PSU Myths (a must see post!)

Im very restrictive when it comes to company's views on things because its usually angled. But in this case I do agree.
Why? Because I have been telling people this for years.
Most common enthusiast seem to believe from pure assumptions that Motherboard manufacturers and Graphic card manufacturers and PSU manufacturers don't have common standards. Well ATX standard is on and there are different versions of it to. There are other standards to like EPS12V and so one.

I my self used a Tyan motherboard K8WE until recently and two daulcore opteron processors so this is a 2 way system with 4 cores total also I used two GTX260 cards in SLI and some other components like SSD drives so basically its had 2x the components of a normal PC.

Just like your URL shows is that etch rail is made for a specific purpose.

Take my setup for instants I used then. The 8 pin EPS12V connector supplying CPU power have 2 rails. 12V1 and 12V2. Etch of them feeds One CPU that was how Tyan designed the motherboard.
Then The 24 pin ATX connector feeds the motherboard and its expansion cards like my 2 graphic cards with 12V3 rail.
Then I have 2x 6 pin PCI-E connectors that is feed by 12V4 rail and then I also have 1x 6 pin pci-e connector that uses the 12V3 rail and a forth one using the 12V2 rail.
Etch Graphic card had one 12V4 rail connected to it and one card had the 12V3 and the other the 12V2 to even the load as much as possible.

So in the end CPU one used 12V1 and CPU two used 12V2 and both graphic cards used 12V2,12V3 and 12V4 and the other components like hard drives used 12V4.

I use the same 4 rail PSU to power my i7 and two GTX260 cards.

But one must first understand how rails are meant to be implemented to properly connect them and do load balancing. Sure if you simply don't know and are not aware then a single rail PSU is simple to connect. One rail so you cant go wrong.

I use both multiple rail PSU at home from 2 up to 4 rails and single rail PSU and all works in practices just fine. Whats best well I would say it depends on what its gonna be used for. I have seen people use 2 rail PSU's rated 600W in fileservers where one rail was only connected to the PCI-E connectors and the other connected to the 4/8 pin connector feeding the CPU and also to the Molex and Sata power connectors so in the end the PSU could only supply some 200W. Powering up say 15 drives is no problem with 600W it uses about 350-400W during boot I know from testing but try doing it with one rail supplying say 200-250W. Just wont work because the overload protection will kick in before the power level starts to drop. And the user in this case hade added som more drives and dident understand why his PSU tripped.
So yes bad planing and it can fail with multiple rail PSU's.

Right PSU for the right task.

One claimed advantage of multiple rail PSU's is lower noise and ripple and better efficiency at lower power draw then a single big rail. I have seen test on this but I haven't seen any lately so I wounder my self how the multiply rail PSU stands up to the single rail PSU to day considering how fast the development is going. Supplying power with multiple rail PSU is no problem if one knows what one is doing but how are the other aspects of a multiple rail PSU in comparison to a single rail one?

If one is gonna market multiple rail PSU to enthusiast one can not only say that it works just as well. Because if it does why select one if a single rail is fool prof to connect wile a multiple rail version can take a lite bit of know how and planing to select for the system. If there are other advantages then those needs to be especially highlighted.

Other wise I think people will take what ever is the simplest and is most fool prof.

Also I think its very important to mark etch connector, Molex, SATA, PCI-E especially with there corresponding rails thy use. I have encountered PSU's that have 5-6 rails and don't have any markings or even a manual stating more then the number of rails. So the manual should have a detailed schematic and explanation of that to. other wise its not easy balancing the rails properly.

And I personally don't see the necessity for 40A on etch rail but sure. The once I use are rate 20A etch but total of 72A (4 of them) @ 50C. Still that works fine for most high end systems. Actually it should do just fine even with a GTX480 if connected correctly considering that one can use up to 3 rails for a card like that and the limiting factor is not the rails but the PCI-E connectors if your going to stay with in PCI-E 2.0 specializations.
On advantage is the added safety of the 20A limit of the ATX standard. But thats my personal preference. I tend to get a PSU that stays with in the the standards of ATX or EPS12V or what ever I need.

So a psu using a split method should be better than one not using this method...?
Like, "hey we split the cpu core to 100, it sound cooler but have no effect.. or it lower a bit, but hey! you got 100 core´s now"
Sry if i didn´t have the time to read about some adds, and due to this i´m misstyping some info. But my point of view is it can NEWER be better to split upp the main power to smaller one´s and call it a better system.

Well your assuming that a single rail is equal to say 3 or 4 rails have the same overall property's like efficiency, ripple, noise. Well thats probably not the case. I have read articles about that. As fare as i know in the past single rail high power psu's have been considered less efficient at say lower loads or have poorer property's concerning noise and ripple from the rail.
Then there is the question if the multiple rail psu is a real multiple rail PSU not just a big rail divided by multiple OCP's. But I cant say that its easy finding a good test of a single rail PSU state of the art VS a real multiple rail PSU preferably with 4 rails at 20A etch and compare there properties I mentioned. Until then people will still argue whats best.

So this is the problem with marketing multiple rail PSU's to enthusiast.
Single rail PSU manufacturers just have to say. Dont worry one rail fore everything! So no worry's. Well sounds simple enough? But did any on ask about any other property's. For the same reason budget PSU's sell on Watts so do single rail PSU sell on simplicity. Its easy to understand there point.
Multiple rail PSU manufacturers only tries to prove that multiple rail PSU are as good not better. I know thy most likely are. I use both not problems but "Why" should one select a multiple rail PSU? Thats what Antec need to state when thy try to sell there products.

Understanding the advantage of single rail is easier then multiple rails so if you cant relay the message you doomed to fail.

Thy need to prove advantages of multiple rails compared to single rail. If ripple is less then prove it. If efficiency is better then prov that. Then people will purchase them if those advantages are preferable.

And the 1 vs 100 core comparison fails for the same reason. You assume 1 advance and fast core is as good as 100 simpler cores. Thats also not the case.
Ask a computer engineer or scientist. Its been a know fact for a long time that computer power per watt increases in a system the more parallelism thy have. So yes 100 cores should be able to do the same work with less power or do more work with the same amount of power.

But the biggest problem is how do you make most software parallel. Works good for supercomputers very often but more traditional software for a PC is not so easy.

No argue about that, if you have a variable system power usage, ofc the multiple psu <should> have a great advantage here.

So what's all this rubbish about multiple 12 volt rails?
A bit old article but interesting.

Some things might have changed in the last 4 years but according to that there are almost no REAL multiple rail PSU's.

Well there are Single Rail PSU and then there are multiple rail PSU that either are one rail divided up in many rails etch with a OCP circuit so in reality its still one rail. Also like he states in the article true multiple rail PSU's the rails can't be connected together usualy becuse of risk of damading the PSU but if its a single rail PSU that have the rail divided up with individual OCP protection one can actually connect the rails together becuse thay share the voltage regulator unlike real multiplie rail PSU's. I have done that my self on "multiple" rail PSU's of low quality but on the other hand I expected them to be fake rail PSU's to so no surprise there.

Then there are individually voltage regulated multiple rail PSU's. As fare as I understand thy are considered Multiple rail PSU's but thy still share the main transformer. So unlike the first case where thy share both the main transformer and voltage regulator and the rails where only created by separate OCP circuits this also have separate voltage regulators thats why there rails should never be combined sens voltage can differ. There are those enthusiast PSU's out there that allows the rails to be combined after a switch is flipped that most likely allows the rails to share one voltage regulator and effectively becoming on rail again.

So what I understand those that have independent voltage regulation per rail even if thy share the transformer are multiple rail PSU's but to be the ultimate in TRUE multiple rail PSU thy should also have independent transformers and apparently most seems to lack that.

So one transformer, one voltage regulator and a OCP per rail. So a TRUE 4x12V rail PSU should have 4 transformers, 4 voltage regulators and 4 OCP's one of this components per rail. But thats makes them expensive and thats obvious.

But even the most expensive multiple rail PSU's seems to be one main transformer and then maybe independent voltage regulators on the more expensive models.

Independent voltage regulators per rail should reduce interference between rails but not as much as if thy also had there own transformers as fare as I can understand from doing some digging.
So a true multiple rail PSU shudder have lower noise and ripple according to the article but considering that most PSU's are not real multiple rail PSU's the "advantage" is not really there.

So the question is if one want something better the a single rail PSU is there actually any true multiple rail PSU's on the market?

Very nice to get some feedback.

But I'm still a bit confused about what your saying and what the PSU's actually are designed to do.

I know the advantage of current limiting is often highlighted to for safety. I my self see that as an advantage. But I checked the PSU I have its rated at 18A per rail and I'm pretty sure its OCP is rated 1.5A above that so 19.5A should trip the PSU on any of the +12V tails. The ATX standard should limit the rails to 20A if I'm not mistaken for safety reasons. You also say that you increased the output to 40A per rail beyond the ATX specification to ensure enough power gets to graphic cards etc.
Even with 20A limit per rail the PCi-E slot and PCI-E 6 and 8 Pin connectors depending on version and model are limited to 75W or 150W according to PCI-E specifications so 20A +12V rail can still deliver 240W, more then enough for one or more PCi-E connectors so even with 3x 20A rails on can supply the most demanding GTX480 or HD5970. Seems almost unnecessary to me to increase the OCP protection to 40A. Better to have more rails and less PCI-E connectors per rail and power wont be a problem for 3 or 4 way SLI or CF systems. 4 rails capable of max 18-20A should be enough for a 2 way SLI/CF high end system with even the most demanding CPU.

I know what happens when you run 60A over a molex for example and its cables during short circuit. It have happened and well The PSU will not trip and the cable wont burn but thy will turn in to ash including the copper or aluminum wiring and connectors will melt within a few seconds.
But then it also depends on what diameter the cable is and how many there are of them.
if the resistant is low enough it will trip.

But 40A during short circuit is a LOT and I have a hard time seeing the benefit of a multiple rail PSU thats not with in ATX specifications of 20A per rail if safety is you marketing point. The wires and connections should make it if the PSU trips at 21-22A immediately but the resistant in the cables and connectors when you have 40A ore more available can keep the current under the OCP limit because of the resistant and resistant usually increases with temperature so the the outcome can still be the same. The higher the OCP limit the more likely the outcome is permanent damage to the PSU, wiring and components but well 40A is still a lot I think.

I see a lot of multiple rail PSU's allowing what I would call dynamic load balancing amongst rails but how is that beneficial to safety when its more likely to cause damage then a multiple rail PSU allowing maximum of 20A per rail before tripping?
To me it seems to be more of a insurance to users that imagine that a PSU that have say 4x18A rails cant power 1 or even 2 CPU's and even 2 graphics cards at once. I know thy can I have don it my self and put the system under heavy synthetic load. But I have also have one unique 12V rail per Graphic card and then thy share 2 rails in common and the CPU have a dedicated rail and also shares on whit one of the graphic cards. So load balancing is still possible. Not exactly to the same extent but still enough for it to not be a problem for a well designed PSU and correctly connected to the components from my experience.

Seems like your doing one thing and saying something else wit the 40A OCP limit and Safety marketing Point?
As I sad earlier I my self use two multiple rail PSU's at least maybe more, one 2 rail and one 4 rail but all are should we call it old school versions that do not allow any load balancing and currents over 20A per rail. Most of them are rated about 18A per rail and OCP protection 1-2A above that.

Then I also have single rail PSU's.

It seems like load balancing with OCP's rated above 20A is some compromises from a selling point and the question is, is it really as safe or safe enough then a classic multiply rail PSU's staying with in ATX specifications allowing maximum safety or is its safety more like a single rail PSU's allowing 60-80A over one rail before OCP kicks in? I theory it should offer something in between but in reality if things go south will it act more like a 20A limited multiple rail PSU and shut down or more like a single rail PSU and keep the current flowing and cables melting because the OCP didn't trip?

I my self is all for Multiple rail PSU's. Thy should if constructed properly tough expensive provided some benefits but I don't feel so sure about multiple rail PSU's with OCP protection fare above recommended limits and dynamic load balancing features.
I my self got in to the Multiple rail PSU dilemma when building my second Opteron system a few years ago and multiple rail PSU's fallowing specifications exactly was recommended and most users seems to prefer multiple rail PSU's with 20A limit per rail that liked to build this kind of high performance systems. I have taken that back to more traditional desktop system sens thy demand lots of power or even more then some multiprocessor systems considering how much GPU's uses to day.

Why not just select a single rail one then if one is not afraid of things going wrong?
If on really wants safety why not stick to a PSU with 20A limit per rail?
To me 40A OCP seems like one wants both but cant decide. I'm not much of a believer in eating the cookie and keeping it at the same time.

Would be interesting to know why you decided to go for this type of design and specifications. I'm not a PSU expert just a consumer that likes to know what I'm getting and why. I'm a bit skeptic to this multiple rail PSU's allowing OCP fare over 20A and dynamic load balancing between rails.

Also is Antecs new PSU's of this type one rail PSU's divided by 4 OCP's creating synthetic rails and sharing main transformer and voltage regulator or do thy have individual voltage regulators and transformers to create real separate rails to minimize one rails influences over another like ripple?

I know this was a long post but It's hard to find good facts about PSU's and there workings. I'm probably wrong to about some things but well cant be right until you know your wrong.

EDIT:
If I might make a suggestion for future PSU's. I know some that have multiple rails allows them to be combined in to on rail. Why not do something similar but add the ability to set the PSU's +12V rails to 20A OCP or 40A OCP with a switch so if the user whants even more added safety he can stay with in ATX specifications or triad some of the safety for the benefit of more current to a given rail.
That would make the PSU more attractive from my standpoint and probably for other users that prefer multiple rail PSU's.

No you misunderstood that. My point there was not that the connector physically is limiting the current, thy are limited or rated to 75W och 150W according to PCI-E specifications. Thats there maximum rated load. I know there is no physical limitation until the thing burns more or less. My point was that 20A OCP limit per rail according to ATX specifications is 240W total power per rail and thats enough for 150W + 75W connector or 3x75W connectors so why the 40A limit then? 20A seems enough per rail?
I have seen multiple rail PSU's with 2 or more rails dedicated to PCI-E connectors.
Sure your can run more then there rated wattage over a 6 or 8 pin PCI-E connector but that was not my point that was just a misunderstanding. My point was that 40A or total of 480W per rail seems unsafe and unnecessary because I doubt a 6 Pin PCI-E connector will take that and 20A is enough for 2 or 3 connectors.

600W for one card is extreme for any PSU but if you ignore the maximum ratings of the PCI-E connectors you can supply the card with 3 separate rails. Even with in ATX specification with OCP at 20A that could come up to some 630W if you don't run more then 150W over motherboard considering it might not take that and then 480W over 6+8 PIN PCI-E connectors typical for high end cards. If you ignore the motherboards 150W rated limit for PCI-E 2.0 Slot then even more is available. But then the question is where and how do the graphic card balances its current draw from different sources. I have seen test so I know thy rarely draw equally. from all 3 sources.

My English is not the best I'm afraid so miss understandings happen.