PSU - Power Supply - Single Rail


NAS

Recommended Posts

Just when you think you've got it all figured out... there is an additional factor to consider.  

Noise immunity.

 

This is from one spec sheet of a disk I found in a quick search.

6f155d.jpg

 

Notice the wide allowance for the actual supply voltages Typically +/- 5%.

 

However, also notice the inability to handle noise on the power supply lines. Less than 100mV peak-to-peak.

On a 12 volt line, this means the regulation of the voltage must be better than 1%.    You can have a supply with plenty of capacity and poor high frequency regulation, and you'll get random errors.

 

I really like the engineering/marketing double-talk in "Note 2"

Note 2. In case of fluctuations in the DC power (spikes, momentary shut off, etc.), the data being recorded or to be recorded from buffer memory are not assured.

In other words, if there is ripple and noise on the supply lines greater than that allowed (1% or greater), the data being recorded or read "may not be assured"

 

Translated to English ... "may not be assured" = "may produce random results"

It is easy to see how adding one more drive to a system may still be within the wattage limits, but result in random errors.

 

The instruments used on the xbitlabs site to measure current samples several times per second.  I doubt if they gave the true "instantaneous" peak current.  To do that you need an oscilloscope, not a digital ammeter.

 

How many of the power supplies even give the ripple ratings.  I found a few, and they said 120mV p-p.  Not as good as  what the disk manufacturer said was needed.

 

Does anybody else not wonder why we see people with multi-disk arrays see errors start when their array size starts to grow?

 

Joe L.

  • Upvote 1
Link to comment

I looked up the ATX power supply ripple specification.  It is 120mV peak-to-peak on the 12 volt rail.

 

Granted, most supplies should be better than the specification, but assuming one is just at the edge of being in spec, it could still potentially cause errors when paired with the disk described above.

Link to comment
100 watts for a MB/CPU + 50 watts for fans

 

I have no idea where you buy parts but you should never need a MB/CPU that uses 100W these days if you're building a small system. My whole HTPC idles at 38W at the plug. Also, exactly how many fans are you using that they need 50W?

 

 

The instruments used on the xbitlabs site to measure current samples several times per second.  I doubt if they gave the true "instantaneous" peak current.  To do that you need an oscilloscope, not a digital ammeter.

 

Did you read their test methodology? They show how they use a scope and why a meter is no good.

 

Peter

 

Link to comment

100 watts for a MB/CPU + 50 watts for fans

 

I have no idea where you buy parts but you should never need a MB/CPU that uses 100W these days if you're building a small system. My whole HTPC idles at 38W at the plug. Also, exactly how many fans are you using that they need 50W?

Agreed... but many of us put together our server before low powered CPUs were available, and others use older MB because they already own them. (My server is about 4.5 years old)

 

The instruments used on the xbitlabs site to measure current samples several times per second.  I doubt if they gave the true "instantaneous" peak current.  To do that you need an oscilloscope, not a digital ammeter.

 

Did you read their test methodology? They show how they use a scope and why a meter is no good.

 

Peter

 

Yes, I read their methodology.   They describe how a scope is needed to track peak current usage here:

http://www.xbitlabs.com/articles/storage/display/hdd-power-cons_3.html#sect0

They also describe how the meter did not detect any true peak used.

 

The article I linked to said this about how they performed their measurements:

[b][color=maroon]We measured the +5 V and +12 V draw (accurate voltages at the output of the above mentioned unit were +5.08 V and +11.82 V) simultaneously with two digital ammeters of the 1.5 accuracy class with the resistance below 0.15 ohm (including the leads' resistance). The refresh rate of readings was approximately 0.3-0.4 sec. The table provides average values for several seconds (current fluctuations usually didn't exceed 30 mA), except for the Start-Up current (the table contains maximum values).[/color][/b]

 

Note they did not use the oscilloscope methodology listed on their site. They used two multimeters.

 

Joe L.

Link to comment

Xbitlabs did nice 2TB power consumption (startup/idle, watts/amps) graphs as well.

 

http://www.xbitlabs.com/articles/storage/display/2tb-7200rpm_18.html

 

It's partially the reason I went with Seagate LP 2TB's. While they need a lot of amps to start up, they idle at the lowest wattage of the available 2TB drives, which I find to be important for a media server.

 

Thanks for the link. I'm building a new unRAID server and wanted to use 2TB drives. Looks like the Seagate LP 2TB's will be my choice, too, albeit due to performance and not because of reduced power consumption during idle/read/write. The power savings would be hard to figure considering you're going to need a bigger power supply that will likely run at lower efficiency when the drives are idle (less than 20% loading). You'd need to see efficiency curves for the power supply you would have used for the WD20EADS drives and the one for the Barracuda LP 2TB drives to compare the difference in power usage at the wall and the degree of difference.

 

A point that needs to be considered,  Even if you have a staggered spin up, unRAID will do a mass spin up if you click the spin up button.

This in turn defeats any staggered spin up set by the controller or POWER UP IN STANDBY jumper/hdparm software setting.

 

In addition, if you are on the edge of power ratings, there will be intermittent drive drop outs just spinning up one drive.

 

Forgot about that part. :P

Link to comment

There are assumptions there.

 

Not everyone is using ECO/GREEN Drives. If they are mixed and matched, issues could arise.

 

The 120W may be a bit conservative. consideration is required for surge which could be 2x that.

Add to that, consideration for CPU, controller cards and fans.

 

Corsair's calculator recommends the 450W for 20 drives.

 

I would consider the Corsair 400W CX for a small 9-12 drive system.

Anything more, I would go higher.

 

I'm using a 500W Seasonic on a 9 drive capable system.

 

I'm at 8 drives now, when I add the 9th drive, The system becomes unstable.

 

Granted I'm not using a single rail CPU, Which is why I purchased the Corsair 400W CX.

With 8 drives, mixed types at 18A. I'm having stability issues when going to the 9th drive.

Most of the issue arises during spin up while the array has put drives to sleep.

For example Array has drives spun down. I do something on one drive, and it is spun up.

There is a chance the spin up will not be successful, the drive goes off line and the array kicks into redundancy mode.

 

if you use splitters of any kind that may case your problems. my friend runs a 1000w coolermaster and is havig problems with 24 hdds until i told him to stop using splitters. no more intermittant drives spinning down when another spins up etc.

Link to comment

Being slightly selfish i though i would ask for advice on dimesiong a PSU for my potential new build.

 

I will add this to the wiki as an exmaple of how to calculate when complete.

 

Opinions (alot of guesswork on this first draft)

 

 

AMD Sempron 140 2.7GHz Socket AM3 1MB Cache Retail Box Processor

Count,Amps,Total Amps: 1 0.2 0.2

 

Kingston 2GB (2x1GB) DDR2 1066MHz/PC2-8500 Hyperx Memory Non-ECC Cl5(5-5-5-15) Unbuffered

Count,Amps,Total Amps: 1 0.2 0.2

 

ASUS M4A78L-M AMD760G Socket AM2+ Gigabit Ethernet DVI VGA HDMI Out 8 Channel Audio MATX Motherboard

Count,Amps,Total Amps: 1 0.4 0.4

 

X-Case RM 413 4u Rackmount for 13 x 3.5 +5.25+LCD+Locking front door..

Count,Amps,Total Amps: 1 1 1

 

Supermicro AOC-SASLP-MV8, 8-Port SAS/SATA Card

Count,Amps,Total Amps: 1 0.2 0.2

 

Cruiser Flash

Count,Amps,Total Amps: 1 0.1 0.1

 

WD15EARS

Count,Amps,Total Amps: 13 1.75 22.75

 

Case fans

Count,Amps,Total Amps: 4 0.07 0.28

 

Total 25.13 Amps

 

Link to comment

if you use splitters of any kind that may case your problems. my friend runs a 1000w coolermaster and is havig problems with 24 hdds until i told him to stop using splitters. no more intermittant drives spinning down when another spins up etc.

 

In my case, I am not using splitters. It's a modest setup with a fairly decent CPU (2.6 Core 2 Duo) which is idle most of the time. I know there will be comments about the CPU. But I've tested a single core celeron rated at 35w. It did not save any power ???  Also limiting the 2.6 to 1.6ghz does not save any power.

 

I think the issue is the mixture of 3 seagate 7200RPM drives and 6 WD Green drives and a couple of controllers.

I'm going to switch out the 500W seasonic for a 400W corsair single rail and see how it goes.

 

If I put the last drive in an external sata case the system functions fine.

Link to comment

Being slightly selfish i though i would ask for advice on dimesiong a PSU for my potential new build.

 

I will add this to the wiki as an exmaple of how to calculate when complete.

 

Opinions (alot of guesswork on this first draft)

 

 

AMD Sempron 140 2.7GHz Socket AM3 1MB Cache Retail Box Processor

Count,Amps,Total Amps: 1 0.2 0.2

 

Kingston 2GB (2x1GB) DDR2 1066MHz/PC2-8500 Hyperx Memory Non-ECC Cl5(5-5-5-15) Unbuffered

Count,Amps,Total Amps: 1 0.2 0.2

 

ASUS M4A78L-M AMD760G Socket AM2+ Gigabit Ethernet DVI VGA HDMI Out 8 Channel Audio MATX Motherboard

Count,Amps,Total Amps: 1 0.4 0.4

 

X-Case RM 413 4u Rackmount for 13 x 3.5 +5.25+LCD+Locking front door..

Count,Amps,Total Amps: 1 1 1

 

Supermicro AOC-SASLP-MV8, 8-Port SAS/SATA Card

Count,Amps,Total Amps: 1 0.2 0.2

 

Cruiser Flash

Count,Amps,Total Amps: 1 0.1 0.1

 

WD15EARS

Count,Amps,Total Amps: 13 1.75 22.75

 

Case fans

Count,Amps,Total Amps: 4 0.07 0.28

 

Total 25.13 Amps

 

You're counting everything from the 12V rail. RAM, I think, gets power mostly from the 3.3V rail. The flash drive, from 5V. The fans will probably use more than 0.07A unless you get very slow moving ones (~500RPM). For example, the Scythe Slipstream 120mm 1200RPM is rated at 0.26A. The estimate for the motherboard, I think is very conservative. I'd figure at least 1A for the motherboard. Don't really know about the other parts since I can't find any specifications for them.

 

By the way, if you haven't bought those parts yet, skip the gaming/performance memory. Value memory (DDR2 800?) at 1.8V is a better option.

 

In my case, I am not using splitters. It's a modest setup with a fairly decent CPU (2.6 Core 2 Duo) which is idle most of the time. I know there will be comments about the CPU. But I've tested a single core celeron rated at 35w. It did not save any power ???  Also limiting the 2.6 to 1.6ghz does not save any power.

 

Did you lower voltage as well? Lowering the frequency isn't really going to help much. Undervolting is what lowers power consumption, albeit it's usually not very significant at idle.

Link to comment

In my case, I am not using splitters. It's a modest setup with a fairly decent CPU (2.6 Core 2 Duo) which is idle most of the time. I know there will be comments about the CPU. But I've tested a single core celeron rated at 35w. It did not save any power ???  Also limiting the 2.6 to 1.6ghz does not save any power.

 

Did you lower voltage as well? Lowering the frequency isn't really going to help much. Undervolting is what lowers power consumption, albeit it's usually not very significant at idle.

 

I was not able to on that motherboard.

Link to comment

Reading along this thread I've got a question bubbleing out of me.  I have the same motherboard mentioned above.  2-7200.11 1.5 TB drives, 3 7200.11 500GB drives.  I'm pretesting one drive so it's active but the others are sleeping.  I have my server plugged into  kill-a-watt meter so I can see what is being drawn from the wall.  It says 107 watts and 1.43 Amps.  When all the drives are spun down the watts settles down to 82 or so. 

 

My question is why am I only drawing 1.43 Amps from the wall?  Does the psu really output 15 times the current it takes in?

Link to comment

My question is why am I only drawing 1.43 Amps from the wall?  Does the psu really output 15 times the current it takes in?

Question #2, yes, but at 1/15th (average of all the rails) the voltage.

 

Amps * Volts = Watts

Watts / Volts = Amps

 

Pretending for an instant it was only a 5 volt supply

107 Watts / 5 Volts = 21.4 Amps

 

Pretending for an instant it was only a 12 volt supply

107 Watts / 12 Volts = 8.9 Amps

 

Pretending the load was split evenly between 12 and 5 Volt rails

53.5 Watts / 12 Volts = 4.45 Amps

53.5 Watts / 5 Volts = 10.7 Amps

 

4.45 Amps sound about right for the load on the 12 volt rail for the 5 disks you are running.

 

Joe L.

 

Link to comment
My question is why am I only drawing 1.43 Amps from the wall?

 

The AC power input = Volts x Amps x Power Factor. You'll notice the meter read 107W and 1.45A. Well, 1.43A times 120V = 172W, yet the meter read 107W. So, you can not use the AC current directly for power measurements without including the power factor term.

 

Does the psu really output 15 times the current it takes in?

 

Yes, it does. With power conversion equipment, the formula to be concerned about is Power Output = Power Input - Losses. Another variation is  Power Output = Power Input X Efficiency. So, the output power will always be less than the input power. At 107W input power the power supply is likely outputting around 80W.

 

Then, take the Power Output and divide by the DC Voltage and you get the DC Current.

 

Peter

 

Link to comment
  • 8 months later...

Hey Folks,

 

I'm new round here and i've been reading up. I havent completed my server build yet but i'm almost finished sourcing parts and i thought I'd throw this in the mix if anyone's has any comments. My final decision with power supply comes down to either the recommended Corsair CMPSU-650TXUK TX or the Coolermaster Silent Pro 600. I havent heard any talk of Coolermaster Power supplies but this one is modular and it seems to meet all the requirements. Is there any particular reason they havent been mentioned or has people just got preferred brands?

Link to comment

Hey Folks,

 

My final decision with power supply comes down to either the recommended Corsair CMPSU-650TXUK TX or the Coolermaster Silent Pro 600. I havent heard any talk of Coolermaster Power supplies but this one is modular and it seems to meet all the requirements. Is there any particular reason they havent been mentioned or has people just got preferred brands?

 

Typically price, awareness and trends.  The Coolermaster looks like a good PSU based on the multiple reviews I read.  The OEM is Enhance (French), not as common as the Seasonics of the world.  Specs look good...go for it to be different!

Link to comment
  • 2 weeks later...

what about the Seasonic S12II 520W

its about 60€ here in the Netherlands

and it seems it also effective when less power is drawn

http://www.seasonicusa.com/M12II-Bronze.htm

 

http://www.newegg.com/Product/Product.aspx?Item=N82E16817151094&cm_re=seasonic_s12ii-_-17-151-094-_-Product

 

 

reviews

http://www.hardwaresecrets.com/article/Seasonic-S12II-Bronze-520-W-Power-Supply-Review/954/10

http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=185

 

 

also the M version is modular (Seasonic M12II)

 

 

edit*

 

Confusingly their website states that it is a 2 rail design but according to the article on hardwaresecrets

 

As a final note, although Seasonic manufacturing quality is flawless, they should take care of the mistakes they are doing on the product label (listing two +12 V rails when in fact the unit has a single-rail design)

 

and on jonnyguru

 

You may be asking yourself, "Hey! He didn't say which 12V rail goes where! Why that no good so and so, I oughtta..."

 

To which I reply: calm down, people. Neither the manual nor the cabling made any indication of what went to what rail, so I popped the cover off. This unit does not have two 12V rails at all - it's a single 12V design. There is no multirail overcurrent protection in there. That'll make load testing easier.

Link to comment
  • 3 weeks later...

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.