So basically you can buy the highest end Seaonic and be underpowered on 12 V1.

7800GTX ~100W
Another 7800GTX ~100W
Three Raptors 70W
DVD burner 10W
Another burner 10W
Case fans 5W

295W ... right there is 25 amps and Seasonic S12-600 will not provide but 20 right? Nevermind cathod tubes, 12V water pump, etc..

Add that to an overclocked pentium dually...say at 3.8 thats another 300W just on processor (a single 3.8 uses about 150).. So 25 more amps the 12V2 can't provide for either being capped @ 20A.


I'm not so sure you should say in your PSU reviews anymore comments about overkill on PSUs when you review them. Seems like this system needs a $300 Zippy 700 which has 45A single rail.

Absolutely not. You have taken the opposite meaning of what I've said. ALL of the total 12V current is available on any 12V lead if it is not being demanded on the other leads.

Not if CPU is using 300W like my example. Here they show a P670 using 162W. http://www.xbitlabs.com/articles/cpu/di ... 670_3.html

I assume two of them, like a 840XE overlocked to 3.8 would use double the 670's power rendering the 12v1 almost powerless for such things like SLi GTX's..drives..fans..

You assume wrong; it's not double. It's around +30% for Intel dualcores right now, maybe a bit more.

Also, the xbit number you cite appear to be for total system AC power draw, NOT individual CPU power draw. No way a 670 pulls 162W DC.

Your 7800GTX numbers are too high, too.

But let's say you really did have 600W power draw on 12V lines from any system of components... you'd have a problem with enough power. Period. Never mind how the 12v is distributed.

We will continue to make recommendations on PSU power as we see fit. The example you cite is the most extreme conceivable. If you're building a server, you'd go to server class components. And if you're not, why would you possibly want to saddle yourself with such a bloated PO... as you're talking about?

You assume wrong; it's not double. It's around +30% for Intel dualcores right now, maybe a bit more.
-------------------
You're guessing. How can one chip use 167W and the second, identical chip, not use the same?

Also, the xbit number you cite appear to be for total system AC power draw, NOT individual CPU power draw. No way a 670 pulls 162W DC.
----------------------
Youre guessing again and wrong. Or are you saying a whole athlon 3800+ system listed on same chart uses only 59 watts as xbitlabs same graph describes when AMD's TDP is 67W for that processor alone? It's trivial to measure CPU power these days. All you do is use a Fluke AC/DC current probe to the isolated +12V supply lines feeding into the CPU VRM. Intels 670 TDP is 132W and many know how erronous that is. Plus other sites have measured it as well and come with similar numbers. I provided one link but you can search around in interested what CPU's really use.



Your 7800GTX numbers are too high, too.
-------------------------
Every site says it uses over 100W.. then if you start overcloking it's more.

We will continue to make recommendations on PSU power as we see fit. The example you cite is the most extreme conceivable.
----------------------------
Hardly extreme. Excessive maybe. Your "high end" gaming machine OTOH is more like low/mid end therfore a poor example to base ones high end needs upon. I mean single core instead of dual core. Low end 3.2Ghz single core instead of 3.8. Single video card, instead of Sli or Crossfire. No Raid on drives. I say this becuase I have bottomed out a couple "high wattage" PSUs before.. Usually what happends while running 3dmarks which has lots of disk hashing 100% processor, and 100% graphics cards load the machine will just shut down. Course this never happend with high end PCP&C.. it's not exactly what i'd call quiet and needs extensive mods to make livable and I water cool everything else with a triple length rad and three 120mm Nexus fans.

In view of the info in this thread is there really any advantage/disadvantage to hooking up the 4-pin 12V auxiliary connector to the motherboard vs. just using an adaptor off a regular molex? In some cases (P-180) this could be a better option for tidy wiring.

I was under the impression that this was a big no-no but since they're really drawing power from just one +12V supply... or am I misunderstanding this?

Mike,

Is this only true if the power supply does not have a 240va current limiter on the rails? Do the seasonics, s12-600 in particular, have current limiters on the rails?

I ask this because I need to find a new PSU on my storage server with 18 drives in it. If I get the s12-600 would all 12v amperage be available as if it were a single rail PSU? I was consider making some kind of adapter to pull 12v off the 4pin connector and take the 5v and another ground off a regular molex.

This PSU does not need to power anything else but hard drives (no mobo, no video, etc.)

psu requirements are vastly overated by manufacturers and other so-called experts in the field. Theres more bullshi.t about psus than any other component. A high end system with an FX57 and 2 7800gtx in sli does not require a. 500w, b. 2 12v rails or c. more than 22 amps on the 12v rail. Of course, this is with the proviso that its a good quality psu.

Thank you for a good informative article. You have cleared things up for me. One questions what about the pcp&c 850. There seem to be 3 separate circuit boards each containing a power supplies and I have seen this in a few power supplies. You’re article would apply or not apply to these few power supplies?

Also I have always preferred one large 12v rail over the new multiple rails. Any computers I have built computer seem to run better and smoother with a power supply with just one large 32amp -36amp 12v rail. I have in fact scored 1000 more in 3dmark just by switching to a single rail power supply. (Not including the pcp&c & seasonic 600w) Is there a reason for this? Thank you

Quoted from here, at the end of the first post. That is, of course, unless you are talking about something entirely different.



In addition, I don't believe this 4-in connector will actually fit. The shapes of the pins are different.

It is sometimes suggested that the dedicated 4pin ATX12V lead for the CPU is by itself on the 12V2 "rail" because the CPU neads "cleaner" power. At first sight that is understandable because the 12V1 "rail" probably carries a bit of RF noise due to the length of the cables and the fact that some electric motors (although small ones) are connected to 12V1 - in particular one might expect to see a suddenly different current draw on 12V1 when a hard drive is spinning up.

However this theory seems to be nonsense since every motherboard surrounds the CPU by its own high spec power regulator, so even if the 12V input had quite bad RF noise or fluctuations in voltage that would not get through to the CPU in any way at all. (A similar point is that even high end audio equipment does not benefit from being connected to the wall using special $5,000 power cables constructed of silver, but that is well off topic ...)

This suggests that the only difference between all the different 12V connectors in your PC is the length of wire connecting them to the PSU: the longest drive connectors, which could be up to around 3 feet long could have a voltage drop of up to around 0.1V (3 feet of 20AWG wire has a resistance of around 0.03 ohms; current in that wire could be 8 amps max if four drives are connected although the full current does not flow through the full length of the wire).

Note: I called 12V1 and 12V2 "rails" although from reading this thread they are probably connected together inside the PSU or at best they are derived from the same 12V regulator but they pass through two separate current limiter circuits.

Mike,

Thanks for this very nice summary. I have just read your latest PSU roundup and was surprised to see something confirmed that I have suspected for a long time!!!!

I am beginning to get a little confused by the 8-pin EPS connector. It is found on more boards these days to provide "more stable power" to the CPU. Do you think this is really necessary for the latest dual core systems, or would a 4-pin connector suffice? If you have a 8-pin mobo connector but only a 4-pin connector on the PSU is it safe to use a 4 to 8 pin adaptor?? After reading your article and the ATX and SSI guidelines I suspect it is, as I am guessing that both come from the same source within the PSU, but I would like some confirmation if possible.

Cheers.

Sorry I'm sort a noob at this. I read through the thread but I am still confused about something.

I just bought a Zalman 460B-APS PSU (based on the review here) and on the 4-pin connector that's wrapped together with the 20-pin connector, there's a warning sticker that says don't connect this with the 4-pin ATX 12V connector, it may damage the cpu and motherboard. The motherboard I have is an Asus P5B Deluxe and it has both a 24-pin connector and the 8-pin 12V connector.

Does the warning only mean to say don't connect that 4-pin connector with one of the 4-pin ATX 12V connectors into the 8-pin connector as a pair, or does it mean, if I'm using the 8-pin EATXV12, I should only have the 20-pin connector cable in the 24-pin connector?

Basically I want to know if it's safe to put the 4-pin + 20-pin connectors into the 24-pin connector and put both of the 4-pin 12V connectors into the 8-pin EATX12V connector.

Thanks in advance

Yes.



It's absolutely safe.

Mike - thanks so much for posting so clearly on a such a muddy topic. You do us a big favor.

I have a question about the "independent" 240VA protection for the 12V rails. I understand that the regulation is all done at one point, and that (except for a few supplies) all ATX units source the 12V from a single transformer, but I'm not real clear on the protection circuits for the individual 12V rails. I've been inside maybe half a dozen recent supplies from as many different manufacturers and have yet to see any form of protection on the "independent" 12V rails - no fuses, no PTC's, no FL's, nothing. Every supply had the grounds and the 12V lines brought to a single area of the main PCB and soldered together. On the modular units I looked at, some had separate "zones" in the distribution panel for the different 12V rails, but still no visible method of independent current limiting.

Have you seen an example of independent 20A/12V current limiting on recent ATX supplies? I'm wondering if ALL manufacturers didn't simply omit this part of the "recommended" spec.

Thanks again.

You're welcome. I expect that not all points are covered by my piece, simply because different mfgs are taking varying approaches to the problem as they keep increasing the power capacity. I refer to the many >700W PSUs that are now being flogged. IMO, those PSUs (and their marketing depts) should be flogged. Oh, and let's not let the vidcard makers off the hook -- ATI and nVidia need to be lined up as well.

As to whether the 240VA current limiting is really there, yes, it is, in some but certainly not all PSUs. We've had a couple instances in the lab where we managed to exceed 20A on a 12V line and caused the PSU protection circuit to trip. Off the top of my head, I can't recall which make/model. It's certainly not a fuse, it's some kind of auto-reset device. Turning the PSU off, removing or lowering the load and then powering it back up clears the problem.

We've had many more instances where the total 12V load exceeded the total rated 12V current -- ie, generally at near maximum load -- and tripped the current limiter/protection. You can tell it's that because if the load is redistributed to that the 5 and 3.3V lines take more of it and the 12V load is reduced while the total power draw is kept the same, the protection does not trip.

In any case, going back to the safety issue, I doubt very much that you could exceed 240VA off a single wire set. The only ones you might be able to do it with are the AUX12V and PCIe power connectors. And even there, the odds are very low; 240W is a LOT for any CPU or video card to draw for any length of time -- even in a momentary peak. I don't believe there are any CPUs or GPUs being made today that can pull that kind of power.

What about dual vidcards, you say, or dual CPUs? Well, the number of cable sets automatically doubles -- you need a 4x12V for dual CPU and twp 6-pin (or 8-pin) PCIe connectors, which are always one plug to a lead.

ATi's forthcoming R600 GPU's are supposedly in that ballpark:

http://www.vr-zone.com/?i=4622

sorry guys but i'm confused as heck after reading all the info in this thread. Some one please break it down for me, is dual +12v rails good or bad? or it has pros and cons? I see all SLi certified PSU now have 3 or even 4 +12v rails.

thnx in advance

From what I've read on this subject multiple rails have no real and practical advantages. Many people say you need multiple lines for better voltage regulation. Tagan Dual Engine with genuine dual transformers achieved best voltage regulation in the history of SPCR. But lo & behold - Silverstone Olympia (reviewed here by Jonny Guru) - boasts a single, massive 12V rail with 54A (sic!) and also regulates voltages like a dream. Seasonics have single lines and who cares?

Very informative. I actually learned something today

I'm still having trouble trying to find the answer to a question. I'm looking at the EVGA 8800gts, their website says



Do I have to assume that those Amp numbers are for the 12v lines as a whole? Since I can't seem to find any 12v rails that supply 26amps/30amps individually?

And what is the actual amp draw of these video cards?

In this case, would the Enermax 82+ 425watt suffice when running SLI? All other things being typical of quiet builds that characterize builds in the gallery.

Is there enough safety overhead or should I just go with the 525w?

You need to go to the Enermax site to examine the detailed output specs. A combined total of 33A is claimed for the 425W. The 525W version has 40A. According to xbitlabs, the XTX GeForce 8800 GTS 512MB reaches 110W peak in 3D gaming, while the generic GeForce 8800 GTS 512MB hits about 100W. So assume 110W max -- two would be 220W. Now you had to estimate the maximum peak 12V demand of your processor. If it's a modern dual or quadcore, 120W would be the tops no matter what processor you have. Then add any other 12V current draws -- say max 6A (60W) from HDDs, mobo, whatever... So you have a max theoretical of 400W.

The 425W model would handle this, but there's not much headroom, and if you pushed it to the limit for any length of time, it would get loud. You can assume that with most PSUs, even quiet ones approved by SPCR, they stay quiet only to about 50~60% of maximum load. The best might make it to 70%... Anyway, the 525W would be a smarter choice. The 625W might even be better. (These comments apply a bit less if you have a separate chamber for the PSU like in the Antec P18x series of the Fusion/2480 -- the PSU will stay quieter to a higher load.)

Ok, I think I get all this but just to make sure and because I fried my previous 7800GT card (memory)

I'm running the following system with a Seasonic S12-430W:

- Q6600 (G0) @ 3Ghz
- 3 SATA drives
- 8800GT
- 2x2GB RAM
- Audigy 2 ZS
- Gigabyte P35-DS3P

Fans:

- Front 12cm Kama Bay @12V
- Front Papst 12cm @ 7V
- Rear Tri-cool 12cm @ 7V
- Scythe Ninja @ 7V
- Zalman VF700 @ 5V

Since the Seasonic supports PCI-express but does not provide a power connector I have to use a 2xMolex to PCI adapter.
I have used 2 molex connectors from the same line to power the 8800GT.
According to what I have been reading the amps will be provided through the single rail if needed, is that correct?
I only have the 8800GT connected to the rail (shortest rail)
Also, if the rails would provide seperate amps how do you recognize them as V1 and V2?

Thx

this is something works, thanks for the help

I recently did some modifications to a Nexus Value 430 PSU and I noticed that all 4 12V lines were connected to one 12V supply. what they did is each group of the 12V outgoing lines all connect to different spots on the circuit board 12V1, 12V2, 12V3, and 12V4 all have there own wire landing area. they then used a piece of copper wire to jump from this landing zone to the single internal 12V buss. I suspect that the piece of copper wire is there "safe" guard system. sort of like a fuse as there is no reason why they needed to do that if not for some safety factor. all of the 12V wires could have just landed directly to the 12V internal buss.

i'm guessing they're fairly efficient, on the order of 80-90%, but a little bit of power does get dissipated as heat.