What power supply for 24 drives ?

A

Aesma

[H]ard|Gawd
Joined
Mar 24, 2010
Messages
1,860
Hello :) I'm posting this here since it's quite a specific use, and I want answers from people with similar setups.

I'm actually interested in two PSUs, one for 24 drives along with an HP expander, and another PSU for 24 drives + motherboard/quadcore Xeon/RAM/HBAs/expander. The drives will likely be green ones but I'm not ruling out 7200rpm entirely. The Xeon could be on a 1155 or 2011 socket, I'm not sure yet. In both case a Norco 4224 would be used (or rather, an Inter-Tech 4U-4324L, but it's the same case).

One imperative is that they must be at least gold certified, available in the EU (you can check here : http://geizhals.at/eu/?cat=gehps ), and discreet. If one model can be used for both situations while staying efficient and not too costly, then that's an advantage, I like having several identical parts when troubleshooting things.

Some of the PSUs I'm looking at are lacking in Molex connectors, sometimes there are only two for a 1000W PSU ! Is this a major problem or is using SATA power=>Molex adapters fine ?

Thanks for your input.
 
I have a motherboard and 10 hard drives running off a 380w poer supply. At idle the power demand is 75w. Peak demand is under 150w.

Explain why you think 1000w is necessary. Most likely 380w is enough for you.
 
7.5 watt max spinup or so? Multiply by 24 would be 180 watts. Include an over estimated 200 for MB/CPU at load, and a buffer so 600 should be low end of overkill.
 
yeah, most people overestimate the needs especially for a server.

imho. I'd go with a Seasonic gold. x750 or x850 is fine for the norco. I run a x750 in my 4220. I haven't cracked it open in a while but I believe the back plane uses molex to connect. 5 drives per molex). I'll double check later when i get home is nobody adds.
 
The 1000W was just an example to talk about the number of molex, although I'm looking in this power bracket (in fact into the 1200/1400 range) for my gaming rig, which has several video cards and runs at full power 24/7. For the PSU to be the most efficient and quiet it must be far from its maximum output, but not too close to the minimum either.

For powering the drives I don't have the case yet but from my understanding it has 2 molex per group of 4 drives, so 12 molex. Half of them are there for redundancy, with some suggestions to connect them all anyway.
 
For a PSU to be most efficient, it must be within 100-200 watts of the maximum theoretical load value. Quietness has absolutely nothing to do with % load, and everything to do with efficiency and absolute load numbers.

You will need to make your own custom wiring. There are no consumer grade power supplies designed to power tons of hard drives.
 
Aesma said:
For powering the drives I don't have the case yet but from my understanding it has 2 molex per group of 4 drives, so 12 molex. Half of them are there for redundancy, with some suggestions to connect them all anyway.
Click to expand...

so to this. yes, just use SATA to Molex adapters (got mine of newegg) to make up what's lacking in molex. I'm using it for 2 rows.

the redundancy, don't bother to connect if it's not from a different power supply.


As far as your gaming rig. I'm using Seasonic 1000w Platinum since it's release. stable as can be. Though if you say several vid cards (you're running 4, which?) Need to know so know what's the max wattage at full load. And I find it hard to believe you're running "full" power 24/7 unless you're running benchmarks and the like when you're not actually using the machine.

Again, the seasonic's are quiet. really with the type of machines "hard" folks run, it's almost a nonissue how much sound the power supply eeks out.. i never hear them no matter what load I'm putting on the machine because the case fans are magnitudes louder.
 
Tsumi said:
For a PSU to be most efficient, it must be within 100-200 watts of the maximum theoretical load value. Quietness has absolutely nothing to do with % load, and everything to do with efficiency and absolute load numbers.
Click to expand...

Actually, this is not true of modern high-efficiency PSUs. Get one 80+ Gold or Platinum and they will hold 90+% efficiency all the way down to 20% load (or even lower). In many cases, maximum efficiency is found near the middle of the load curve (60% load or so). Reputable manufacturers publish the load curves on their websites.

Its still silly to buy a bigger PSU than you need - just that efficiency at lower loading is not really why.

Tsumi said:
You will need to make your own custom wiring. There are no consumer grade power supplies designed to power tons of hard drives.
Click to expand...

Agree that you will want to do custom wiring. Get a modular PSU and build your own power cables. Getting a massively over-sized one (and spending $massive to do so) is just silly when all you want is more cables to distribute power. For your Norco case what you really want to do is build a single cable running from the PSU and up one row of Molex and a 2nd cable from the PSU running up the 2nd line of Molex connectors with a short tail to two SATA power connectors for any 2.5s or SSDs you put on the shelf above the hotswap bays.

Also - for this application - look for single-rail designs. It will save you the hassle of trying to allocate load to each rail.

For your 24 drive applications (both with and without the MB) something in the 500-600 watt range, 80+ Gold/Platinum, single rail & modular is what you want. Smaller and you need to do some engineering math to be sure its happy - bigger is just wasteful.
 
Last edited:
PigLover said:
Actually, this is not true of modern high-efficiency PSUs. Get one 80+ Gold or Platinum and they will hold 90+% efficiency all the way down to 20% load (or even lower). In many cases, maximum efficiency is found near the middle of the load curve (60% load or so). Reputable manufacturers publish the load curves on their websites.

Its still silly to buy a bigger PSU than you need - just that efficiency at lower loading is not really why.



Agree that you will want to do custom wiring. Get a modular PSU and build your own power cables. Getting a massively over-sized one (and spending $massive to do so) is just silly when all you want is more cables to distribute power.

Also - for this application - look for single-rail designs. It will save you the hassle of trying to allocate load to each rail.

For your 24 drive applications (both with and without the MB) something in the 500-600 watt range, 80+ Gold/Platinum, single rail & modular is what you want. Smaller and you need to do some engineering math to be sure its happy - bigger is just wasteful.
Click to expand...

What I was referring to was the myth that you need to buy far above your max load levels in order to achieve maximum efficiency, which is far from the truth.
 
Tsumi said:
What I was referring to was the myth that you need to buy far above your max load levels in order to achieve maximum efficiency, which is far from the truth.
Click to expand...

But what you wrote was "For a PSU to be most efficient, it must be within 100-200 watts of the maximum theoretical load value", which is highly inaccurate and will mislead many readers who come to this forum looking for good information.
 
PigLover said:
But what you wrote was "For a PSU to be most efficient, it must be within 100-200 watts of the maximum theoretical load value", which is highly inaccurate and will mislead many readers who come to this forum looking for good information.
Click to expand...

It is very accurate due to the fact that most computers are never at full theoretical load, and in fact sit somewhere around 10-15% of maximum theoretical load while idling and 60-80% of maximum theoretical load while gaming and running other intensive programs.

Even though PSUs have made great advances in low load efficiency, they still do take a hit in efficiency at load levels below 20%, and the hit gets worse the lower the load %.
 
Tsumi said:
It is very accurate due to the fact that most computers are never at full theoretical load, and in fact sit somewhere around 10-15% of maximum theoretical load while idling and 60-80% of maximum theoretical load while gaming and running other intensive programs.

Even though PSUs have made great advances in low load efficiency, they still do take a hit in efficiency at load levels below 20%, and the hit gets worse the lower the load %.
Click to expand...

FUD. And completely unrelated to the topic of this thread. Go to the 80+ website and read about real power efficiency curves and how modern, efficient PSUs really work...
 
PigLover said:
FUD. And completely unrelated to the topic of this thread. Go to the 80+ website and read about real power efficiency curves and how modern, efficient PSUs really work...
Click to expand...

I don't think you even know what I'm trying to say anymore.

I am very confident that my knowledge of power supplies is well above yours. I don't need to read up on anything, because I have already done so.
 
Tsumi said:
I don't think you even know what I'm trying to say anymore.

I am very confident that my knowledge of power supplies is well above yours. I don't need to read up on anything, because I have already done so.
Click to expand...

Uh huh. Which is why you keep writing things that are not true...like this 100% false, completely misleading statement from your first post in this thread: "For a PSU to be most efficient, it must be within 100-200 watts of the maximum theoretical load value.". Unfortunately, everything else you've written in this thread is nothing but trying do anything but admit you were wrong.

To the OP, Aesma, you've gotten some good advice from several posters in this thread. As long as you ignore the FUD from Mr "I'm so confident in my ignorance that I don't need to read anything Tsumi", you will be well served.
 
Last edited:
Uh.. HP SAS expanders do not support staggered spinup. For a 20 disk system with a core 2 duo it got up to around 700 watts for a few seconds at initial power on spinning up 20 disks at once so this is something to keep in mind...
 
Thanks for the answers everyone, I'll answer in detail later. I'd just like to say that I always look at the tests of the 80+ website and that's what convinced me that I wanted a far bigger PSU than seems necessary, not just for efficiency reasons but also noise, durability, and margin since I like to overclock (and my setup is fully watercooled). And yes that "gaming" rig (my main rig really) is always at 100% thanks to BOINC, including on the GPUs.

houkouonchi : good point (and more to the point, since I kinda hijacked my own thread), I was concerned about that, in fact I currently power 14 HDDs with a dedicated 500W FSP/Fortrum Aurum, it works fine but I haven't measured what happens when I press that power button.
 
For all intents and purposes, I'd just ignore the 80Plus rating or the whole thing in general:
http://www.hardocp.com/article/2011/10/04/80_plus_irrelevant_to_you_when_buying_psu/

It really doesn't matter that much. Yes SATA power to molex adapters should work just fine.

Anyway, for that 24 drive DAS you're talking about, a solid 500W to 550W PSU should be enough. So something like the Seasonic S12II-520 520W PSU should do the trick. If you must have Gold for whatever reason, the Seasonic SSP 550RT or the Seasonic G-550 550W PSUs are good choices.

For the actual server, just to err on the safe side, I'd go with at least a 750W PSU to 850W PSU. I second the recommendation for the Seasonic X-750 750W or X-850 850W PSUs.
 
PigLover said:
Uh huh. Which is why you keep writing things that are not true...like this 100% false, completely misleading statement from your first post in this thread: "For a PSU to be most efficient, it must be within 100-200 watts of the maximum theoretical load value.". Unfortunately, everything else you've written in this thread is nothing but trying do anything but admit you were wrong.

To the OP, Aesma, you've gotten some good advice from several posters in this thread. As long as you ignore the FUD from Mr "I'm so confident in my ignorance that I don't need to read anything Tsumi", you will be well served.
Click to expand...

Fine, I might have worded that wrong. I meant for most efficient general usage operation. But you do not need massive headroom for most efficient and most quiet operation, which was the main point I was addressing. Especially in OP's case, where most of his power consumption is in hard drives that will shut themselves down and consume very little power.
 
It's bit too simplistic to just say you need XXX watt PSU to support 24 drives properly.
It depends on the drives, the PSU in question and the setup they are used in.

If you can't configure for staggered spinup, then you need to spec the PSU to spinup all the drives simultaneously (plus whatever else the PSU is starting up at the same time) - given that a typical 12V spinup current can be around 2A, that's a fairly hefty 48A hit on the 12V rail(s), not including any draw by other components.
High performance drives might have even higher requirements - though on the flip side, green drives might have lower demands, especially those that can be configured for low current spinup.

If you use a twin rail PSU, then the rail you are running your disks off would need to supply 48A - this might be problematic given that many multirail PSUs are limited to 20A per rail - you'd really need 3 rails just to power the drives.
You may find that PSUs with 4x 12V/20A rails are in the 1000W rated range.

Personally I'd spec a lower rated single rail PSU, which could supply 50-60A on its 12V rail (depending on your exact components). Have a look at the Corsair AX range, probably the 650 or 750 model - though several reputable makers do similarly specced units with 80+ Gold rating.

One issue you may find is that the specs quote continuous power rating, but rarely, if ever, mention the peak power capabilities. You need a 12V rail which can supply 50-60A for a few seconds, but which probably won't draw more than 15A in normal use - you need the rail to do this without tripping any overcurrent sensors, and without any long term detrimental effects. Of course, if you spec a PSU with a continuous rating of 50-60A, then no problem, but what if you want to spec a lower rated PSU - say one with a 30A max continuous load rating?
The bottom line is that some PSUs might do it and some won't - hence you can get conflicting reports from the field about just what PSU you need. You can't tell from the specsheet - though good online reviews can help.


If you can stagger the spinup, then the peak 12V demands are lowered, and you can use a lower rated PSU.
 
Danny Bui said:
For all intents and purposes, I'd just ignore the 80Plus rating or the whole thing in general:
http://www.hardocp.com/article/2011/10/04/80_plus_irrelevant_to_you_when_buying_psu/

It really doesn't matter that much. Yes SATA power to molex adapters should work just fine.

Anyway, for that 24 drive DAS you're talking about, a solid 500W to 550W PSU should be enough. So something like the Seasonic S12II-520 520W PSU should do the trick. If you must have Gold for whatever reason, the Seasonic SSP 550RT or the Seasonic G-550 550W PSUs are good choices.

For the actual server, just to err on the safe side, I'd go with at least a 750W PSU to 850W PSU. I second the recommendation for the Seasonic X-750 750W or X-850 850W PSUs.
Click to expand...

I concur sir!
 
The only recommendation I can make is this: SeaSonic X750 Gold 750W 80 PLUS GOLD Certified Full Modular Power Supply

You will not be disappointed and any extra wattage is just waste.
 
I have a Norco RPC-4224 with an LSI MegaRAID 9261-8i and an HP SAS Expander. Xeon E3-1220v2, 8GB of RAM and Toshiba DT01ACA300 drives...Rosewill FORTRESS 550w 80+ Plat PSU, and it works great. Never did any testing...but I have no problems at all with power.
 
I am amazed at how much power people think computers need.

I just rebooted my TV server, Asus motheroard with I3 CPU, 10 WD green drives. Booting up with 10 HDs the system peaked at 107w (Kill-A-Watt). The system idles at 77w and peaks at 110w at 100% load (Primes95).

My Kill-A-Watt appears to give the correct reading for a 100w light bulb. So I tend to trust it.

According to my simple test, 3 motherboards and 30 hard drives could be powered by a 380w power supply.

---

The power supply I use is an Antek 380w something or other.

But the power supply used really does not matter much. The computer is going to cost about $100/year for power. The difference in a proper sized power supply or a more/less efficient power supply is noise in the total cost of setting up a computer.
 
GeorgeHR said:
I am amazed at how much power people think computers need.

I just rebooted my TV server, Asus motheroard with I3 CPU, 10 WD green drives. Booting up with 10 HDs the system peaked at 107w (Kill-A-Watt). The system idles at 77w and peaks at 110w at 100% load (Primes95).

My Kill-A-Watt appears to give the correct reading for a 100w light bulb. So I tend to trust it.

According to my simple test, 3 motherboards and 30 hard drives could be powered by a 380w power supply.

---

The power supply I use is an Antek 380w something or other.

But the power supply used really does not matter much. The computer is going to cost about $100/year for power. The difference in a proper sized power supply or a more/less efficient power supply is noise in the total cost of setting up a computer.
Click to expand...


Agreed. It's the oldest marketing trick they've been using since the P4 days when PSU manufacturers used the power hungry CPU (and later GPU's) as an excuse to throw as many watts as they could at consumers. Even places which give you the "max" wattage used per device in a system so you can calculate what you need, isn't very accurate. Most of the time the majority of you system is idling. You only need enough wiggle room in-case you most likely device needs more power for a longer time frame than the other devices. Those charts often tell people what each component uses at max and then people add them up and think that's what their system will use 24/7.

Which is why you can get away with PSU's that are 500W and your system on paper may say the max total wattage a system can use is 600W because the chance all your components are maxing out the power supply is rare. Most likely it's the CPU and GPU, and these days the CPU's are marginal in comparisons.
 
Danny Bui said:
For all intents and purposes, I'd just ignore the 80Plus rating or the whole thing in general:
http://www.hardocp.com/article/2011/10/04/80_plus_irrelevant_to_you_when_buying_psu/
Click to expand...

That article points interesting facts but I wouldn't conclude that 80Plus should be ignored, after all if a PSU hasn't 80+ there is not a chance it's a good PSU, unless you have specced/designed it yourself for your own use.

What I get is that a good 80+ rating doesn't make a good PSU, which I was already aware of, I don't intend to venture into buying from unknown brands and without reading other reviews.

Also about modular PSUs, there are advantages but also disadvantages, my current Cooler Master modular PSU had a "branch" burn because of a faulty contact, it also killed one expensive pump.

Now that I think of it, would it be actually possible to get redundancy ? I'm using the FSP Aurum 500W as an example, it seems to be a good PSU, one is replacing a Seasonic S12 I had since those came out (I even got it from Germany since it was available there first) and that finally died and seems to perform similarly, but more efficiently. It's also quite cheap.

So, what if I put two of those in the Norco, and powered them at the same time, would it help at the peak power draw of start-up, and then also provide redundancy if one dies ? Or is that risky for some reason ?

That would only be for the DAS, not the server.
 
Billy_nnn said:
If you can stagger the spinup, then the peak 12V demands are lowered, and you can use a lower rated PSU.
Click to expand...

The HBAs would be LSI based, probably the M1015 since I already have one of those. The HP expander is the most available and cheaper than the others too, to the point I could buy a spare one, so it's my choice. From what I understand, no staggered spinup with those.
 
The only way to make a properly redundant power supply system is with either a redundant control system or to get one of those redundant server power supplies. You cannot wire the 12v of one power supply with the 12v of another, due to differences in voltage regulation. One power supply will have a slightly higher voltage than the other, and will feed current into the lower voltage power supply, possibly creating a short circuit and damaging your power supplies.

Here is some good information.
 
GeorgeHR said:
My Kill-A-Watt appears to give the correct reading for a 100w light bulb. So I tend to trust it.
Click to expand...
Uh huh. Unfortunately it's been shown that while the KAW will be accurate with other devices, it won't be accurate when it comes to PSUs. Here's a wealth of info on it:
Here's Paul Johnson's post, PSU reviewer of HardOCP.com, about the inaccuracy of the Kill-A-Watt:
http://hardforum.com/showpost.php?p=1032190998&postcount=7

Another Paul Johnson's post:
http://www.jonnyguru.com/forums/showpost.php?p=27745&postcount=26

In addition, three other PSU experts backs up Paul Johnson's statement:
Oklahoma Wolf of JonnyGuru.com:
http://hardforum.com/showpost.php?p=1034843536&postcount=21
http://hardforum.com/showpost.php?p=1034555880&postcount=17

JonnyGuru of JonnyGuru.com and the former product manager for BFG acknowledge the inaccuracies of a KAW (Post #7 in regards to Post #2):
http://www.jonnyguru.com/forums/showthread.php?t=5977
http://www.jonnyguru.com/forums/showpost.php?p=29907&postcount=2

Redbeard of Corsair also acknowledges the inaccuracies:
http://hardforum.com/showpost.php?p=1032811067&postcount=22

Related discussions:
http://hardforum.com/showthread.php?t=1447774
http://hardforum.com/showthread.php?t=1509210
http://www.jonnyguru.com/forums/showthread.php?t=3102
http://www.jonnyguru.com/forums/showthread.php?t=2695
http://www.jonnyguru.com/forums/showthread.php?t=2284
GeorgeHR said:
According to my simple test, 3 motherboards and 30 hard drives could be powered by a 380w power supply.
Click to expand...
And it's fairly obvious that your simple test doesn't explain the whole story.

Aesma said:
What I get is that a good 80+ rating doesn't make a good PSU, which I was already aware of, I don't intend to venture into buying from unknown brands and without reading other reviews.
Click to expand...
Fair enough.
Aesma said:
So, what if I put two of those in the Norco, and powered them at the same time, would it help at the peak power draw of start-up, and then also provide redundancy if one dies ? Or is that risky for some reason ?
Click to expand...
No. While you can certainly power the DAS with two PSUs, you just won't get redundancy as Tsumi has shown.
 
Danny Bui said:
Uh huh. Unfortunately it's been shown that while the KAW will be accurate with other devices, it won't be accurate when it comes to PSUs.

And it's fairly obvious that your simple test doesn't explain the whole story.
Click to expand...
I never claimed it gave an accurate result. But for you to make the claim it is not accurate would require you to build a similiar system and make a measurement.

My UPS does give an accurate result. Charge the UPS to full power. Start up the computer. Unplug the UPS. Run at the computer at test load until the UPS shuts down. A bit of math gives a reasonable value for the wattage.

My UPS has a power display also. I am sure it is not accurate enough for you.

----

If you want to claim a better answer: Buy a $500 power meter. Build a system. Measure the power used.

Just using published claims is worthless. By the way the claims that the Kill-A-Wat is not accurate only applies to power supplies of a certain design. According to your links it appears that design is rare.

---

My Kill-A-Watt does accurately reflect the 3w difference for green hard drives going from idle to continuous reads. It also accurately reflects the power difference of various CPUs at full power. It even reflects the power used by my USB tuners well.

I am sure there are errors but since most people seem to think that 500-600w power supplies are reasonable for the OP, my claim seems not as far out of line as you make it out to be.
 
GeorgeHR said:
I never claimed it gave an accurate result.
Click to expand...
In roundabout way you did. You specifically said "My Kill-A-Watt appears to give the correct reading for a 100w light bulb. So I tend to trust it." So you basically said that you trust what readings that KAW give you due its correct measurement of a 100W light bulb. Thus, you're implying that it's accurate enough to you.
GeorgeHR said:
But for you to make the claim it is not accurate would require you to build a similiar system and make a measurement..
Click to expand...
No I can say it's not accurate because I have evidence and facts that show that the KAW itself can be inaccurate.
GeorgeHR said:
My UPS does give an accurate result. Charge the UPS to full power. Start up the computer. Unplug the UPS. Run at the computer at test load until the UPS shuts down. A bit of math gives a reasonable value for the wattage.

My UPS has a power display also. I am sure it is not accurate enough for you.

----

If you want to claim a better answer: Buy a $500 power meter. Build a system. Measure the power used.
Click to expand...
Went off the deep end here a bit don't you think? Anyway, a Brand Power meter is actually only $150, not $500.
GeorgeHR said:
Just using published claims is worthless. By the way the claims that the Kill-A-Wat is not accurate only applies to power supplies of a certain design. According to your links it appears that design is rare.
Click to expand...
Huh? The inaccuracy occurs with PSUs using APFC which any good PSU these days have. APFC is not rare these days at all.

GeorgeHR said:
I am sure there are errors but since most people seem to think that 500-600w power supplies are reasonable for the OP, my claim seems not as far out of line as you make it out to be.
Click to expand...
While I fully agree that people overspec on the PSU, what I have an issue with is people using data from their own KAW to justify going lower with the PSU. In other words, your general conclusion is right but you're backing it up with poor data.
 
I would recommend you hire homer simpson and have him and bart run power lines directly to you nuclear reactor, I mean hard drives.. That's goona sound nice when they all spin up..
 
Well I'm in France so newegg is not an option. If I was in the US I would probably be buying SE-3016 JBODs to be honest.

I like Seasonic but they're really expensive.

I've read the links so redundancy is out, however I could just plug 12 drives on one PSU and 12 on the other, or 16/16/16 across two cases.

When my Seasonic S12 died recently my server was down for 10 days because I wasn't about to power it with the noname PSUs I had lying around, and I had to read reviews and then order and wait, so I'm thinking 2 good PSUs are better than one excellent PSU.
 
Billy_nnn said:
FYI, the Corsair AX range are based on the Seasonic X Series!
Click to expand...

I think only certain models in the AX line are Seasonic sourced. Easily verifiable reading through some of the hardocp reviews.

and yeah, i'd second the corsair route if actual seasonic is not available to you.
 
You must log in or register to reply here.
Back
Top