Dual 12v rails inferior to single?

I'll throw this out there because it has been on my mind for a while now. I have been shopping around, and decided to go with the antec truepower 550 watt server edition. I love it so far. This model has the single 36 amp 12v rail. I know...pretty impressive.

There are a few reasons why I think a dual 12v design is inferior to a power supply with a strong single rail.

1) Most dual 12v psu's only offer 12-16 amps on each rail. Well, in my case, this isn't enough. A 630 pentium 4 processor is supposed to take 200 watts or power or around there. Well, if the psu has a 15 amp 12v rail, then 15 amps multiplied by 12 volts gives 180 watts....less than what a 630 requires.

2) Two 12v rails that are only 15 amps are MUCH cheaper than a solid 30-36 amp single 12v rail. It makes sense, doesn't it? Just put in two lower powered rails rather than making a power supply that can supply tons of amps on a single rail.

I really am not sure of this, but it makes sense that a single rail would be superior to a dual rail if the single rail was strong enough. While the idea of dual 12v rails and splitting the load between CPU and peripherals is a good one, I think that the CPU rail would need to be MUCH stronger for this to work. Since when does a hard drive use as much power as a CPU?

Maybe this is why the OCZ Powerstream and Modstream psu's use a single 12v rail? Any comments? If I am wrong, someone tell me please

-Collin-

You're right, having a single 12 line is probably better than two under powered lines.

The point of having two is for the other to compensate for fluctuation on the other. As long as the rails are strong enough, it should be better than a single rail, but in our case, I dont think that they're quite strong enough to suit our massive overclocking needs.

Maybe someday they'll be strong enough for us not to worry.....of course I dont think power consumption is going to go down in the near future, atleast not if Intel and their prescott have anything to say about it!

My power supply, the PC Power & Cooling 510 ATX-PFC Ultra Deluxe(short name) has only one strong 12v line. It is one helluva power supply, I dont need 2 lines to power my system.

Well the specs for ATX12v 2.01 and EPS12v 2.0 state that its for safety reasons that it have separate rails.

ATX12V 1.2.1. -- Increased +12 VDC output capability
System components that use 12V are continuing to increase in power. In cases where expected current requirements is greater than 18A a second 12 V rail should be made available. ATX12V power supplies should be designed to accommodate these increased +12 VDC currents.

ATX12V 1.2.4. -- Separate current limit for 12V2 on the 2x2 connector:
The 12V rail on the 2 x 2 power connector should be a separate current limited output to meet the requirements of UL and EN 60950.

EPS 12V 6.1.1 -- 12V Power Rail Configurations
There are two types of 12V rail configurations for systems: ‘Common plane’ and ‘split plane’ processor power delivery. The ‘common plane’ system has both processors powered from a single 12V rail (+12V1) from the power supply. The ‘split plane’ system has both processors powered from separate 12V rails (+12V1 and +12V2), one dedicated to each processor. The system, in both cases, has an additional 12V rail to power the rest of the baseboard +12V loads and dc/dc converters. +12V1, +12V2, and +12V3 should not be connected together on the baseboard to ensure that 240VA protection circuits in the power supply operate properly.

Click to expand...

It's not the amperage that's the problem, it's the fact that probagation delay and the inductors resisting changes in current that cause digital PSUs to have a finite response time. If the load suddenly changes, the voltage could change quite a bit before the circuitry can react.
The easiest way to explain is to use an analogy. Let's suppose that you're modulating the water valve for a garden hose so the sprinkler will water a certain area but not go outside the area. If the water pressure suddenly changes (like if someone just started to take a bath), you'll open or close the valve more to make up for the difference, but it's not instantanous.

BTW, I had a computer technician ask me about why hard drives were failing in a dual Britney server, particularly in the morning when the users are going online and putting a lot of load on the server. The drives will often log something about bad power before failing. There is a UPS and the PSU is relatively high quality (he told me it was a Fortron), so I asked him if he had an oscilloscope to check the 12v power. The problem was exactly what I suspected: the Britney CPUs transitioning from idle to full load and back put spikes on the 12v that could damage other circuits. The voltage will dip as low as 11v and spike as high as 13v when the load changes. I told him to put a noise filter on that line to the disk drives and so far, there have been no problems.
I'm not sure how much power a Britney CPU uses at full load (or even exactly what a Britney CPU is), but the best answer I've heard so far was, "About as much power as a small floodlight." (150w?)
But I'm now curious. Just how fast is a Britney CPU and what makes them great for servers? Oh, and what's the price?

Dell_Axim said:

It's not the amperage that's the problem, it's the fact that probagation delay and the inductors resisting changes in current that cause digital PSUs ...

Click to expand...

Switching AC-DC power supplies are analog devices.

Intel knows what they are doing. You can bet that dual 12V lines are superior, or the spec would not go in that direction.

That being said, there are limits to this stradegy. While I have heard of no single cpu system that runs poorly from a single, dedicated 18A 12V line, I have heard of dual cpu systems that have issues. If you read the EPS12V spec posted above, you are supposed to be using an EPS supply for these systems, and for high-power processors it should have 3 12V lines. One for each cpu, and one for the rest. So at this juncture, the only problems I have seen from dual lines are the simple result of misapplication of the power supply.

Not that a single 36A 12V line is a bad thing, but current single processor systems (even OC'ed) do run comfortably off a single 18A rail (even a 15A rail if you are talking about a Fortron). And if you have serious 12V video card draw (PCI Express), the isolation of the cpu and video load will allow tigher regulation a greater efficiency.

Switching AC-DC power supplies are analog devices.

Click to expand...

When I opened a computer PSU, there is a lot of digital logic to control the power circuitry. It might be better to say "mixed A/D" since there's still a lot of analog circuits. But the big transistors are (idealy) either fully on or fully off. That's digital to me. That's very unlike most small wall warts that contain no digital circuitry at all.

That's like saying a computer is an analog device because there are resistors and caps in it. I've seen you post 'answers' to power supply questions by asking if they have an analog or digital supply. Stop making these confusing statements. These are all analog devices. And whatever digital circutry exists in them has for decades, so whatever you choose to call them it's just confusing to go around asking people whether their supply is analog or digital. A supply is a supply, from that perspective.

larva said:

Intel knows what they are doing. You can bet that dual 12V lines are superior, or the spec would not go in that direction.

That being said, there are limits to this stradegy. While I have heard of no single cpu system that runs poorly from a single, dedicated 18A 12V line, I have heard of dual cpu systems that have issues. If you read the EPS12V spec posted above, you are supposed to be using an EPS supply for these systems, and for high-power processors it should have 3 12V lines. One for each cpu, and one for the rest. So at this juncture, the only problems I have seen from dual lines are the simple result of misapplication of the power supply.

Not that a single 36A 12V line is a bad thing, but current single processor systems (even OC'ed) do run comfortably off a single 18A rail (even a 15A rail if you are talking about a Fortron). And if you have serious 12V video card draw (PCI Express), the isolation of the cpu and video load will allow tigher regulation a greater efficiency.

Click to expand...

i agree with alot of this as i tend to with almost all your posts larva but ive found with heavily overclocked prescotts they can require a tremendous amount of amps on the 12v rail. my 600w powerstream couldnt carry my current overclock past 4.9ghz stably but both my 520w ps could run it fully stable up to 5.2 with more volts on it. i borrowed a friends 600w ps and had the same results.

i do agree with you on the fortron 15a rating as thats usually good for at least 18-20a. the fortrons are great psus as they tend to under rate thier psus whereas the majority of makers will over rate them or rate them @ startup temps rather than over a period of time.

That's like saying a computer is an analog device because there are resistors and caps in it. I've seen you post 'answers' to power supply questions by asking if they have an analog or digital supply. Stop making these confusing statements. These are all analog devices. And whatever digital circutry exists in them has for decades, so whatever you choose to call them it's just confusing to go around asking people whether their supply is analog or digital. A supply is a supply, from that perspective.

Click to expand...

I'll make clear the difference between a digital and an analog PSU. In a digital PSU, the power transistors (in theory) are either on or off, meaning only two states (discounting transitions and imperfections), and therefore could be considered digital. In an analog PSU, the power transistors are operated somewhere between fully on and fully off, and thus have a theoretically infinite number of possible states, hence analog.

BTW, many newer audio amps are called digital audio amps simply because of the usage of the output transistors!

Edit: Oh, I should have mentioned why digital is better for most computer applications. To simplify, it has been discovered that transistors are the most efficient at the two extremes: on and off. But how is it possible to turn a stream of pulses into a constant current, you ask? Good question. Inductors, capacitors, and transformers (as well as diodes and transistors) are connected to form a low pass filter. Low pass filters can also work as integrators. A digital PSU uses the mathematical property of integration to convert pulses into steady DC current. However, the low pass filter used as an integrator creates a side effect: the control circuits can only influence the output voltage through the filter. All sorts of problems are caused by this and this also makes digital PSU design hard. Oscillation can result from the time delay, load transients can cause dramatic changes in voltage before it can go back into regulation, and many other problems. Extremely smart engineers (many are PhDs) have found workarounds like bigger capacitors (which increases size and cost) and active voltage correction. One design used some high speed analog circuits to correct the output voltage of a digital PSU before the change in modulation gets through the low pass filter.

One way to sum it up is: "If you think doing the taxes is hard, digital PSU design is 1000x worse." (I have heard one digital PSU engineer say that, not sure if it's true or not, but I guess that there's a reason why a lot of people have to work to design a single digital PSU.)

Ok, I'm a total volt/amp idiot but I thought the Dual 12V rails "loaned and borrowed" volts from each other as needed, so each Rail is variable in capacity. Guess I should do some reading unless you gyuys can explain the concept in Grimm's Fairy Tale version.

Slake said:

Ok, I'm a total volt/amp idiot but I thought the Dual 12V rails "loaned and borrowed" volts from each other as needed, so each Rail is variable in capacity. Guess I should do some reading unless you gyuys can explain the concept in Grimm's Fairy Tale version.

Click to expand...

No, these lines are seperate. There is no "loaning".

Dell_Axim said:

I'll make clear the difference between a digital and an analog PSU. In a digital PSU, the power transistors (in theory) are either on or off, meaning only two states (discounting transitions and imperfections), and therefore could be considered digital. In an analog PSU, the power transistors are operated somewhere between fully on and fully off, and thus have a theoretically infinite number of possible states, hence analog.

Click to expand...

This is still not the issue with any of the posts here. What you are calling a 'digital' PSU is more correctly called a 'switching' PSU. This is what PC power supplies have been for decades, and almost entirely continue to be. They are what they have been, there is no need to quiz each person as to which they have. These are design details above user's heads and don't translate into helpful advice.

The way to tell which type is which is to look at transformers. Switching power supplies (what you are calling digital) have ferrite core transformers, so if you've been inside PC power supplies anytime recently, you know that's them. Linear power supplies (what you are calling analog) use 60Hz laminate steel transformers. These are rare in the PC power supply biz, and the only quality one I can point to is the Fortron AX500-A Blue Storm. And in fact, this unit has a ferrite transformer switching stage also, which is the dominant topology anyway.

Dell_Axim said:

BTW, many newer audio amps are called digital audio amps simply because of the usage of the output transistors!

Click to expand...

That would be false advertising. These amps use a "Class D" switching scheme, but are indeed analog devices. The term "digital amp" is a misleading marketing device and we have no need for an analog (no pun intended) of it here. The consequences of whatever design points you are throwing around needlessly were long ago mastered by big-time PC power supply companies are not a source of concern.

Im going to have to agree with Larva, and im sure OK Wolf would also agree. Dell_Axim your not exactly wrong, your just using the wrong term to describe the same thing.

hawtrawkr said:

i agree with alot of this as i tend to with almost all your posts larva but ive found with heavily overclocked prescotts they can require a tremendous amount of amps on the 12v rail. my 600w powerstream couldnt carry my current overclock past 4.9ghz stably but both my 520w ps could run it fully stable up to 5.2 with more volts on it. i borrowed a friends 600w ps and had the same results.

Click to expand...

My mistake, I guess it is possible to strain an 18A line with a single cpu. This sort of thing is why I still have a northwood, more power to you if you can afford the consequences of this kind of power consumption. The thing here is voltage always declines when you reach the current limits of the supply, so the difference in voltage is symptom rather than problem. The higher current (amps) is the operative agent here. Sounds like 5GHz (on a Prescott) is roughly the crossover point, where 18A rails are concerned.

hawtrawkr said:

i do agree with you on the fortron 15a rating as thats usually good for at least 18-20a. the fortrons are great psus as they tend to under rate thier psus whereas the majority of makers will over rate them or rate them @ startup temps rather than over a period of time.

Click to expand...

Yes, the 15A AX500-A drives loads about as stiff as any dual 12V unit, but I made sure to get the TPII480 vs the TPII430 to get a CWT 18A rating. This is probably the same thing if tested by actual output. And also remember that we are on the first generation of dual line units. Hopefully single line currents in excess of 18A will be available soon. How the rate of these increases compares to power consumption remains to be seen, and will determine just how radical of a setup you have to have before dual lines becomes a concern.

One point that may be of use at this juncture is that you don't have to spend a fortune to get a single 36A rail if you feel it is the better bet for your needs. And this is 36 Fortron amps, so you know its a beast:

http://www.newegg.com/Product/Product.asp?Item=N82E16817103478

I've avoided this unit in the past as I don't like the limitation of an 80mm fan, but the 36A alternatives have 80mm exhaust fans also. It has proven possible to drive stiff loads with this supply and a 29cfm Sanyo 80mm fan without consequence.

larva said:

also remember that we are on the first generation of dual line units. Hopefully single line currents in excess of 18A will be available soon.

Click to expand...

the 600w has its 12v lines rated at 20a and 18a respectively but im not sure how accurate those are. i like the idea of the dual rails though and im waiting for what you mentioned above to happen. the problem might be that by the time they start beefing up the rails the cpus might not be so power hungry. the dothan is a great example when gauging its power consumption to its processing power per clock and then putting it up next to the power hungry prescott with the weakest cpu cycles in the biz. from the looks of things intels moving more towards the dothans older p3 architecture than thier more recent prescott offering and amd hasnt had a problem with the heat/power consumption on anything near the scale of intels latest and ungreatest. so for most every single cpu solution in the world save a heavily oced prescott you were correct in your initial statement about the 18a rating i just wanted to throw out some of my experiences with the inefficient beasts.

thanks for the link to the psu though i wish you could have recommended it to me a month or so ago. if i was going to keep this 660 around id consider selling a 520w ps to buy one and play with but im already thinking of building a x2 system so ill have to sell my p4 hardware and then i probably wont need that monster of a 12v rail. (hell under fsp ratings its probably a good 40a continuous)

I think you've touched on the big picture here. While we as enthusiasts may put of with the electrical consequences of a 5GHz Prescott, and may even welcome the challenge, OEMs and JoePC will not. The P4's evolution was cut short for these reasons, not so much performance inadequacy. It does live on in the dual core processors, though, which are essentially two entire 2.8-3GHz Prescotts on a single package. Single 36A 12V rail is sounding entirely attractive for this sort of use.

This is still not the issue with any of the posts here. What you are calling a 'digital' PSU is more correctly called a 'switching' PSU. This is what PC power supplies have been for decades, and almost entirely continue to be. They are what they have been, there is no need to quiz each person as to which they have. These are design details above user's heads and don't translate into helpful advice.

The way to tell which type is which is to look at transformers. Switching power supplies (what you are calling digital) have ferrite core transformers, so if you've been inside PC power supplies anytime recently, you know that's them. Linear power supplies (what you are calling analog) use 60Hz laminate steel transformers. These are rare in the PC power supply biz, and the only quality one I can point to is the Fortron AX500-A Blue Storm. And in fact, this unit has a ferrite transformer switching stage also, which is the dominant topology anyway.

Click to expand...

Well, engineers in my area call them "digital" and "analog". But I guess that engineers here call them "switching" and "linear".
But certainly, computer PSUs have evolved to meet the needs of new hardware. In the past, 30A or more on the 12v line or multiple 12v lines are only necessary for very large RAID arrays. Now, faster logic is depending more and more on 12v to obtain power. A single 12v rail is a problem because noise from high power logic (like the Britney CPUs in the earlier example) can find its way into other sensitive circuitry (like the disk drives in the earlier example) and cause problems. If the PSU has two rails, one for logic and the other for drives, that problem wouldn't have existed. A noise filter also works and is cheaper and simpler to implement than a whole new 12v rail. So at the least, they could put a noise filter on the line going to the drives.
BTW, the reason why the logic itself isn't disturbed by the power noise is because it actually runs off a regulator. In contrast, drives usually use the 12v without regulators and are more sensitive.

Also note that switching PSUs are starting to use microcontrollers (single chip computers) that can optimise for different conditions. For instance, it could change the operating frequency for different load levels to optimise efficiency, spread out the clock frequencies to reduce EMI, or even disable and enable converter units in a multi converter PSU. Some can even warn about fault conditions, although that's usually only found in certain server PSUs.
Still, I would like to see a PSU with a small LCD displaying the number of watts being drawn, the temperature of the circuits, and the input voltage. Maybe even display how many amps are being drawn from each rail. Even a red LED on the PSU itself to indicate a fault would be good, but that's about as useful as a "check engine" light in a car, so I'd prefer a LCD display.

I have consulted with two senior power supply design engineers on this one, switching and linear are indeed the correct terms.

These details may (or may not) be very interesting, but they don't serve the purposes of the threads into which you are constantly interjecting them. These are all switching power supplies, and you are clouding the issue with design details above users' heads. Again, I simply ask you to stop it. If you have questions or suggestion germane to the resolution of the problems being asked about, great. Otherwise, it's just another form of thread-crapping.

Well, I appreciate all the input. I guess it goes back to...."did I buy the right power supply?"

It kind of makes me worry. Larva brought up a good point that if you have a lot of strain from video cards on the 12v rail, you would want to keep the rails separate. Well, I have two 6600GT's overclocked, so I think that constitutes a lot of strain

So far so good with this antec though. The fans don't slow down when I click on things like with my enermax!

-Collin-