Why a single rail power supply is superior to multiple rails

[HeXploiT]

I found an excellent example of why 1 single 12v rail is superior to multiple rails.
Here's a quote from PC Power & Cooling's website:


Here are the facts: A large, single 12-volt rail (without a 240VA limit) can transfer 100% of the 12-volt output from the PSU to the computer, while a multi-rail 12-volt design has distribution losses of up to 30% of the power supply?s rating. Those losses occur because power literally gets ?trapped? on under-utilized rails. For example, if the 12-volt rail that powers the CPU is rated for 17 amps and the CPU only uses 7A, the remaining 10A is unusable, since it is isolated from the rest of the system.

Since the maximum current from any one 12-volt rail of a multiple-rail PSU is limited to 20 amps (240VA / 12 volts = 20 amps), PCs with high-performance components that draw over 20 amps from the same rail are subject to over-current shutdowns. With power requirements for multiple processors and graphics cards continuing to grow, the multiple-rail design, with its 240VA limit per rail, is basically obsolete. [/i]

---Makes perfect sense to me.
So basically a lesson we can take from this is that if we are using multiple rail power supplies the best thing we can do is spread components out across all rails. If all rails on a multiple rail psu are thoughtfully spread across a system this argument could be rendered void however I know there are a great many people who haven't taken this into consideration and even purposefully use as few rails as possible in order to minimize cable clutter.

 

[jonnyGURU]

You have to realize that PCP&C writes myths to help perpetuate the market. They just started doing single +12V rail because they weren't good at the whole multi-+12V rail thing (they had Ginormous 1kW PSU's shutting down with mere dual 7x SLI set ups!) and wrote the "myth" to correspond with that migration.

PCP&C myths are marketing and should not be quoted as fact.

 

[jonnyGURU]

Originally posted by: Perry404
Here are the facts: A large, single 12-volt rail (without a 240VA limit) can transfer 100% of the 12-volt output from the PSU to the computer, while a multi-rail 12-volt design has distribution losses of up to 30% of the power supply?s rating.

Those losses occur because power literally gets ?trapped? on under-utilized rails. For example, if the 12-volt rail that powers the CPU is rated for 17 amps and the CPU only uses 7A, the remaining 10A is unusable, since it is isolated from the rest of the system.
[/quote]

FUD. Your typical dual +12V rail is problematic because it dedicates a good deal of power to the CPU via the 4-pin or 8-pin that can not be "accessed" by other connectors. But "modern" power supplies engineered for high end systems have four or more +12V rails and the sum of the capability of all of them typically exceeds the +12V output capability of the unit by 20%. You could have a rail with ZERO LOAD on it, and still have 100% of the capability of the +12V available to the other three rails.

Originally posted by: Perry404
Since the maximum current from any one 12-volt rail of a multiple-rail PSU is limited to 20 amps (240VA / 12 volts = 20 amps), PCs with high-performance components that draw over 20 amps from the same rail are subject to over-current shutdowns. With power requirements for multiple processors and graphics cards continuing to grow, the multiple-rail design, with its 240VA limit per rail, is basically obsolete. [/i]

I would call our friends at PCP&C to ask them what "high-performance components" they're talking about that use more than 20A per connector? Or rail? Or over all? Even a high end PCI Express 2.0 graphics card with a 6-pin and 8-pin power connector can only draw 18.75 from those connectors (combined, not each) IF you drew as much power as those connectors are capable of delivering. Of course, we know that these cards really don't even use half that. And even if you count what comes through the slot (as much as 75W) that would typically come off a separate rail (the one the 24-pin uses.)

There's certainly nothing wrong with single +!2V rail PSU's or multiple +12V rail PSU's. I could go either way. What I'm opposed to is people spreading FUD in order to make a sale.

 

[HeXploiT]

Well I guess the next question I would ask is what is the definition of a "rail"?
If you have two rails splitting 500 watts that means each rail can only handle a capacity of 250watts no?
In other words, if all 500 watts can go on either rail do you not in fact have only one rail? I thought the entire purpose/definition of a rail was to split the current.

 

[Sheninat0r]

Originally posted by: jonnyGURU
There's certainly nothing wrong with single +!2V rail PSU's or multiple +12V rail PSU's. I could go either way. What I'm opposed to is people spreading FUD in order to make a sale.

So basically, you're opposed to every company ever that markets their product?

 

[HeXploiT]

Originally posted by: Sheninat0r

Originally posted by: jonnyGURU
There's certainly nothing wrong with single +!2V rail PSU's or multiple +12V rail PSU's. I could go either way. What I'm opposed to is people spreading FUD in order to make a sale.

So basically, you're opposed to every company ever that markets their product?

HA!:laugh:

 

[Mr Fox]

Originally posted by: Sheninat0r

Originally posted by: jonnyGURU
There's certainly nothing wrong with single +!2V rail PSU's or multiple +12V rail PSU's. I could go either way. What I'm opposed to is people spreading FUD in order to make a sale.

So basically, you're opposed to every company ever that markets their product?

There is a very big difference between a large steaming pile of fecal matter, and responsible sales, and marketing.


It is what it Is.... as much as what PCP&C was once a market leader, their marketing has been almost as far out there as OCZ now that they are partners the steaming pile has grown in my humble opinion.

But ... PCP&C still makes some of the better hard wired PSU's out there if they survive the partnership.

Companies that market by Viral Methods are generally viewed as marginal by folks that are truly enthusiasts.

In high end hardware generally you get what you pay for....

 

[jonnyGURU]

Originally posted by: Sheninat0r

Originally posted by: jonnyGURU
There's certainly nothing wrong with single +!2V rail PSU's or multiple +12V rail PSU's. I could go either way. What I'm opposed to is people spreading FUD in order to make a sale.

So basically, you're opposed to every company ever that markets their product?

Haha!!!! Nice!

Originally posted by: Perry404
Well I guess the next question I would ask is what is the definition of a "rail"?
If you have two rails splitting 500 watts that means each rail can only handle a capacity of 250watts no?
In other words, if all 500 watts can go on either rail do you not in fact have only one rail? I thought the entire purpose/definition of a rail was to split the current.

Ah.. So you don't understand the concept to begin with, therefore believe the FUD. That's exactly what PCP&C count on.

When it comes to splitting up a +12V rail into multiple +12V rails, they don't just take the source and divide it up across multiple rails. All they're doing is limiting how much amperage is delivered to anyt particular connector or connectors.

If they DID evenly divided the source, there REALLY WOULD BE "trapped" power in a particular rail. That's what makes this "myth" such FUD. Because the sum of your rails is ALWAYS greater than the actual output of the source. In other words, there is an overlap of capability that compensates for any potential power that could be "trapped" by any one rail.

Take a look at ANY multiple +12V rail PSU and you'll see what I mean....

Rosewill Xtreme 850W: Four +12V rails @ 20A each. 4 x 20A = 80A. 80A * 12V = 960W. The source can only do 768W. 960W > 768W.

Gigabyte Odin GT 550W: Four +12V rails @ 18A each. 4 x 18A = 72A. 72A * 12V = 864W. Combined +12V capability is only 492W! 864W > 492W.

Look, fact is that any one +12V "rail" can put out ANY amount of power as long as it doesn't exceed the capability of the source. Typically, these rails are 20A or less because that's what's deemed "safe" by safety standards. But unless the PSU manufacturer puts too many connectors on one "rail" or doesn't have enough rails, there isn't a problem.

 

[4x4expy]

I bought the Silencer 610w for my new system. Only after did I find your site and learn the little that I now understand about PS design and function. Still no regrets in my purchase. I don't think PCPnC has hit rock bottom yet. But being better informed, now my pool of PS's for final selection would look different.

 

[jonnyGURU]

Originally posted by: 4x4expy
I bought the Silencer 610w for my new system. Only after did I find your site and learn the little that I now understand about PS design and function. Still no regrets in my purchase. I don't think PCPnC has hit rock bottom yet. But being better informed, now my pool of PS's for final selection would look different.

You shouldn't regret buying a Silencer 610W and I doubt PCP&C will ever hit "rock bottom."

They have a strong engineering team and good testing equipment. Nothing slips by them. Only if OCZ fires everyone but Doug will PCP&C be in trouble.

 

[johanh13]

Hello jonnyGURU,

I my opinion, this is not simply a myth. It is logical and sensible and can easily be proven. Imagine that you have four houses that are fed by a single transformer. The transformer is rated at 20KW. Now let us assume that three houses are using 1KW each, while the fourth is using 17KW. No problems, everything is fine. This is the power and flexibilty of a single rail design.
Now imagine that each house has its own 5KW transformer. No single house can exceed a power draw of 5KW! In this case there would be only an 8KW total power draw (leaving 12KW "trapped"), however, house number four would be non-functional.
In the case of the PCP&P 610 watt Silencer, it has a 49 amp 12 volt rail, I realize that you could not pull the entire 49 amps off a single 12v connnector. What I really like about a single rail is that it is a more balanced load. I have used PCP&C power supplies for years and have not been let down yet...

Originally posted by: jonnyGURU
You have to realize that PCP&C writes myths to help perpetuate the market. They just started doing single +12V rail because they weren't good at the whole multi-+12V rail thing (they had Ginormous 1kW PSU's shutting down with mere dual 7x SLI set ups!) and wrote the "myth" to correspond with that migration.

PCP&C myths are marketing and should not be quoted as fact.

 

[jonnyGURU]

Originally posted by: johanh13
Hello jonnyGURU,

I my opinion, this is not simply a myth. It is logical and sensible and can easily be proven.

I didn't say it was a "myth". I said it was blown out of proportion for the sake of marketing.

Originally posted by: johanh13

Imagine that you have four houses that are fed by a single transformer. The transformer is rated at 20KW. Now let us assume that three houses are using 1KW each, while the fourth is using 17KW. No problems, everything is fine. This is the power and flexibilty of a single rail design.
Now imagine that each house has its own 5KW transformer. No single house can exceed a power draw of 5KW! In this case there would be only an 8KW total power draw (leaving 12KW "trapped"), however, house number four would be non-functional.
In the case of the PCP&P 610 watt Silencer, it has a 49 amp 12 volt rail, I realize that you could not pull the entire 49 amps off a single 12v connnector. What I really like about a single rail is that it is a more balanced load. I have used PCP&C power supplies for years and have not been let down yet...

Originally posted by: jonnyGURU
You have to realize that PCP&C writes myths to help perpetuate the market. They just started doing single +12V rail because they weren't good at the whole multi-+12V rail thing (they had Ginormous 1kW PSU's shutting down with mere dual 7x SLI set ups!) and wrote the "myth" to correspond with that migration.

PCP&C myths are marketing and should not be quoted as fact.

Bro.... Not to sound arrogant, but you don't have to explain PSU theory to me. I'm not dogging PCP&C's quality at all. See my last post in this thread. I only questioning their marketing tactics. They justify everything they do by stating what everyone else does as wrong.

Besides, your analogy is seriously flawed and shows that you not only didn't read my fourth post, you didn't fully read my third post nor do you comprehend rail distribution in a PSU with multiple +12V rails.

A more accurate analogy would be a transformer that's rated at 20KW with each of four houses capable of drawing 10KW each. Of course, in the real world this would never happen because the houses have the potential to overload the transformer. But it does happen in the real world of power supplies because you DON'T want power "trapped" and you do want as much power as possible available to all of the connectors without exceeding a specified limit.

When they split up rails, the NEVER take 80A, for example, and split it into four 20A rails. Because if they did, that would trap a lot of potential power and in that case what PCP&C's marketing would really hold water. But again, it's NOT what they do and you should know that before coming into a discussion armed with poor analogies. I'm not going to reiterate what I said. Just go back and read my posts and the examples I provide.

 

[SerpentRoyal]

A good example of a high quality PSU is an automotive battery. As long as the battery is fully charged, it should deliver a lot of clean current and voltage via the anode and cathode. No BS multiple rails artificially capped with current-limiting electronics. KISS. If you try to draw a lot of current with small diameter cable, then the cable will overheat and break the circuit.

It costs a little more $ to design a PSU with a sigle high current +12VDC output tap. The PSU must be able to handle low impedance load since there will be a lot of devices connect in parallel to the +12VDC output tap.

 

[jonnyGURU]

Originally posted by: SerpentRoyal
A good example of a high quality PSU is an automotive battery. As long as the battery is fully charged, it should deliver a lot of clean current and voltage via the anode and cathode. No BS multiple rails artificially capped with current-limiting electronics. KISS. If you try to draw a lot of current with small diameter cable, then the cable will overheat and break the circuit.

Right.... the wire breaks the circuit by overheating and dripping flaming insulation all over the inside of your chassis. :roll:

Originally posted by: SerpentRoyal
It costs a little more $ to design a PSU with a sigle high current +12VDC output tap. The PSU must be able to handle low impedance load since there will be a lot of devices connect in parallel to the +12VDC output tap.

Incorrect. It actually costs LESS to make a single output +12V unit. When I convinced Ultra to go to a single +12V rail, they were actually ecstatic because their BOM cost went down a buck or two.

All a multiple rail PSU is is a single source split off into multiple "rails." There is ALWAYS a lot of devices connected in parallel to the +12V output whether the PSU is single or multi +12V rail.

In fact, a lot of the single rail PSU's are actually multi-rail PSU's with the OCP board removed and shunts connecting the separate +12V outputs on the PCB. This "lack of redesign" allows the OEM to go either way because a lot of system builders still follow the IED 69050.

Again, I'm neither for nor against single vs. multiple rail PSU's.... as long as the PSU is designed CORRECTLY. I AM against FUD and the general mis-information people like to spew out.

 

[johanh13]

If, in your example, they NEVER take 80A and split it into four 20A rails, what DO they do then?

Also, what is wrong with my analogy? I exaggerated the differences to make a point. For example, the first three houses may be using gas/oil for heating and cooking, while house four is all-electric (heat pump, electric water heater, electric stove). In my analogy, the houses are the peripherals. Oh well...

Originally posted by: jonnyGURU
...
When they split up rails, the NEVER take 80A, for example, and split it into four 20A rails. Because if they did, that would trap a lot of potential power and in that case what PCP&C's marketing would really hold water. But again, it's NOT what they do and you should know that before coming into a discussion armed with poor analogies. I'm not going to reiterate what I said. Just go back and read my posts and the examples I provide.

 

[jonnyGURU]

Originally posted by: johanh13
If, in your example, they NEVER take 80A and split it into four 20A rails, what DO they do then?

GOOD LORD MAN!!!! READ!!!!!

Originally posted by: jonnyGURU
When it comes to splitting up a +12V rail into multiple +12V rails, they don't just take the source and divide it up across multiple rails. All they're doing is limiting how much amperage is delivered to anyt particular connector or connectors.

If they DID evenly divided the source, there REALLY WOULD BE "trapped" power in a particular rail. That's what makes this "myth" such FUD. Because the sum of your rails is ALWAYS greater than the actual output of the source. In other words, there is an overlap of capability that compensates for any potential power that could be "trapped" by any one rail.

Take a look at ANY multiple +12V rail PSU and you'll see what I mean....

Rosewill Xtreme 850W: Four +12V rails @ 20A each. 4 x 20A = 80A. 80A * 12V = 960W. The source can only do 768W. 960W > 768W.

Gigabyte Odin GT 550W: Four +12V rails @ 18A each. 4 x 18A = 72A. 72A * 12V = 864W. Combined +12V capability is only 492W! 864W > 492W.

Look, fact is that any one +12V "rail" can put out ANY amount of power as long as it doesn't exceed the capability of the source. Typically, these rails are 20A or less because that's what's deemed "safe" by safety standards. But unless the PSU manufacturer puts too many connectors on one "rail" or doesn't have enough rails, there isn't a problem.

Originally posted by: johanh13
Also, what is wrong with my analogy? I exaggerated the differences to make a point. For example, the first three houses may be using gas/oil for heating and cooking, while house four is all-electric (heat pump, electric water heater, electric stove). In my analogy, the houses are the peripherals. Oh well...

Your analogy is wrong because it's based on your misunderstanding of how rail division works. Your analogy would only make sense if they DID take 80A and split it into four 20A rails. Then, by the other houses not using all of the power available does it leave untapped power at the source, but that's NOT how it works. If you have a PSU with multiple +12V rails that totalled 80A, it probably would NOT have four 20's. It would probably have six 20A rails.

For example.... This 1200W from Topower:

http://epowertec.com/power_ep-1000p10.html

Has 84A on the combined +12V capability. It has six 12V rails each capped at 20A each. What is 20 X 6? Anyone? Anyone? It's 120. 120A > 84A. Is it not?

 

[SerpentRoyal]

More rails = more parts. The engineer can specify lower grade electronics to save $ when migrating to a single rail design.

With highly automated production line, the cost of a well-designed single rail PSU will be slighly higher than the cost of a multiple-rail PSU.

 

[jonnyGURU]

Originally posted by: SerpentRoyal
More rails = more parts. The engineer can specify lower grade electronics to save $ when migrating to a single rail design.

They CAN, but what makes you think they DO? I guarantee you that in 99% of the cases where the same platform is used for both single and multiple +12V rails are implemented, the same grade electronics are used in both.

Originally posted by: SerpentRoyal
With highly automated production line, the cost of a well-designed single rail PSU will be slighly higher than the cost of a multiple-rail PSU.

Go on with yourself. I know for a fact that you're absolutely wrong.

 

[SerpentRoyal]

And how many years have you designed high performace PSUs for aerospace application? If the same platform is used, then the cost savings will be reduced part counts.

 

[jonnyGURU]

Originally posted by: SerpentRoyal
And how many years have you designed high performace PSUs for aerospace application? If the same platform is used, then the cost savings will be reduced part counts.

LOL! I haven't "designed high performance PSUs for aerospace applications" but I have and do work in the PC industry, specifically in the power supply category and know that you are wrong.

I'm not doubting your competence or intellegence, but your experience doesn't apply here. It's like saying because I work on top fuel dragsters that I know how to diagnose the OBD in a Escalade.

With computer power supplies they start with a base unit that typicaly only has one +12V output. Some may have two, but that's a moot point right now. The +12V source typically runs to a second PCB where this rail is split into multiple +12V rails with an OCP on each. This output then either goes back to the main PCB via wires or traces (depending on how the second PCB is implemented. If it's mounted elsewhere in the housing, it may just go out to the output wires. A vertically mounted PCB may run power through traces or pins at the base of the board, but this makes it difficult to retro the PSU back and forth between single and split designs)

To make this a single +12V design, IF the OEM has opted to even make this a viable option, they simply remove the second PCB and run all output wires out from one source. On some PCB's, this source may have separate landings, especially if the base unit was not meant to be retro'd back and forth between single and split +12V rails, so they'll use shunts to bridge the traces together.

Examples of PSU's that are designed to be retrofitted back and forth between single and multiple +12V rails are Seasonic and Channel Well. You can find variations of both with either single or multiple +12V rail outputs with virtually no change in component quality, value, etc. unless the OEM does so to reduce BOM cost or increase performance/reliability (i.e. Corsair specifying Japanese capacitors in their units.)

Examples of PSU's that were not designed to be single +12V rails but were modified to output on a single +12V rail are the Ultra X-Pro 750W (SevenTeam) and all of the X3 units (Andyson). These units were designed with a four +12V rail output design. Per Ultra's request, the split rail PCB was removed and shunts were installed on the seperate +12V output landings. NONE of the other components were changed in this design, nor did they need to be.

I'm sure there are LOTS of things you know a hell of a lot more than me. Perhaps even SMPS design. But unlike a lot of people here, I don't open my mouth unless I know what I'm talking about and on this subject, I know what I'm talking about.

Have a good Thanksgiving Serpent.

 

[johanh13]

SORRY GURU! I totally overlooked some key elements in your posts... but now I am enlightened! THANKS

Originally posted by: johanh13
If, in your example, they NEVER take 80A and split it into four 20A rails, what DO they do then?

Also, what is wrong with my analogy? I exaggerated the differences to make a point. For example, the first three houses may be using gas/oil for heating and cooking, while house four is all-electric (heat pump, electric water heater, electric stove). In my analogy, the houses are the peripherals. Oh well...

Originally posted by: jonnyGURU
...
When they split up rails, the NEVER take 80A, for example, and split it into four 20A rails. Because if they did, that would trap a lot of potential power and in that case what PCP&C's marketing would really hold water. But again, it's NOT what they do and you should know that before coming into a discussion armed with poor analogies. I'm not going to reiterate what I said. Just go back and read my posts and the examples I provide.

 

[YabbyU]

I have a question. If I were building a new rig with an EVGA 8800Ultra that has power requirements of:

Minimum of a 500 Watt power supply.
(Minimum recommended power supply with +12 Volt current rating of 34 Amp Amps.)
Two available 6-pin Molex hard drive power dongles

If I used the 1200 watt PS from Toppower

I would use (2) 12v rails @ 20amps each to power the card to be greater than 34 amps?

or

Since the the total output is 84amps would 1 rail power the card.



I'm not familiar with multiple rails....does each 12v rail only have (1) 6pin Molex?

 

[jonnyGURU]

Originally posted by: YabbyU
I have a question. If I were building a new rig with an EVGA 8800Ultra that has power requirements of:

Minimum of a 500 Watt power supply.
(Minimum recommended power supply with +12 Volt current rating of 34 Amp Amps.)
Two available 6-pin Molex hard drive power dongles

If I used the 1200 watt PS from Toppower

I would use (2) 12v rails @ 20amps each to power the card to be greater than 34 amps?

or

Since the the total output is 84amps would 1 rail power the card.



I'm not familiar with multiple rails....does each 12v rail only have (1) 6pin Molex?

The 34A requirement is taking into consideration the +12V demands for the entire computer. Not just the graphics card. So you would look at the 84A and say "ok... 84A > 34A, we're good."

The PCI-e connectors on the Topower are on +12V5 and +12V6, so there's actually more power available to these connectors than even the mechanical specification allows.

 

[YabbyU]

10-4......Glad I found this thread, I've always been cornfused about mutiple rails.

thanks jonny

 

[Howard]

Certain limitations can happen. I believe I came across one such issue when researching the Enhance ENP-5140GH (model # off the top of my head), 12V1 was limited to 16A and 12V2 was limited to 18A, with 32A total. IIRC, 12V2 was for CPU and 12V1 was for mobo/graphics.

Let's see if I care enough to check that now...

EDIT: I was right about everything about maybe CPU - the full 8-pin connector draws from both 1 and 2, when split, it's probably only 2 that it draws from.