PSU 12V Rails: Multi vs. Single

[JCNiest5]

Okay, you experts out there, as title says, which offers better stability or usability as far as Power Supplies is concerned? I see many manufacturers are claiming multi-rails is a more future-proofs as others are saying single rail remains the most common technology. Why should I choose multi-rails OR single rail instead?

As always, thanks in advance...

 

[Dangman]

Depends on the manufacturer, power supply in question, how the rails are distributed, etc.

The number of Rails should not be the only consideration when choosing the power supply. Have to factor in how much power total is available on the +12V rail, whether it can actually provide that power, efficiency, etc.

 

[JCNiest5]

So, which is better? Multi-rails or Single rail? That's my real question if I wasn't too clear in my thread starter post.

 

[RamonGTP]

Single rails are better becuase you get to use ALL the power available on the 12v rail.

From PCP&C Founder:

One 12V rail is better because all of the power supply's capacity is available to the system. With a multi-rail 12V design, as much as 30% of the PSU's capacity can be trapped on under-utilized rails. For example, if one 12V rail rated at 18A is for the CPU, and the CPU only draws 8A, the remaining 10A cannot be utilized by other components in the system.

http://www.anandtech.com/casecoolingpsus/showdoc.aspx?i=3222&p=2

 

[E.T]

I see many manufacturers are claiming multi-rails is a more future-proofs

... for them because you'll come back to buy new PSU sooner when you can't utilise full capacity of your current PSU so easily.


Most PSUs have only one transformer and "multiple" rails are completely virtual and nothing more than artificially current limited wire groups. (adding own transformer circuitry for them would add lot of complexity to already cramped PSU and costs)
Here's simplified diagram:
http://img401.imageshack.us/img401/382/12vrailsnj5.gif
Current values only for example and unlike in your house, overloading one "fuse"/current limiter shutdowns your whole house at once instead of just that output wire.

Lately high power PSUs have been having two smaller transformers (easier to fit than one bigger) but outputs of those seem to be normally first combined and then again split to wire groups with artificial current limiters. Only exceptions to this (I know) are Enermax Galaxys but they again dropped the bowling ball on their toes by distributing output of two 12V transformers so that only other transformer/half of 12V capacity is available for other parts than CPU(s).


Reason behind this deceit is that members of certain group hit their heads together (too hard) and decided that "240VA is enough for everyone" which lead to 20A current limit in 12V. With rising power needs and output PSU makers just couldn't rise that limit for retaining compliance with "standard" so they added second wire group with own current limiter et cetera. But instead of telling the truth that you can't draw full output from any combination of connectors you want they decided to advertise it as new "Jesus feature".
Bottom line: It's 100% full kludge for circumventing artificial limit. Also it's very effective in hiding real 12V output capacity when Joe Average Vegetable... err Consumer stares number of "multiple rails"

And actually this 240VA limit requirement has been removed in silence couple years ago but most PSU makers still cling to it because they don't want to admit that all what they said about it was basically BS.
Seasonic made PSUs of last couple years have been without these artificially current limited wire groups, despite what sticker says. You'll find hint to that from UL Level 6 marking under UL sign (UL sign is that "mirrored" LR like sign) which tells they aren't bothered by this 240VA limit and if wanted, push out full 12V output through any connector... if cable and connector just don't burn away.

Also latest Corsair branded PSUs have been officially without limiter (Like PC P&Cs) and I know they won't be left as last brand to do that.

 

[JCNiest5]

Single rails are better becuase you get to use ALL the power available on the 12v rail.

From PCP&C Founder:

One 12V rail is better because all of the power supply's capacity is available to the system. With a multi-rail 12V design, as much as 30% of the PSU's capacity can be trapped on under-utilized rails. For example, if one 12V rail rated at 18A is for the CPU, and the CPU only draws 8A, the remaining 10A cannot be utilized by other components in the system.

http://www.anandtech.com/casecoolingpsus/showdoc.aspx?i=3222&p=2

Got it..that's the answer I was looking for. Now, single rail here I come!

 

[jonnyGURU]

Okay, you experts out there, as title says, which offers better stability or usability as far as Power Supplies is concerned?

Neither. Split or single +12V rail PSU's typically have absolutely no bearing on stability, etc. because the split is almost always done off the same +12V source.

Depends on the manufacturer, power supply in question, how the rails are distributed, etc.

The number of Rails should not be the only consideration when choosing the power supply. Have to factor in how much power total is available on the +12V rail, whether it can actually provide that power, efficiency, etc.

So, which is better? Multi-rails or Single rail? That's my real question if I wasn't too clear in my thread starter post.

Danny told you... it depends on the particular unit and how the connectors are distributed. If single vs. multiple +12V rail is actually a decision making criteria for a power supply and not it's capability to provide power, efficiency, etc. I'm going to ask that you set the screwdriver down and step away from the computer.

Got it..that's the answer I was looking for. Now, single rail here I come!

And PCP&C's marketing FUD snags another lemming.

Naturally something you find on PCP&C's website is going to try to sway you towards single +12V rail since that is what they sell.

 

[Mithent]

I'm pretty confused about the HX620W's 'rails'. It only has one +12V transformer, and then the PCB claims there are +12V1 and +12V2, and then the modular interface and the sticker say there are three rails. Then I hear there's no OCP on these rails - so that basically means it's a single-rail power supply?

 

[WhoBeDaPlaya]

Correct. Same thing for the HX520

 

[E.T]

I'm pretty confused about the HX620W's 'rails'. It only has one +12V transformer, and then the PCB claims there are +12V1 and +12V2, and then the modular interface and the sticker say there are three rails. Then I hear there's no OCP on these rails - so that basically means it's a single-rail power supply?

Like jonnyGURU just said practically all PSUs are single 12V source designs in reality and it's just split to separate wire groups with current limiters.
But Corsairs HXs are one of the UL Level 6 PSUs made by Seasonic.

Looking at the Seasonic main PCB inside the Corsair PSU reveals only two rails, labeled 12V1 and 12V2. There is no third rail. This is illustrated well at Hardware Secrets. Although I can not say that these rails are or are not somehow electronically separated in the PSU's circuitry somewhere, I did find that there was no OCP (over current protection or "limiter") on either of these rails as I was able to load any given connector up 30 to 40A with no drop in voltage, system shut down, etc.
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story2&reid=21
Further info about design and why there are 12V1 and 12V2 markings on PCB is in review of M12 700W



And I'm sure that jonnyGURU has already said it enough many times why most of PC P&C's "multiple vs single rail" claim is marketing BS so here's it again for letting him rest this time:
All 12V comes from one source so all of it is available to any wire group and current limits of those are generally high enough that even if you put only couple ampere load to one of them you can draw "unused" amperes from other wiregroups and it just makes drawing bigger currents harder because you can't draw it from all combinations of cables/connectors.
http://img401.imageshack.us/img401/382/12vrailsnj5.gif

Only exception to this are Enermax Galaxys in which outputs of two smaller 12V sources/transformers aren't combined and there are two separate groups of these virtual rails:
http://www.bjorn3d.com/read.php?cID=960&pageID=2546

 

[[Spectre]]

Single rails are better becuase you get to use ALL the power available on the 12v rail.

From PCP&C Founder:

One 12V rail is better because all of the power supply's capacity is available to the system. With a multi-rail 12V design, as much as 30% of the PSU's capacity can be trapped on under-utilized rails. For example, if one 12V rail rated at 18A is for the CPU, and the CPU only draws 8A, the remaining 10A cannot be utilized by other components in the system.

http://www.anandtech.com/casecoolingpsus/showdoc.aspx?i=3222&p=2

... for them because you'll come back to buy new PSU sooner when you can't utilise full capacity of your current PSU so easily.


Most PSUs have only one transformer and "multiple" rails are completely virtual and nothing more than artificially current limited wire groups. (adding own transformer circuitry for them would add lot of complexity to already cramped PSU and costs)
Here's simplified diagram:
http://img401.imageshack.us/img401/382/12vrailsnj5.gif
Current values only for example and unlike in your house, overloading one "fuse"/current limiter shutdowns your whole house at once instead of just that output wire.

Lately high power PSUs have been having two smaller transformers (easier to fit than one bigger) but outputs of those seem to be normally first combined and then again split to wire groups with artificial current limiters. Only exceptions to this (I know) are Enermax Galaxys but they again dropped the bowling ball on their toes by distributing output of two 12V transformers so that only other transformer/half of 12V capacity is available for other parts than CPU(s).


Reason behind this deceit is that members of certain group hit their heads together (too hard) and decided that "240VA is enough for everyone" which lead to 20A current limit in 12V. With rising power needs and output PSU makers just couldn't rise that limit for retaining compliance with "standard" so they added second wire group with own current limiter et cetera. But instead of telling the truth that you can't draw full output from any combination of connectors you want they decided to advertise it as new "Jesus feature".
Bottom line: It's 100% full kludge for circumventing artificial limit. Also it's very effective in hiding real 12V output capacity when Joe Average Vegetable... err Consumer stares number of "multiple rails"

And actually this 240VA limit requirement has been removed in silence couple years ago but most PSU makers still cling to it because they don't want to admit that all what they said about it was basically BS.
Seasonic made PSUs of last couple years have been without these artificially current limited wire groups, despite what sticker says. You'll find hint to that from UL Level 6 marking under UL sign (UL sign is that "mirrored" LR like sign) which tells they aren't bothered by this 240VA limit and if wanted, push out full 12V output through any connector... if cable and connector just don't burn away.

Also latest Corsair branded PSUs have been officially without limiter (Like PC P&Cs) and I know they won't be left as last brand to do that.

Got it..that's the answer I was looking for. Now, single rail here I come!

I guess I am going to have to sticky this because the FUD being spread by PC Power and Cooling just keeps getting worse.

One last time:

Look at this power supply and assume it really is a multirail OCP limited unit: http://www.newegg.com/Product/ShowI...800+800W+Patent+Piperock+Modular+Power+Supply

We see it has a 12v capacity of 64A yet each of the 6 12v rails is rated for 20A. Well if we just add that we get 120A. 120Ax12v=1440w which certainly is not possible from a 800w power supply that we know is capped at 64A on the 12v rails. Because of this there is more capacity on each individual rail than is generally necessary for any device that will be on that rail and there is no real issue with running out of capacity on a given rail because the sum of the rails is greater than the real total cap as we have just seen. So is power really trapped...well no not really since if 12v1 needs 20A it can draw 20A even if 12v2 is capable of 20A but is unloaded or lightly loaded. This can occur in any arrangement of required power across the 6 rails up to the 64A cap in total and up to the 20A OCP limit for each rail. So power never real stays anywhere as it is not being delivered unless the components draw it in which case the power supply delivers the necessary amount and then it is governed by the same limits as a single 12v rail in regards to OCP.


There is no trapped power


PC Power and Cooling was big on the multi12v rail bandwagon....until they miss allocated their 12v rails with their true discrete multiransformer/multirail setup and got burnt. So now their marketing indicates that only a single 12v rail unit is correct....which is not true obviously...if you do multi12v rails correctly they are fine. PC Power and Cooling just didn't:

This is the wrong way to allocate the 12v rails and connectors with the design that Win-Tact provided PC Power and Cooling.

 

[RamonGTP]

What is the point of multi rails then? What happens when you exceed the load on one of the 12v rails but the other still has some left, like the example given by PCPC?

I always look at multi rails like the circuit breakers in your home. You have multiple 15a circuits, but they're all coming from the same source. That source, just like the power supply cannot support loading ALL of the circuits to full capacity.

However, if you load ONE of the circuits to capacity, that one circuit will trip, even though the main power source still has plenty of juice left, so what you end up having to do is plug some of your devices into a different outlet that is branching off a different circuit to balacne the load.

Seems to me that multi 12v rails are doing the EXACT same thing, so (to me) there does seem to be some merit to what PCPC is saying.

 

[[Spectre]]

like the example given by PCPC?

Seems to me that multi 12v rails are doing the EXACT same thing, so (to me) there does seem to be some merit to what PCPC is saying.

Look at this power supply and assume it really is a multirail OCP limited unit: http://www.newegg.com/Product/ShowI...800+800W+Patent+Piperock+Modular+Power+Supply

We see it has a 12v capacity of 64A yet each of the 6 12v rails is rated for 20A. Well if we just add that we get 120A. 120Ax12v=1440w which certainly is not possible from a 800w power supply that we know is capped at 64A on the 12v rails. Because of this there is more capacity on each individual rail than is generally necessary for any device that will be on that rail and there is no real issue with running out of capacity on a given rail because the sum of the rails is greater than the real total cap as we have just seen. So is power really trapped...well no not really since if 12v1 needs 20A it can draw 20A even if 12v2 is capable of 20A but is unloaded or lightly loaded. This can occur in any arrangement of required power across the 6 rails up to the 64A cap in total and up to the 20A OCP limit for each rail. So power never real stays anywhere as it is not being delivered unless the components draw it in which case the power supply delivers the necessary amount and then it is governed by the same limits as a single 12v rail in regards to OCP.

 

[Mithent]

Further info about design and why there are 12V1 and 12V2 markings on PCB is in review of M12 700W

Ah, thanks for that - I'd read jonnyGURU's review of the HX620W but hadn't seen that one.

I know that nearly all only have a single +12V transformer, but that doesn't mean that nearly every power supply is single-rail in the common usage of the term, it depends on how they're separated and OCP applied? I meant that the HX620W seems to only be effectively single-rail in that there's no real separation or OCP, and the claimed rails are really just made up, which isn't the case for every PSU with rails.

 

[jonnyGURU]

What is the point of multi rails then?

Slow shorts can be seen as loads and the short circuit protection in the PSU may not see it as a short and may not shut the PSU down. With the "limit" on the rails, such a short will create a load that exceeds the OCP, over current protection, and prevent the wire from overheating, causing the insulation to melt, causing small fires inside your chassis.

What happens when you exceed the load on one of the 12v rails but the other still has some left, like the example given by PCPC?

Nothing. The example given by PCP&C really doesn't exist. The only time you would really run into that scenario is if you had a PSU with two +12V rails, one goes to only the CPU and the other only has 18A on it. Say the combined +12V rating is 30A. If the CPU is only using 6A, then there is 6A of unusable power. But this is a rare arrangement and unless you were running SLI on a 500W power supply, probably not something you would ever see in the real world.

In the real world, you have examples like the one Paul offers above.

I always look at multi rails like the circuit breakers in your home. You have multiple 15a circuits, but they're all coming from the same source. That source, just like the power supply cannot support loading ALL of the circuits to full capacity.

Exactly.

However, if you load ONE of the circuits to capacity, that one circuit will trip, even though the main power source still has plenty of juice left, so what you end up having to do is plug some of your devices into a different outlet that is branching off a different circuit to balacne the load.

Right. But why should you be allowed to overload that circuit and burn your house down? It is very much the same thing and if we were to remove those breakers from your house and you plugged a bunch of crap into a power strip and overloaded that circuit and overheated the wiring, there would be flames shooting out of your wall... no?

 

[Basalt]

So to rephrase the original question what's better?

A power supply with a single line that can deliver up to full PSU capacity on the rail (obviously a heavier guage rail) as needed? Or a power supply that limits amperage power along multiple rails at a lower net values?

Seems to me, after calculating component demands, the only real world possibility of currently of getting caught under powered is if, say a 20 amp rail is insufficient to efficiently meet the demands of a particular component setup on like say a power hungry sli setup or a large drive array.

But then the modern mother board offers an optional 12v connection allowing a second rail to supply video power, and multiple rails can feed multiple drives.

So in reality, since larger capacity drives tend to offer more connection options, the real question becomes what is my net power need, and how efficiently can the PSU meet that need. That and is the build quality of the PSU capable of meeting the claimed capacity.

Single rail vs multi rail becomes totally moot?

Am I grasping this correctly?

 

[E.T]

So to rephrase the original question what's better?
A power supply with a single line that can deliver up to full PSU capacity on the rail (obviously a heavier guage rail) as needed? Or a power supply that limits amperage power along multiple rails at a lower net values?

If cables/connectors are distributed well to virtual rails and you won't be using any extreme/out of normal components they shouldn't cause any problem. In desktop/gaming PCs those componets include CPU and graphic cards.
In case of file server and without staggered spin-up also HDDs could overload virtual rail because hungriest drives draw nearly 3A peak from 12V meaning ten of them could trip OCP.


In case of fault situation "multiple rail PSUs" are safer because they shutdown more easily while "single rail PSU" needs much lower resistance situation to notice it as short circuit.
Already that 240VA can do lot of burning so somewhat higher limit than that, like ~30A (starts to be common in rails for connectors of graphic cards) probably doesn't lower safety much.
I could try asking opinion of Safety Technology Authority to that but one thing is sure:
When we go to these PSUs with near 100A or over in 12V tripping OCP starts to require short circuit comparable to soldering 12V and ground wires together meaning shorted component probably gets burned away sooner than PSU shutdowns!

...the real question becomes what is my net power need, and how efficiently can the PSU meet that need.That and is the build quality of the PSU capable of meeting the claimed capacity.

Build quality affects always, regardless age of PSU design (old 12V weak or modern 12V strong) and is 12V split to multiple virtual rails or not.
And in case of low quality PSU they can lack whole OCP totally (and other protections) and continue until something breaks/blows up in PSU... fortunately they would do that often much before advertised output capacity.

I know that nearly all only have a single +12V transformer, but that doesn't mean that nearly every power supply is single-rail in the common usage of the term, it depends on how they're separated and OCP applied?

Exactly.
Instead of only one big OCP in 12V source most PSUs are equipped with additional lower set limiters (commonly 18A or 20A) and then wire groups connected to these are named as 12V1, 12V2, 12V3, 12V4 etc.

 

[RamonGTP]

Slow shorts can be seen as loads and the short circuit protection in the PSU may not see it as a short and may not shut the PSU down. With the "limit" on the rails, such a short will create a load that exceeds the OCP, over current protection, and prevent the wire from overheating, causing the insulation to melt, causing small fires inside your chassis.



Nothing. The example given by PCP&C really doesn't exist. The only time you would really run into that scenario is if you had a PSU with two +12V rails, one goes to only the CPU and the other only has 18A on it. Say the combined +12V rating is 30A. If the CPU is only using 6A, then there is 6A of unusable power. But this is a rare arrangement and unless you were running SLI on a 500W power supply, probably not something you would ever see in the real world.

In the real world, you have examples like the one Paul offers above.



Exactly.



Right. But why should you be allowed to overload that circuit and burn your house down? It is very much the same thing and if we were to remove those breakers from your house and you plugged a bunch of crap into a power strip and overloaded that circuit and overheated the wiring, there would be flames shooting out of your wall... no?

Well of course, if a PSU has a single large rail, it should be desigend to operate safely at that load. I'm certainly not condoning unsafe PSU's. PCPC tests their PSU's to capacity at 50C load, so i'm sure that isn't an issue. My point is simply that you cannot simply dismiss what they are saying so blindly, the situation they are discribing can happen and HAS happened.

http://www.extremetech.com/article2/0,1697,1932958,00.asp

 

[jonnyGURU]

Well of course, if a PSU has a single large rail, it should be desigend to operate safely at that load. I'm certainly not condoning unsafe PSU's. PCPC tests their PSU's to capacity at 50C load, so i'm sure that isn't an issue. My point is simply that you cannot simply dismiss what they are saying so blindly, the situation they are discribing can happen and HAS happened.

http://www.extremetech.com/article2/0,1697,1932958,00.asp

I don't dismiss it "blindly." I am far from blind of the matter.

SilverStone PSU's were notorious for shutting down under heavy load because the ST65ZF only had two +12V rails with the CPU being on one rail and everything else on another and the ST60F put all of the PCIe connectors on the same +12V rail (yet split the CPU up, via the 8-pin, across two rails!) You can't dismiss the valid purpose of multiple +12V rails because one or two companies didn't spread the connectors out right. That's like dismissing using zeppelins as a feasible means of transportation because of the Hindenburg. Oh... wait....

There are plenty of PSU's out there that have multiple +12V rails that do not have a problem. Thermaltake Toughpower, Cooler Master Real Power Pro, Etasis Gaming Series.....

The real reason for going single +12V rail? Laziness and to save money. No thought has to be put into rail distribution and it costs less to not have the OCP circuitry in place.

What I dismiss is FUD. PCP&C spreads FUD. I'm not an advocate for or against single +12V rails. I just oppose FUD. When I helped design the Ultra X3, it was my suggestion to go single +12V rail. So maybe I'm being devil's advocate here, but I didn't try to spread a bunch of marketing FUD telling people that if the +12V was split up across four +12V rails that you would have "trapped power." I opted for a single +12V rail because it's easier to design your wiring diagram when you don't have to think about what connector should go on what rail, you don't have to explain to people that think "their graphics card requires 26A on the +12V rail" that they have to look at the "maximum combined +12V rating" of the PSU, and it saves money on the BOM because you get to leave about $1.50 worth of parts out of the unit.

The fact that PCP&C test their PSU's to capacity at 50°C has nothing to do with safety or multiple +12V rails. The wire and insulation PCP&C use is no different than most other PSU's. On a short, the wire will act as a slow burn fuse and the insulation can melt and start a fire. Sure, shove a paper clip across the +12V and ground and the PSU shuts itself down immediately, but let's try that short with something slow burn that doesn't trip the SCP.

 

[[Spectre]]

Well of course, if a PSU has a single large rail, it should be desigend to operate safely at that load. I'm certainly not condoning unsafe PSU's. PCPC tests their PSU's to capacity at 50C load, so i'm sure that isn't an issue. My point is simply that you cannot simply dismiss what they are saying so blindly, the situation they are discribing can happen and HAS happened.

http://www.extremetech.com/article2/0,1697,1932958,00.asp

Yeah PC Power and Cooling did it wrong too which is why they are now on the marketing path they are on rather than redoing the design, the thing is properly designed untis don't do it and 99.9999999999999% of people will never run into the issue. Even with the FUBAR'd units.

 

[Dan_D]

SilverStone PSU's were notorious for shutting down under heavy load because the ST65ZF only had two +12V rails with the CPU being on one rail and everything else on another and the ST60F put all of the PCIe connectors on the same +12V rail (yet split the CPU up, via the 8-pin, across two rails!)

About a year later, I have my answer as to why my friends computer kept shutting down under load. I went through two of those power supplies and he ended up with a PC Power & Cooling unit instead. (Which worked.)