Problems between recent ATX12V v2.xx compliant power supplies and many motherboards have been raised often in the SPCR forums. The most common symptom is an explicable failure to boot. This problem seems to crop up with surprising frequency in the SPCR forums.

In the past few weeks, we've tried to replicate the problem in our lab with a variety of different motherboards and power supplies. We also spent time talking with both PSU and motherboard engineers from many companies. We believe we now understand the problem: It relates to high efficiency PSUs and minimalist low power systems.

It turns out that high efficiency power supplies -- such as those that meet 80 Plus standards but also many others that don't -- typically need a higher minimum load on the 12V line(s) than older or less efficient power supplies.

It also turns out that some recent vintage motherboards have a bit of a delay before the 12VDC power is provided to the processor. This might be to stagger the 12V power load so that the surge to the PSU is reduced. As many PSUs have a current limiter to keep the maximum power on any DC output to <240VA (to meet UL/EN safety standards), this may be useful with high power systems; If the current limit is triggered at boot, the PSU (and the system) would power down immediately.

In a minimalist system where on-board graphics is used, this 12V turn on delay in some motherboards causes the initial total current on the 12V line to drop below 1A. If the delay is long enough, then the PSU will not start.

Too low a 12V load on startup is the most common cause of a high efficiency PSU not starting with the kind of low power minimalist system that many SPCR readers build.

Usually an older or less advanced, less efficient PSU will not exhibit this problem with the same system; their minimum 12V power requirement is lower.

Adding more load on the 12V line is the obvious way to overcome the problem, but the common strategy of adding another hard drive does not always work. The typical HDD may take a little too long to ramp up to full current demand on startup, due to its electro-mechanical nature.

A more sure-fire way of ensuring startup is to add a video card. This is not always desirable, as total heat and power consumption will be increased; low power/thermals are usually key reasons for such systems in the first place. Other add-ons such as a sound card, modem card, or TV tuner card may help breach the minimum 12V load. Another fan or two could also work.

A technical solution is to add an artificial load of 6~8W on the 12V line. This could be anything from a DC light bulb to a passive >10W resistor. The problem again is that total power will be increased. However, if you can figure out how to add this load, you can probably also figure out how to add a switch to take the extra load out after bootup. (NOTE: Be aware that sleep mode could well put the system back down to less than minimal load, however. This will cause problems.)

Some PSU makers are aware of this problem, and some of them are working on a dynamic loading solution that will ensure proper operation even with extremely low load -- without affecting high efficiency.

------------------

ADDENDUM, Aug 24/06

As mentioned above, higher efficiency PSUs generally tend to need higher minimum power on the 12V line in order to simply run. Typically, we're talking about 1A or greater. Older, less efficient PSUs have much lower minumum current needs, under 0.5A and often ZERO.

In some fairly recent motherboards, there are various time delays implemented in order to ensure that the PSU (and motherboard) is not subject to a huge current surge when everything turns on all at once. Many Asus boards have been identified as doing this -- though you won't get Asus to talk about it -- I tried -- they will say it's proprietary information they don't want to share with competitors. They are not the only board makers doing this.

The practice began in the peak of the Prescott era when startup surge became quite serious, and the board makers extended the practice to AMD boards as well.

So this means, for example, that there could be anywhere between tens and hundreds of millseconds between different portions of the board and components being powered up. Just how much delay there is and how much power the CPU/VGA draw affects whether one of these high efficiency PSUs will actually start. Sometimes, adding HDDs will help, sometimes not -- they may not pull current soon enough after the power button is pressed to change the current demand the PSU "senses".

I don't have concrete information about the delays involved.... however, some of the Asus boards that would not start with some high efficiency Seasonic PSUs also would not start with some high efficiency PSUs from Antec, Fortron-Source, and Enhance.

A sure-fire way to tell whether too-low 12V start current is the problem is to hook up a known working bog-standard generic 300W PSU to the afflicted system. If max power was the problem, it would have a hard time starting -- or not start at all. But invariably, with these too-low 12V start current situations, such PSUs (even several years old ones that long precede 24-pin ATX outputs, etc) will start the system fine.

The reality is that most of the better brands -- like Seasonic and the others mentioned above -- are going for high efficiency because this is one of the big differentiators between PSUs today, and also very high power output. There are few PSUs that put less than 80% of the total power rating available on the 12V rail. For a 400W PSU, this typically means 320W is available on the 12V lines. Trust me, you simply don't get a surge that big at startup with 99.99% of computers, enthusiast or not.

The upside of all this is that most PSU makers -- well, Seasonic, at least -- are aware of the issues here, and they are implementing solutions. The simplest one is to add just enough internal resistance on the 12V rail to ensure that there is enough current draw to start the PSU even with no 12V draw from the outside. This naturally drops the hard-earned efficiency down a notch, but it is in fact, what some PSU makers have done.

I know that Seasonic has quietly implemented an active circuit that automatically inserts enough of a load so that the 12V line always sees the minimum load, at least, but then this extra resistance is removed when the load gets higher, so that turn-on is never a problem, AND high efficiency is maintained at normal and high power operation.

Which Seasonic models? I believe all the current sleeved output cable S12s, the S12-80+ models and the soon to come M12s.

Where does it say this? The only reference I could find is in the example configurations section. This section is preceeded with the statement:

This seems to clearly state that what follows are examples only, and doesn't form part of the specification per se. Furthermore, it gives an explicit warning about system compatibility if the PSU designer assumes these examples will work in all systems.

I also don't see where it states anywhere that a motherboard and/or system needs to draw 1A from the 12V at power-up.

TomZ,

Your point is well made. It's not a requirement.

When you compare the current ATX12V to earlier (pre 1.3) versions, however, you'll find that the earlier versions used lower min 12V current in their examples. The 1A used in all the examples in v.2.2 is a recognition by Intel that higher rated, more efficient PSUs need a higher minimum 12V current: This is the way it was explained to me by a couple of PS engineers.

One thing I should also mention is that most of the very high efficiency PSU samples in the lab, including all the 80 Plus approved models, had trouble booting with an Asus A8M-VM CSM v1.01 (latest BIOS) running with any A64 CPU, any HDD, a GB of RAM, and onboard graphics. Adding a PCIe graphics card -- any one of several -- allowed the system to boot. So did 6-7W extra dummy load on the 12V line. Older PSUs and less efficient PSUs did not have this trouble.

Also, the practice of adding dummy resistors inside the PSU to ensure high enough 12V load at all times appears to be fairly widespread. It's a cheap and easy solution for broad compatibility (not only with the current range of systems but also for backward compatibility). Most retail buyers are not choosy about efficiency. PSUs for bigger system integrators are a whole nuther story. Such practices (dummy resistors in the PSU) generally would not be accepted.

The more technically inclined could use one of these circuits…



Basically it provides a load for x number of seconds during power-up.

R1 is the load resistor, for example 12 Ohms for 1A current draw.
R2 is to limit the base current during initial power-on.
R3 and C1 determine the amount of time Q1 remains on after power-on.

Seems like a lot of trouble, but it would probably fix the start-up issue and still reduce power consumption.

D2.

Thanks for the info! This makes me worried, of course. I'm currently looking at an Epox board with the same chipset as the one from Asus mentioned above. I will buy a graphics card for it, but probably a couple of months later. Well I guess I could just connect something to make it start.

Using the fan headers could be a good idea, maybe even for other things than fans, because I can easily turn it off with SpeedFan. It should work as long as I'm not connecting too power hungry devices. Can I connect a resistor to a fan header? Trying to regulate a resistor with PWM sounds strange, however dropping from 100 % to 0 % is another story. The resistor would only draw power when booting up and shutting down, like 30 seconds maximum. Unless I'm in BIOS...

Ironically, it may be more efficient overall to buy an older and less efficient power supply, so as to avoid having to add extra loads such as high-powered graphics cards and resistors in order to get the more efficient power supplies to boot with a low-powered system.

I would suggest that we exercise some caution when developing general advice regarding this topic. MikeC has given some general observations that may apply to some PSUs, but I don't think we can conclude that all high-efficiency PSUs have this design. In addition, we also cannot assume that all systems built with certain chipsets and/or motherboards will have the issue.

I think the best way to make use of this information is to ask these questions of the PSU and motherboard OEMs. Usually these specifications, at least for the PSUs, are not hidden. For example, the NeoHE specifications page on the Antec web site clearly states a 1A minimum:

http://www.antec.com/specs/NeoHE430_spe.html

I also got confirmation from Seasonic that their S12-500 and S12-600 also have 1A minimums on the 12V rail. I got this information by writing them a support e-mail.

TomZ,

What you'll find is that almost all the PSUs that are ATX12v v1.3 compliant (or earlier) have 0.0~0.5A as the 12V min current while those spec'd for ATX12V v2.xx compliance almost always have 1A as the min current on 12V1 AND 12V2. On the surface of it, this suggests that each of those lines needs 1A, for a total of 2A.

But this doesn't explain what happens with those PSUs that really have only one 12V rail. Based on our experiments in the lab, it would appear that 1A on the 12V lines in total is enough.

after an expensive lesson with it and an older system. Basically, I thought the power supply was broken, so I sent it to Seasonic for warrantee work (shipping bricks around is *NOT* cheap)-; then I bought another S12-500 to swap into a system I was building which had the same power-up issue (ironically using the Asus CSM motherboard that Mike used). Seasonic verified that it was an issue w/ not enough load, so I emailed the SPCR folks (I guess this means Mike and asked that they include testing for this oddball case so others wouldn't have to waste their money on what I consider a bug in their design.

What Seasonic probably should have done was do the resistor hack (via a plug you can put on the end of a connector) to it so I wouldn't have to sell the S12 and RMA the other one. Man, that was a waste of money.

I'm running an Enermax Liberty and couldn't be happier now. The modular plugs also made fitting it into the Q-Pack much easier to boot (thanks to Seasonic for the dumb design so I could find a better PSU ...

Thanks for testing it Mike. It'd be interesting to see if Seasonic's new modular PSU has the same issue to see if they learned their lesson...

I would not advise using a fan header, drawing close to 1A through a fan header is likely to burn out the output circuitry of the motherboard. You would connect this sort of thing through a molex connector.
Dutch2's circuit would work fine (though it doesn't address standby issues) and can be put together for a few bucks.

You must have misunderstood my post. The reason for using a fan header (and not Molex) is that I only need a resistor, SpeedFan will do the rest of the work. And I did say "It should work as long as I'm not connecting too power hungry devices." We don't even know how big resistor we need. Someone else posted 1 A (the requested total draw is 1 A, so we probably need something less than that), it wasn't me. When I suggest using something instead of a fan I mean something similar in terms of power usage.

I would not recommend using the fan header. The idea is to add a load to the power supply because the motherboard is deliberately pausing before drawing any current from the +12V line. Most likely, that includes fan headers on the motherboard.

Thanks Devon, I think you're right. Although it's pretty easy to find out when the fans start spinning.

PCIE too? Does this mean that I have to use a graphics card that have a separate power connector to solve this issue? The PCIE slot is designed for 75 W max AFAIK. Even some X800 cards have no connector, the power goes through the motherboard.
I wonder if all those cards have a connector:

That's a very good question. No, I don't believe that all of them had a PCIe connector. I guess I could be wrong about the motherboard blocking all +12V power. Perhaps it just blocks the CPU line, although I suspect that this is probably up to the specific motherboard.

I'm not so sure about the delay on the fan headers, at least not on my MB. I have a 120 mm papst fan connected to the MB through a Zalman Fanmate. If I turn it to the lowest setting, my system won't boot. But if I crank it up to full speed (slightly less is enough, but why risk having to turn the master switch off and start all over ) the system boots fine. I don't think it would if there was a delay on the fan header.

Guess I'm proven wrong then. Glad to hear it.

Hehe… I wouldn’t exactly call you “proven wrong”. I only tried it on a single MB. Other MBs may be different.

Have an OEM Seasonic SS-500HT which won't boot.

My setup is:

Epox MB
Athlon 64 3200+
Geforce 6800 GT
Samsung 160 GB HD
2*512 MB RAM
Soundblaster PCI sound card
Netgear wireless PCI network card

One chassi fan, one fan on the graphics card and one for the CPU

The only thing I could add is another hard drive, which I will do after this post. Update: Didn't help.

I bought my PSU cheap on eBay, so maybe it is just broken.

http://cgi.ebay.co.uk/ws/eBayISAPI.dll? ... 6871413140

You also bought it as-is . This is hardly a low-power system with that 6800 in there.

Would this work?...

http://www.pcpowercooling.com/products/viewproduct.php?show=VLR



I know it's meant to plug into the +5v line, but you could easily move the pins so that it plugs into the +12v line. However, it might not be big enough.

It's possible that the system is already providing *almost* enough load and adding one of these might be just enough to put it over the line.

Any thoughts?

If it draws 1A on the 5V line (5W), it will therefore draw 2.4A on the 12V line. That's now 28.8W, which probably will exceed its power-dissipation capability.

You could make something like this yourself -- all you need is a ~25ohm 10W resistor + spare 4-pin molex plug like the one picture. That would pull ~6W (0.5A) on the 12V line, which is enough (in combination with whatever else is in the PC) to provide the minimum 12V load for PSU startup. Just position the resistor somewhere where its heat won't hurt anything. Maybe clamp it to the chassis.

Any idea where I could buy a 25 ohm 10 watt resistor?

I did a little googling and I find a lot of 10 ohm 25 watt resistors but not much the other way around.



EDIT: I think I spoke too soon. This looks like the right thing...

http://www.marvac.com/detail.aspx?ID=24831


Before I order this thing, are you positive that this is the correct size and rating?

That's overkill, it's a high precision resistor.... not necessary for such a "dumb" application. Try this one for 39 cents. Any electronics parts store should have them -- even your local radio shack.

Yeah, I know it's a little overkill but I like the metal casing because of the heatsink action. In addition, the mounting lugs on the sides should make it easier to mount to a metal surface inside the case.

The price doesn't really matter that much to me. Anything under $5 is cheap enough for a one-off device. It would be different if I was making a lot of them and trying to sell them for a profit.

I'm about to go work for Community Professional Loudspeakers this summer, and those look a lot like the power resistors used by the Dyna-Tech driver protection circuitry we use on our lower-end models, or models where they'll be installed in inaccessible locations.

Mike, I hope you'll let us know when the new crop of high-efficiency power supplies becomes available that doesn't need third-party modifications to work. This whole scene is becoming ridiculous. I'm glad I bought a Gateway laptop instead of building another machine. At least I can shout at Gateway if something stops working. But this whole business of manufacturers selling products that can't account for the wide array of loading conditions they'll face is just insane to me.

I don't think this issue is a widespread as you might perceive. There are lots of people building systems with these supplies (like myself) with no problems at all.

In my situation, it's a micro-ATX motherboard with integrated video, the MSI K8NGM2-FID (GeForce 6150 chipset). It probably doesn't help that it's an AMD CPU which uses less power than equivalent Intel CPUs, and there's no fans on the motherboard other than the CPU fan. And there's only one case fan. All of this adds up to a system that doesn't draw enough power on startup to keep the PSU happy.

I've added extra hard drives to the mix but they seem to have a startup delay that prevents the PSU from getting sufficient load fast enough.

I'm wondering if there's any BIOS settings that affect power usage that can make the board _less_ efficient.

I don't know about your MSI, but I've noticed that some ASUS boards come with an option "Time before PWR_OK" or something like that. Essentially, this is just the delay between the when the motherboard receives power for the first time, and when it sends a signal to the PSU to confirm that it is receiving power. On the A8R32-MVP, it's set to 400 ms, but has options down to 100 ms. I haven't tested this in person, but there's a chance that it might help those who are having trouble booting ... at least if you have a backup PSU so you can get into the BIOS in the first place.

I have tried to get some answers from MSI to my K8NGM2-FID & NeoHE problems. The delay issues was also talked: "We have conmunicate with our Power Dept, they said they didn't design any delay circuit on the board. We recommend you use other brand PSU to have a try, as we here have received other customers' reflection, there maybe some compatibility issue between the Antec PSU and the board. Sorry for any inconvenience caused!".
Only solution seems to be a dummy load or use of a less advanced, less efficient PSU, if integrated GPU is used.