Ericloewe / @gamedude9742 ) but as I haven't seen many competent opinions regarding the quality of the newer PSU, this is presented as a FYI rather than a recommendation.


TL; DR - Precalculated Guesses for the Lazy Geek

Okay, so you don't want to mess with all that crap. You just want to know what to buy! The following makes an assumption of a fan for every four drives and a reasonable amount of memory, and an HBA for more than 4 drives. These are ballpark numbers. Guesses based on back-of-a-napkin math. You should do your own homework. And yes of course you're not likely to hit those peak numbers except perhaps during bootup, but they're still a sane thing to shoot for.

1) For an Avoton C2550/C2750 (18-35W board, 12W memory):
  • 1-2 Drives: 132W peak, 46W idle -> SeaSonic G-360
  • 3-4 Drives: 202W peak, 71W idle -> SeaSonic G-360
  • 5-6 Drives: 297W peak, 118W idle -> SeaSonic G-450
  • 7-8 Drives: 367W peak, 134W idle -> SeaSonic G-450
  • 9-10 Drives: 437W peak, 150W idle -> SeaSonic G-550
  • 11-12 Drives: 507W peak, 166W idle -> SeaSonic G-650 or X-650
2) For an E3-1230v3 (32-98W board+CPU, 12W memory):
  • 1-2 Drives: 195W peak, 75W idle -> SeaSonic G-360
  • 3-4 Drives: 265W peak, 91W idle -> SeaSonic G-360
  • 5-6 Drives: 360W peak, 132W idle -> SeaSonic G-450
  • 7-8 Drives: 430W peak, 148W idle -> SeaSonic G-550
  • 9-10 Drives: 510W peak, 174W idle -> SeaSonic G-650 or X-650
  • 11-12 Drives: 585W peak, 195W idle -> SeaSonic G-750 or X-750
3) For an E5-1620v3 (speculative: 70-210W board+CPU, 24W memory):
  • 1-2 Drives: 319W peak, 125W idle -> SeaSonic G-450
  • 3-4 Drives: 389W peak, 141W idle -> SeaSonic G-550
  • 5-6 Drives: 484W peak, 182W idle -> SeaSonic G-650 or X-650
  • 7-8 Drives: 554W peak, 198W idle -> SeaSonic G-750 or X-750
  • 9-10 Drives: 639W peak, 229W idle -> SeaSonic X-850
  • 11-12 Drives: 709W peak, 245W idle -> SeaSonic X-850 or X-1050
 
Last edited by a moderator:

INCSlayer

FreeNAS Experienced
Joined
Apr 4, 2015
Messages
201
Thanks
27
#2
Sweet a power calculation guide i could understand.

*does the math*
Throws out the 850w psu for a 1500w one luckily im still in the planning phase

24drives require a lot of power it seems :p
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,776
Thanks
3,018
#3
Usually for something beyond 12 drives, you want to get a redundant power supply. The fact of the matter is that 12 drives are unlikely to all spin up at exactly the same time and actually consume all your power. You'll notice that in most rackmount designs that there are two power supplies, either of which are capable of holding the load, often just barely, but when teamed they are both just lazily feeding power.

The Supermicro 6028R for example comes with a dual 920W supply. If you take my 12-drive E5-1620v3 number from above (709W peak) and add another 140W on there, you come out around 849W, just shy of 920W.

However, as you get into the larger number of drives (24), you can a little-more-safely rely on statistical unlikeliness. The Supermicro SSG-6048R actually also comes with a dual 920W supply and I'm certain that it can power up safely with both supplies, though I'd be a little hesitant to try it with a single.

Going further, you get up to the Supermicro SC946ED, which runs 90 drives off of four 1KW PSU's. That's just a storage enclosure and powering all those drives might take (35W * 90 ~=) 3150W. They've rated the supply at 2KW though, so obviously they don't expect all the drives to spin at the same exact moment.

The sizing guide I've given here is more intended to help builders of smaller systems avoid making underpowered systems. Once you get out to a larger number of drives, it isn't as linear as what I suggest above.
 
Joined
Oct 22, 2015
Messages
14
Thanks
3
#4
Thank you for this guide, it was very helpful! I went through and got some estimates for a system I'm about to build:

Product - Total Potential Watts, Average Idle Watts
Supermicro X10SL7-F Micro ATX LGA1150 Motherboard - 60, 10
Intel Xeon E3-1231 V3 3.1GHz Quad-Core Processor - 91, 24
Crucial 16GB (2 x 8GB) DDR3-1600 Memory - 24, 8
8x WD Red 3tb Drives - 280, 48
3xFractal Design HP14-PWM 78.1 CFM 140mm Fan - 45, 9

Total Potential Watts = 500watts
Average Idle Watts = 99watts
Total Potential Watts x 1.25 = 625watts

I'm thinking I'm a bit on the high side for some of these estimates but I spent the morning researching and these are what I found. I had originally thought I would go with a SeaSonic SSR-450RM but maybe I should consider a SSR-550RM or SSR-650RM
?
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,776
Thanks
3,018
#5
There was some discussion recently regarding the SSR vs the G. You might want to take a look and see if you can find it.
 
Joined
Apr 20, 2015
Messages
10
Thanks
8
#6
Honestly it can be a fair amount of work to come up with a good number, take some time and look at data sheets.
I went through several data sheets when I did the estimation for my system. I shared those number in this post because they can be useful to others.
Some are very common components among the FreeNAS users here.
 

mattbbpl

FreeNAS Experienced
Joined
May 30, 2015
Messages
237
Thanks
35
#7
Thanks for posting, it's very helpful. I have a question that will probably be pertinent to a number of users here based on common purchase recommendations.

Your guidelines for redundant power supplies are pretty clear (size so you only need one, but you can feel more comfortable approaching the limit of that power supply). Since power supply efficacy degrades over time, when purchasing a used chassis with what we assume (to be safe) are used power supplies, do we need to add extra buffer to that limit - perhaps continuing to abide by the 80% rule in spite of the redundancy? Or is the likelihood of failure during peak consumption time (i.e. simultaneous spin up during a restart) low enough to not warrant such considerations?

I realize there's no "absolutely right answer" to this question. Just asking for personal thoughts from those with more experience on the matter.
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,776
Thanks
3,018
#9
Well, no, that's just some random redundant PSU chatter, not really guidelines. Guidelines get complex.

Consider for example: We used to sell the AIC RMC4E, a 24-drive chassis with a 950W power supply. The 950W power supply was actually a 3+1 redundant supply, made up of four 350W PSU modules. Any individual module could fail and life was fine.

But! A lot of people wanted to be able to plug into two separate power sources, so that if one failed, the server would live. This wasn't possible with the RMC4E; it meant that two PSU's had to plug into each power source, and that if one power source failed, suddenly only two 350W PSU's (~600W of capacity) were shouldering the load. We had to tell people that EITHER they had to plug all four PSU's into the same power source, OR that they needed to plug each PSU into four separate power sources.

So my answer is really more one of "you have to think about it."

If you've built a system that runs close to the capacity of one of the PSU modules, I'd say you have to be very conservative about how you behave in a scenario where only one PSU module is shouldering the load.

There's a perfectly legitimate argument to be made that spiking a PSU to near its rated watts for a few seconds should not be horribly injurious to a PSU; I'd even tend to believe that is more likely to be true for a high-quality server-grade PSU module that you're likely to find in a Supermicro or HP or Dell box. I can tell you the AIC PSU modules were manufactured by some crappier outfit (maybe Zippy?) and had a horrible reputation for burning out after a few years, but at the time they were the only show in town.

In the end, you have to figure out what your risk tolerance is, especially bearing in mind that if you blow out a module it might cost you an arm and a leg to acquire an appropriate replacement. My *strong* preference is to make sure all modules are supplying power and then you're probably just dandy. If you lose a module, then carefully think about what peak loading might be, and how old the supplies are, etc.
 

Jailer

Not strong, but bad
Joined
Sep 12, 2014
Messages
4,098
Thanks
1,031
#10
Thanks for posting, it's very helpful. I have a question that will probably be pertinent to a number of users here based on common purchase recommendations.

Your guidelines for redundant power supplies are pretty clear (size so you only need one, but you can feel more comfortable approaching the limit of that power supply). Since power supply efficacy degrades over time, when purchasing a used chassis with what we assume (to be safe) are used power supplies, do we need to add extra buffer to that limit - perhaps continuing to abide by the 80% rule in spite of the redundancy? Or is the likelihood of failure during peak consumption time (i.e. simultaneous spin up during a restart) low enough to not warrant such considerations?

I realize there's no "absolutely right answer" to this question. Just asking for personal thoughts from those with more experience on the matter.
To add to this question, what is considered "old" as far as power supplies go? I've got an "old" ATX power supply that I'm using in my FreeNAS build that is supposed to be a very good power supply when it was built. At what point under normal usage scenarios would a power supplies specs start to fall off?
 

tvsjr

FreeNAS Guru
Joined
Aug 29, 2015
Messages
943
Thanks
319
#11
To add to this question, what is considered "old" as far as power supplies go? I've got an "old" ATX power supply that I'm using in my FreeNAS build that is supposed to be a very good power supply when it was built. At what point under normal usage scenarios would a power supplies specs start to fall off?
The reality is that the 80% derating value mentioned above is extreeeeeemely conservative. If the power supply's output falls off 20% and hasn't totally flamed out, the rest of the computer is so old that it's worthless. I've got power supplies that are 30+ years old that still make at or above rated power.

The 80% value gives good headroom for a number of things... production variances (every power supply has some single digit percentage tolerance of how much power it actually produces versus its rating), derating over time, thermal variation (supplies make less power when they're hot), etc. Look at it as your 'oops' factor to cover multiple issues.
 
Joined
Oct 19, 2015
Messages
16
Thanks
0
#12
I've recently put toghether a freenas box, and i was really interested in the power consumption. I didn't do any estimates before the setup, because the chassis already came with 2x750w psu's. But what i've found using a kill-a-watt meaduring device is that the harddrives don't actually take that much power at spinup vs. normal drive usage.

Basically, for a 2U chassis, with: -2x750w power supplies (but only one connected for the test)
-single L5630 cpu
-12 GB ram (1.5v ddr3 ecc reg)
-S5520UR motherboard
-5 system fans
-Intel Rs2wc040 hba
-intel expander
-12 sata 1tb drives(mixture of HGST and seagate constelation ES and CS )

At powerup, peak consumed power is 260W, after everything settles down, and freenas finishes booting, system stabilizes at 185-190w.
I guess the difference can be attributed to the fans and harddrives.
But even if we ignore the fans, a 70w difference, split by 12 drives, comes down to 5.8w difference per drive between spinup and whatever state the drives are in after boot. So not even close to the 15-20w i was usually taking into consideration when sizing psu for a server.

Sent from my SM-N9005
 
Joined
Jan 21, 2014
Messages
85
Thanks
0
#13
Thank you for finally having a guide for calculating the required power supply. My build is kind of undersized (I realized this before I read the guide) but now I'm just blown away by what power supply i SHOULD use. Currently I'm using a X10SL-7 build running with 12 disks. A few weeks ago I got myself a new chassis which allows me expension up to 20 drives. Now I wanted to scale my power supply properly to be on the safe side. When doing the math I end up requiering a power supply in the 1200 Watt range.

As you stated in an other post things change a bit when having more then 12 drives running but still it should be more than the 900 Watt suggested by you for the 12 disk setups. My question now is: What is/are an effordable / recommended power supplie(s) for this case? The Seasonic G-Series does not offer that big power supplies and on a first look the redundant power supplies cost a furtune.
 
Joined
Apr 20, 2015
Messages
10
Thanks
8
#14
... But what i've found using a kill-a-watt meaduring device is that the harddrives don't actually take that much power at spinup vs. normal drive usage.
...
You have to consider two factors.
1. The kill-a-watt (a 20$ device?) hasn't the sampling rate necessary to measure the transients involved during the system startup. The peak power it measured is not the actual peak power. The peaks are usually too short to be measured by a such device.
2. A transient at the DC output doesn't translate in a transient at the AC input. The peaks of power absorbed at the DC are averaged out at the AC input. You should measure the current and the voltage between the power supply and the components to really see what the PS has to sustain.

So your conclusions about the "actual" power required by your hard drives are wrong. Again, the data sheets don't lie. Follow them.
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,776
Thanks
3,018
#15
At powerup, peak consumed power is 260W, after everything settles down, and freenas finishes booting, system stabilizes at 185-190w.
The Kill-A-Watt is not up to the challenge of measuring peak current; you want a high quality power analyzer like the ones made by Valhalla or Fluke to do this, or at least a scope. The Kill-A-Watt is a fine thing to plug into your PC (or refrigerator or whatever) to see how much power is being used over time and how much that might be costing you, but its sampling window is extremely low because it is oriented towards summing power usage over minutes, hours, or days of time, not seconds.

Fortunately what we are really trying to measure isn't the power supply's peak inrush current, but rather the power required to spin up a dozen hard drives. This is a slightly different problem than the traditional peak inrush current problem, because we're not really worrying about the charging of capacitors over the period of a few milliseconds, but rather a peaky sort of beast.

... and I found pictures. And more talking material. You poor sorry bastards! ;-)

The fine folks over at 45Drives have some pretty good info. So, here, they have a chassis with 45 drives, all spinning up "simultaneously".

12V_5V_SU_45_EDIT_2.png


Now, what they're showing here is essentially that as the number of drives increases, the more you can rely on the fact that the drives won't all spin precisely at the same moment. Honestly, you're not likely to get them spinning all at the same moment even on a smaller system, but a smaller PSU has less elbow room to soak up sudden demand, so the larger supply is a more conservative, safer bet.

As a system gets larger, you could actually start to lean on that, and/or also start to control the load through staggered spinup techniques, which would allow you to get by with a smaller PSU. Further, the reality is the numbers I use for calculation are the worst case numbers, and while the spec sheets might SAY 2.1 amps on the 12V rail, the power actually drawn on the 12V may average substantially less. For example, the chart above suggests that the current draw for the drives they used might be closer to averaging 1.1 amps at 12V. It would be interesting to see a graph of the draw of a single drive to more fully understand what's going on here. The spin time of a single drive is typically around 3.5 seconds, and the graph seems to show heavy draw from around 2 to 7.5 seconds, so it seems like the drives may be varying in startup time over a period of maybe two seconds?

Obviously the variables involved in calculating large system PSU's gets complicated quickly. My goal with this sticky was to address the most egregious well-meaning-but-mistaken newbie mistakes and to provide some recommendations that I feel confident would not turn into a mistake, which is why I totally disregard the impact of slightly varied start times and other mitigations and just look at the spec sheets. Here in the shop we've got a scope, a Valhalla power meter, clamp ammeters, and a lot of other toys, but the average forum user just needs something that's going to be fine for their application.
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,776
Thanks
3,018
#16
To add to this question, what is considered "old" as far as power supplies go? I've got an "old" ATX power supply that I'm using in my FreeNAS build that is supposed to be a very good power supply when it was built. At what point under normal usage scenarios would a power supplies specs start to fall off?
I guess the big question is "what defines a very good power supply." A very good power supply will actually have a comfortable margin engineered right in, so that even ten years later it'll be able to blast out its full watts AND still have an engineering margin that doesn't run components at their limits. Engineers design stuff like that all the time.

The real problem is that we're talking PC parts manufacturers here. There's price pressure. Sure someone COULD make a super high quality 550W power supply with large safety margins and high quality components, but do you really want to pay $150+ for a PSU when the cheapies are .99 and the Seasonic G-550 is ? And who is to know that the components in such a PSU are actually over-engineered? Or that there isn't already sufficient margin and quality built into the Seasonic?

So I can't tell you the answers to any of that with any level of comfort. As a result, I've gone for the conservative approach: belt and suspenders. The belt is that we've seen for years that the Seasonics are well-respected and don't seem to be constantly letting out the magic smoke. The suspenders is that I assume they're still crap and calculate sizing for them using aggressive sizing and derating rules.

With all that said, I consider a five year old supply to be "getting old" and a ten year old supply to be "dangerously old." One of the things that you'll find is that the ATX specification has evolved over time, and that means that what was appropriate half a decade ago might no longer be appropriate. Recycling old power supplies therefore becomes somewhat risky. If you've got a PSU that has seen heavy demand placed on it, you're running a higher risk that there could be some reduction in capacity over time. There are no good solid answers.
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,776
Thanks
3,018
#17
The reality is that the 80% derating value mentioned above is extreeeeeemely conservative. If the power supply's output falls off 20% and hasn't totally flamed out, the rest of the computer is so old that it's worthless. I've got power supplies that are 30+ years old that still make at or above rated power.
Absolutely correct.

The 80% value gives good headroom for a number of things... production variances (every power supply has some single digit percentage tolerance of how much power it actually produces versus its rating), derating over time, thermal variation (supplies make less power when they're hot), etc. Look at it as your 'oops' factor to cover multiple issues.
Very nicely said. My goal in this sticky is to provide guidance that gets users to a place where they're figuring a size that is going to work for them. Going a little too big isn't a risk. Going too small is, along with all the various other risks you've just noted, and a hundred other possible issues like fan stalls.
 

Ericloewe

Not-very-passive-but-aggressive
Moderator
Joined
Feb 15, 2014
Messages
16,139
Thanks
3,867
#18
I'd like to add a detail to the excellent OP:

Around 650W, the G-Series starts having to deal with its bigger brother, Seasonic X-Series/Platinum(and Titanium, in the future).

The X-Series is a bit more expensive, but uses a more modern design, has a better fan (120mm San Ace ball bearing, instead of Hong Hua hydrodynamic bearing or Yate Loon ball bearing on newer and older G-Series units, respectively), is fully modular and, optionally, does Platinum efficiency.
The G-Series platform is also optimized for the lower power levels, so the G-750 doesn't quite match the G-550 or G-450 in voltage regulation, whereas the X-Series is a mid-to-high-power platform.
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,776
Thanks
3,018
#19
Thank you for finally having a guide for calculating the required power supply. My build is kind of undersized (I realized this before I read the guide) but now I'm just blown away by what power supply i SHOULD use. Currently I'm using a X10SL-7 build running with 12 disks. A few weeks ago I got myself a new chassis which allows me expension up to 20 drives. Now I wanted to scale my power supply properly to be on the safe side. When doing the math I end up requiering a power supply in the 1200 Watt range.

As you stated in an other post things change a bit when having more then 12 drives running but still it should be more than the 900 Watt suggested by you for the 12 disk setups. My question now is: What is/are an effordable / recommended power supplie(s) for this case? The Seasonic G-Series does not offer that big power supplies and on a first look the redundant power supplies cost a furtune.
You found a non-rackmount chassis that holds 20 drives? That just strikes me as a bad idea (especially for cooling) but then again I work mostly with rackmount gear, so you can disregard my bias. :)

For a large system, you probably don't want to follow my sizing math outlined above, because it rapidly gets untenable above 12 drives (arguably even at 12 drives). I don't have particular recommendations beyond that point, because even the Seasonic recommendation isn't truly mine, but rather that of many happy forum users plus my eyeballing of a bunch of reviews.

My best guess is that you could very well still be fine with a ~900W quality power supply. We used to sell 950W PSU's that powered 24-drive systems which peaked at 6.6A (792W) but that was back in the days of 250GB SATA disks. More recent arrays are still generally in that ballpark. The difference is that we're actively measuring this stuff as we build systems, and have the tools to do so. I am ... very nervous ... about making recommendations that are not provably reasonable on a mathematical basis. However, I do encourage you to read every word I've written, and then to look at the 45Drives data above (bearing in mind that's drives ONLY), to make some educated guesses as to power requirements. A lot of this business is just doing research, then doing your homework, and understanding the realities.
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,776
Thanks
3,018
#20
I'd like to add a detail to the excellent OP:

Around 650W, the G-Series starts having to deal with its bigger brother, Seasonic X-Series/Platinum(and Titanium, in the future).

The X-Series is a bit more expensive, but uses a more modern design, has a better fan (120mm San Ace ball bearing, instead of Hong Hua hydrodynamic bearing or Yate Loon ball bearing on newer and older G-Series units, respectively), is fully modular and, optionally, does Platinum efficiency.
The G-Series platform is also optimized for the lower power levels, so the G-750 doesn't quite match the G-550 or G-450 in voltage regulation, whereas the X-Series is a mid-to-high-power platform.
No objections there.
 
Top