Enermax Liberty

F

frofan

n00b
Joined
Oct 30, 2005
Messages
8
I had a neohe 500, and had to RMA it due to a compatability issue with Asus Mobo. I'm looking for a new PSU and would like to get something with 32+ amps on the 12v rail and would like it to be module. I'll be powering the system below. I have been looking at the Liberty's, but wasn't sure if I should be concerned with it not regulating the voltage on the 12v rails? What modular PSU do you suggest if the liberty is not a good choice?

AMD x2 4400
asus a8n-sli premium
evga 7800 gt co
2 raptors
2 case fans
 
You weren't sure about the Liberty not regulating the 12V rails?!?!

Huh?!!??!

The Liberty as a modular power supply is actually more stable, even at high loads, than MOST NON-MODULAR POWER SUPPLIES.
 
What is the major difference between the liberty 500 and 620? They both have 22amp rails, but the 620 is $50 more. Is it worth the $50 for the 620?
 
The difference is the COMBINED WATTAGE.

NEVER look at EACH RAIL when looking at multiple rail PSU's. Always look at the combined.

The 500W does 32A combined (which is more than enough for 99% of the folks out there) where the 620W does 36A.

All the "each rail" wattage tells you is what the max for each rail is, but that doesn't mean you can max out each rail. You're still limited by the combined capabilty.
 
frofan said:
I was going by the second post in this thread by retrofitter.

http://www.hardforum.com/showthread.php?t=971391&highlight=liberty



Is the liberty a good choice, or is there better?
Click to expand...

Ok.. You weren't sure about the Libert not regulating as much power on the 12V rails. Go to watch your verbiage when your thoughts go to your fingertips. Sometimes the message conveyed is completely wrong. ;)

I'm bumping this thread because I learned something that I think is interesting out about SLI certified dual rail (and tri, and quad..) PSU's, and I don't have enough information to conclude an educated opinion, but the way it sounds... single rail PSU's are all but useless and suggesting dual rail PSU's to help "isolate loads" is completely bogus. Hopefully someone here can clear the mud... otherwise all of this information is just my own ignorant speculation.

We all know that some PSU's have either their rails on truly separate rails while others (most) have their rails regulated coming off of the same plane. But I was recently talking to a guy that told me that apparently nVidia puts a 15A load on each rail via a PCI-e connector for their SLI test. Most PCI-e connectors are on the same rail, not two separate rails. In fact.. I can't think of a power supply that puts each PCI-e on a separate rail. It's almost always on 12V1.

Think of this analogy: Two AA batteries in parallel still provide 1.5V, but can now provide 4A of current instead of just 2A. Not common practice for AA batteries as they're usually run in series to provide more voltage but the same amperage, but we all use AA batteries every day so I thought it was an easier way to grasp this upcoming concept... The variables used in the wattage equatio are different, but the result is still the same: 1.5V X 4A = 6W, 3V X 2A = 6W.

Still confused? Look at this: http://www.allaboutcircuits.com/vol_6/chpt_3/3.html

Ok.. So what they're doing is making these power supplies to where the MOSFET's are running together in parallel. I don't fully understand how the rails can still be considered "separate." But if you look at a Silverstone manual, they clearly state what connectors are on what rail.. so perhaps there's some sort of circuitry there that limits the flow of current from a particular rail unless that rail is overloaded. Some sort of "switching diode" if there was such an animal. But also this could explain why some companies don't disclose what rails power what. Hard to say because despite all of these guys I know that work for these power supply companies that are engineers are not the engineers involved in manufacturing these power supplies over in China, and the one guy I do know that is involved in the manufacturing process doesn't speak English well enough to explain it to me.

Intel specs does say there should be a "load limiter" on each 12V rail that prevents more than a 20A load on a rail, but obviously this isn't getting done. Funny thing about ATX specs... nobody actually TESTS a PSU for ATX compliance. :eek: It's just a paper thing.

So here's my point.. Take an Antec TPII 550, for example. SLI approved. Has two PCI-e connectors, both on 12V1. With a 15A load on each connector, the PSU passes the test... yet the limit on 12V1 is only 19A. Where'd the other 11A come from? It would have to come from 12V2.

Hmmm......

So multiple rails will make a power supply more reliable as it's easier and cheaper to use multiple 12V MOSFET's that are of a lower spec than one big one. But I think we can throw the whole "better stability due to isolated load distribution" theory right out the window.
 
So multiple rails will make a power supply more reliable as it's easier and cheaper to use multiple 12V MOSFET's that are of a lower spec than one big one. But I think we can throw the whole "better stability due to isolated load distribution" theory right out the window.
Click to expand...
Not so fast, not so fast, let's not throw the baby out with the bath water!

Tail of 3 PSU's:

(1) Sparkle EPS 550W has one +12V rail rated at +12V@36A.

(2) Enermax (#701) 600W has +12V1@18A & +12V2@18A, max combined +12V@36A

(3) Antec Neo HE 550W has three +12V rails with a max combined of +12V@36A

Now the first two are under the ATX standards although both are EPS compliant with a 8-pin EPS connector,
#3 is under the SSI/EPS standards.

ATX: http://forums.extremeoverclocking.com/attachment.php?attachmentid=60568 (PDF)

SSI/EPS: http://www.ssiforum.org/Power Supplies/EPS12V_Spec 2_1.pdf

To be continued.... :D
 
davidhammock200 said:
To be continued.... :D
Click to expand...

Well hurry up, because nothing you've said so far has anything to do with put two 15A loads on one 19A rail... without blowing it up. ;)
 
In both of those specs, the only thing that's truly "isolated" is the CPU rail. But that's my whole point... if 12V1 can also get power from 12V2 in an overload situation, how is it that 12V2 is atually isolated from 12V1?
 
jonnyGURU said:
In both of those specs, the only thing that's truly "isolated" is the CPU rail. But that's my whole point... if 12V1 can also get power from 12V2 in an overload situation, how is it that 12V2 is atually isolated from 12V1?
Click to expand...
And all 3 of my examples are "common plain". ;)
 
So? I don't get your point, Dave. Are you getting mine or do you think I'm talking about something else?

Common planes are common prior to the voltage regulation. If they're simply running the MOSFET's in parallel, common plane or not, then the rails are still technically "one rail" with the duty of voltage regulation simply being divided up between two or more transistors.

The combined wattage limit is usually not dictated by the result of the combined capability of those transistors, but by the power that's being fed to that plain prior to regulation.

So even if a two rail power supply is capable of a combined wattage of 384W and each 12V rail is capable of 240W, certainly you can put up to 384W on the 12V rails and not 480W. We know this and are constantly preaching this to our minions.

My point is.. HOW IS IT that I can put a 360W load on ONE RAIL, not both, but ONE RAIL that exceed 240W (using my above example) when that rail is supposedly only rated for 240W. Theoretically, and per ATX spec, that rail should trip overload protection. Instead, it's able to rob Peter (12V2) to pay Paul (PCI-e.)
 
My point is.. HOW IS IT that I can put a 360W load on ONE RAIL, not both, but ONE RAIL that exceed 240W (using my above example) when that rail is supposedly only rated for 240W. Theoretically, and per ATX spec, that rail should trip overload protection. Instead, it's able to rob Peter (12V2) to pay Paul (PCI-e.)
Click to expand...
What dual +12V rail PSU's (except Skyhawk) have allowed you to pull 36A from just one of the rails?

Or 30A or 26A or even 24A?
 
jonnyGURU said:
Well hurry up, because nothing you've said so far has anything to do with put two 15A loads on one 19A rail... without blowing it up. ;)
Click to expand...

It would probably depend on the design of the supply to be able to do something like this. Cheap crap like the Skyhawks don't (or at least didn't) actually have two isolated 12v rails, and this could certainly be possible for something like that. Whether the Skyhawks can managa 30A of sustained 12v current is another story of course.

However something like the Fortron FSP400-60THN, which likely has a current limiter for each 12v going into the common 12v winding probably will shut down on you if you try overloading each 12v by such a large amount. There likely is some leeway allowed there due to tolerances of the components, but when this model was tested by Oleg at Xbit, it did shut down at 29A @ 12v combined. He didn't go into specifics on each 12v though :(

I'm not an engineer, just a service tech, so I'm not sure I can be much more help; but Oleg looks to have a pretty good understanding of the engineering side of these things. It is entirely likely though that to keep costs down and use as few parts as possible, many OEM's are running both 12's off the same regulator.
 
davidhammock200 said:
What dual +12V rail PSU's (except Skyhawk) have allowed you to pull 36A from just one of the rails?

Or 30A or 26A or even 24A?
Click to expand...

So you didn't read what I typed. :p

I used the Antec TPII 550 as an example. The two PCI-e connectors are on 12V1. nVidia load tests for SLI certification by putting a 15A load on each PCI-e connector.

That's 15 + 15 = 30A on ONE RAIL. Theoretically impossible since the 12V rail is rated at 19A.

See my point now?
 
Oklahoma Wolf said:
It would probably depend on the design of the supply to be able to do something like this. Cheap crap like the Skyhawks don't (or at least didn't) actually have two isolated 12v rails, and this could certainly be possible for something like that. Whether the Skyhawks can managa 30A of sustained 12v current is another story of course.
Click to expand...

But it's NOT cheap power supplies.. it's the power supplies on the nVidia SLI list!

Come on guys... quit speed reading me here. ;) Y'all should know I'm talkative and don't believe in Cliff's.

The Antec TPII 550. Supposedly 19A on each rail... but can handle a single 30A load on the 12V1.

I've got the Liberty here still hooked up to the tester. Supposedly 22A on each rail. But I can put 30A on one rail easy as long as there's no load on the other rail. No problem.

What's funny is nVidia's "requirement" completely contradicts Intel's ATX spec.....

ONLY THE CPU CAN BE ON 12V2! Yet nVidia wants to put a 15A load on each PCI-e connector, assuming each PCI-e is on a separate rail. But instead of moving a PCI-e to a different rail from the other, they're just doing without load limiters and eliminating the isolation of the rails! BOGUS! :eek:
 
nVidia load tests for SLI certification by putting a 15A load on each PCI-e connector.
Click to expand...
Are you sure? Is this documented?

If this is true then no "in spec" ATX12V2.xx dual rail PSU could be SLI Certified. :eek:
 
davidhammock200 said:
Are you sure? Is this documented?

If this is true then no "in spec" ATX12V2.xx dual rail PSU could be SLI Certified. :eek:
Click to expand...

Exactly!

Is it documented? I'll see if I can find something. I've got a copy of the SLI certification app somewhere. It might be in there.

I got the information from someone that was trying to get a PSU to pass that had 12V rails on separate planes and a 25A current limit on each rail. It failed and when asked how a PSU with 19A on each rail could fail such a simple test, he was told it was because they put a 15A load on each PCI-e connector regardless of how many rails the PSU has, what connector gets power from what rail and what other loads may potentially be on the power supply.

They're contemplating one of two solutions: Remove the load limit, assuming it can handle a 30A load on 12V1, or moving a PCI-e connector to 12V2. Either solution would take the PSU out of ATX specification.
 
davidhammock200 said:
The ATX spec gives all of +12V2 to the CPU via the p4! :rolleyes:
Click to expand...

Right. Which is overkill.... You look at a power supply with 12, 14, 18.. 22A!!!! on the 12V2 and it's nuts. What's the MOST you're going to draw? A Pentium D draws maybe 10A if memory serves me correct, right?

So it would make sense to put something else on that rail, but that wouldn't be per Intel spec. You could put something other than drive Molexes on the rail and you'd still have clean power... but that's not Intel spec. But I still like the idea better than getting rid of the current limiter.
 
Some OC'ed Prescotts use over 200W's I believe. :eek:

There have been complaints that the Frotron Blue Storm 500W's +12V@15A isn't enough for some Intel OC's!
 
jonnyGURU said:
But it's NOT cheap power supplies.. it's the power supplies on the nVidia SLI list!
Click to expand...

Doesn't necessarily mean the design isn't cost or space oriented though... regulating each 12v seperately means more parts and higher cost. For instance, I know Channel Well pretty well by now - I very much doubt both 12v rails in the True 2's have anything more than a single 12v regulator for the two of them. CWT designs are good, but they're usually a bit on the frugal side. Whether they have the required by spec current limit on each 12v is a matter of speculation, but I will say this - the UL certification page lists numbers for at least the 550 higher than the specs on the label. 21A+21A IIRC. If that's where the True 2 550 shuts down, CWT's overcurrent protection seems a bit liberal.

Edit - and how scary is it I now apparently have CWT's UL file number memorized - e161451. I needs a life I does ;)
 
davidhammock200 said:
Some OC'ed Prescotts use over 200W's I believe. :eek:

There have been complaints that the Frotron Blue Storm 500W's +12V@15A isn't enough for some Intel OC's!
Click to expand...

Ok... So 200/12=16.666. Doesn't leave much for a PCI-e card... but would you want to put anything other than a video card on that rail?

See that's messed up. That's my whole thing. The Intel spec actually ensures you have enough juice for your CPU, and typically the rest of your rig (14A for a couple PCI-e cards, another 6 or 8 for drives, etc.) but the SLI certification completely negs that.
 
Oklahoma Wolf said:
Doesn't necessarily mean the design isn't cost or space oriented though... regulating each 12v seperately means more parts and higher cost. For instance, I know Channel Well pretty well by now - I very much doubt both 12v rails in the True 2's have anything more than a single 12v regulator for the two of them. CWT designs are good, but they're usually a bit on the frugal side. Whether they have the required by spec current limit on each 12v is a matter of speculation, but I will say this - the UL certification page lists numbers for at least the 550 higher than the specs on the label. 21A+21A IIRC.
Click to expand...

Well sure, again.. I don't have an issue with them running the transistors in parallel and having all of the juice come from the same source. Multiple smaller MOSFET's is still better than one big one (unless it's a really high end one like the kind PCP&C use.) But it's not adhering to Intel spec and how is it a PSU manufacturer can say "the 12V1 goes here, and the 12V2 goes there" when in reality they're really getting their power all from the same combined rails?
 
jonnyGURU said:
But it's not adhering to Intel spec and how is it a PSU manufacturer can say "the 12V1 goes here, and the 12V2 goes there" when in reality they're really getting their power all from the same combined rails?
Click to expand...

Not sure I can help you with that one... a little confused about that myself.

jonnyGURU said:
the SLI certification completely negs that
Click to expand...

Which is why I never recommend anything based on either Intel's specs or Nvidia's certifications for high draw systems. Better to recommend a big single 12v supply in the first place than guess at which ATX 2.0 unit can borrow power from where for which rails IMO.
 
Well if UL tested them & got shut down at 21A per rail & the max combined is 36A,
then there must be OC protection for each rail.
 
davidhammock200 said:
Well if UL tested them & got shut down at 21A per rail & the max combined is 36A, then there must be OC protection for each rail.
Click to expand...

There may be, but the difference between the 36A combined rating and the 42A draw they got from it kinda leads me to believe they're perhaps not quite as quick to shut down as they could be, thus allowing for one of the dual 12's to possibly exceed its ratings or even UL's tested results by a fair bit when one isn't being used.

Edit - in going back over this thread, I'm starting to think that some of these units are actually skipping the seperate OC protection on each 12v, rather preferring to just put in a general OC circuit for both. Would save cost and space in the design, and would explain why the Liberty and True 2's can exceed their ratings by so much. Would be interesting to see if the upcoming Fortron quad 12v units and the Enhance/Akasa/Silverstone quads are this way too, since they are supposed to conform to EPS SSI specs. But then, how is Nvidia getting 30A @ 12v3 from the Silverstone? Or, is Nvidia really testing them this way? I'm thinking there's a piece of the puzzle missing.
 
jonnyGURU said:
I got the information from someone that was trying to get a PSU to pass that had 12V rails on separate planes and a 25A current limit on each rail. It failed and when asked how a PSU with 19A on each rail could fail such a simple test, he was told it was because they put a 15A load on each PCI-e connector regardless of how many rails the PSU has, what connector gets power from what rail and what other loads may potentially be on the power supply.
Click to expand...

I don't think that is right because the total required minimum output from +12V rail is 30A for NVIDIA's SLI certification. So this means that they expect the total system draw to be at least 30A, not 30A for only PCI-Express connectors. If they are indeed drawing 30A from the PCI-Express connectors for certification, then the total required +12V rail draw should be at least 40A to take into account of CPU, HDD, fans, etc...
 
TONY!!!

Thanks for your input. The more I thought about it, the more I came to the conclusion that this was crap. :)

I took the Silverstone and put a 15A load on the 12V3. I then put a 21A load on the 12V3 and it shut down as per the current limiter.

Just as it SHOULD!
 
jonnyGURU said:
TONY!!!

Thanks for your input. The more I thought about it, the more I came to the conclusion that this was crap. :)

I took the Silverstone and put a 15A load on the 12V3. I then put a 21A load on the 12V3 and it shut down as per the current limiter.

Just as it SHOULD!
Click to expand...
I opened an Antec SP II 500W & your X-Finity 500W both "seem" to have PCB's after the common plain for both +12V rails. ;)
 
Nope.. It's apparently true...

The SLI testing processed was outsourced in the past. They just recently obtained their own test equipment and are still trying to figure out what's going to work best.

They really did want 15A capability on both PCI-e's allowing for a potential 100% overhead. Which is pretty extreme.... and not very realistic. At least not with a dual rail. Single rail? Sure. Three or four rail? Only if the PCI-e's are split up on two different rails and NOT the 12V2 (if you wanted to adhere to Intel specs.)

I guess we should expect a change in the SLI certification again. :rolleyes:
 
jonnyGURU said:
Nope.. It's apparently true...

The SLI testing processed was outsourced in the past. They just recently obtained their own test equipment and are still trying to figure out what's going to work best.

They really did want 15A capability on both PCI-e's allowing for a potential 100% overhead. Which is pretty extreme.... and not very realistic. At least not with a dual rail. Single rail? Sure. Three or four rail? Only if the PCI-e's are split up on two different rails and NOT the 12V2 (if you wanted to adhere to Intel specs.)

I guess we should expect a change in the SLI certification again. :rolleyes:
Click to expand...
You indicated that the Enermax 535W & 600W couldn't generate a combined +12V@30A,
how did they get SLI Certified months ago, when the spec was +12V@34A?

nVidia should probably stay out of the PSU certification business. :rolleyes:
 
davidhammock200 said:
nVidia should probably stay out of the PSU certification business. :rolleyes:
Click to expand...

How about the SLI RAM certification business :rolleyes:
 
You must log in or register to reply here.
Back
Top