Is this possible from a psu?

ted34 · 30 Apr 2006 at 21:48

I have a tagan 530w powering my rig as in sig except with a fx60 and a x-fi. The rails are always very stable according to mbm. This changes when i run a stress test. Normally the rails stay completely stable but every now and again a mbm alarm comes up saying 5v rail dropped to 2.17v. The 3.3v rail sometimes does similar. Thing is the comp doesnt crash or fail it stress test and the rails only drop for a split second. Is this likely or is mbm lying to me. Id have thought comp would crash immediately if 3.3 or 5v rails dropped that low

Bennah · 30 Apr 2006 at 21:53

Could be just a software problem. Reading the wrong values. Had the same problem with smartguardian.

Tried to read the values with a Multimeter?

If the 5v rail and 3.3 rail dropped that low, no doubt the system would power off.

Joe42 · 30 Apr 2006 at 22:48

Put a multimetre between the red and black connections on a molex connector to measure the 5v rail.

If any of the rails drop below 5% of their normal ratings the psu should shut off if its a decent one, which it is.

ted34 · 1 May 2006 at 11:08

Thanks for quick responses guys. Its a tagan 530w (u22 model). 12v rail dropped for a split second to 5.17v earlier this morning - set mbm alarms off and woke me up

Can measure with a multimeter but it only happens for a split second and is random when it does so would have to sit there watching a multimeter screen for however long it takes

Bennah · 1 May 2006 at 11:41

Hm...

So its donig it at Idle now?

Any bad connections with the psu to devices etc?

Has the PSU been doing it all the time since you got it? Started recently? Added any new hardware?

Shame, I was going to get one of these PSU's. From the things Ive read, they are great for the money. Defo worth a buy anyway.

I think this could be an isolated incident though.

ted34 · 1 May 2006 at 14:23

Its only ever done it when stress testing except this morning. Theres no dodgy connections though and theres not much in my system thats a power drainer. I might install ite smart guardian and see what that reports as im guessing it must be mbm as its got the dfi config (from dfi street) on it allowing it to run nf4

fornowagain · 1 May 2006 at 14:33

ted34 said:
Thanks for quick responses guys. Its a tagan 530w (u22 model). 12v rail dropped for a split second to 5.17v earlier this morning - set mbm alarms off and woke me up

Can measure with a multimeter but it only happens for a split second and is random when it does so would have to sit there watching a multimeter screen for however long it takes

I'm not convinced, in my experience the CPU would crash bigtime. If its long enough for you to see it, its an eternity for the CPU. It could be a sensor spiking on its way out. If its a 100ms long flash a cheap DMM won't register.

Joe42 · 1 May 2006 at 14:36

The cpu has its own voltage regulators and they can usually cope with a wide range of voltages i believe. The chipset and ram are the same.

fornowagain · 1 May 2006 at 14:44

Joe42 said:
The cpu has its own voltage regulators and they can usually cope with a wide range of voltages i believe. The chipset and ram are the same.

Do you know how they work? Regulators do just that, they clip over voltage to a few percent before they melt. They don't up voltage, that's what caps do and they don't support voltage drops like that. Its called smoothing, maybe 0.25v for a few microseconds at CPU current levels. Motherboard caps holding a 7v drop at 100W for 100ms or more, no way.

ted34 · 1 May 2006 at 14:44

I was thinking it would be far more likely to my dfi xpert reporting wrong than the psu. As its in prime small ffts while its doing it id expect immediate crash or at least prime failure. Hasnt done it once today while stress testing and as far as i remember it never happened with my 3700 only dual cores. Cant be that 530w isnt enough power can it:

dfi xpert, ldt@ 1.3v, chipset @ 1.52v
fx60 @ 1.392v
2gb corsair ram @ 2.66v
x850xt
2 dvd-rw drives
3 seagate hd
x-fi
zalman fan controller so all fans are powered from a molex.

It did happen when i had an x800 in there too.

Jimbo Mahoney · 1 May 2006 at 19:14

I'd say with 99% certainty it's the mobo reporting dodgy every so often. almost every mobo I've ever had does this, normally with temps (you know, sometimes reporting 127'C for a second or something...)

Joe42 · 1 May 2006 at 19:39

fornowagain said:
Do you know how they work? Regulators do just that, they clip over voltage to a few percent before they melt. They don't up voltage, that's what caps do and they don't support voltage drops like that. Its called smoothing, maybe 0.25v for a few microseconds at CPU current levels. Motherboard caps holding a 7v drop at 100W for 100ms or more, no way.

The voltage regulators produce a regulated voltage for the cpu or ram or chipset from the 3.3v rail.
Usually they can cope with a variety of input voltages and still produce the same regulated output. So fluctuations in the 3.3v rail would be unlikely to cause problems for components such as ram, chipset and cpu, but might cause problems for the rest of the components on the board that don't have their own voltage regulators and run off the 3.3v rail. I believe thats how it works.
I've seen proposals to change the atx psu standard so that the psu produces an unregulated 5v output and no 3.3, and apparently this would not require a voltage regulator re-design and increases the efficientcy of both the psu and the regulators, so i think they must be pretty flexiable.

fornowagain · 1 May 2006 at 20:25

Joe42 said:
The voltage regulators produce a regulated voltage for the cpu or ram or chipset from the 3.3v rail.
Usually they can cope with a variety of input voltages and still produce the same regulated output. So fluctuations in the 3.3v rail would be unlikely to cause problems for components such as ram, chipset and cpu, but might cause problems for the rest of the components on the board that don't have their own voltage regulators and run off the 3.3v rail. I believe thats how it works.
I've seen proposals to change the atx psu standard so that the psu produces an unregulated 5v output and no 3.3, and apparently this would not require a voltage regulator re-design and increases the efficientcy of both the psu and the regulators, so i think they must be pretty flexiable.

Stop it your killing me

. For a start the cpu is fed from the 12v rail on A64's. I will say it again, listen. Vregs can't produce voltage, they clip voltage its known as a breakdown region. They are a three pin gate device controlled from the third pin which is a resistance setting the gate. The larger the difference from their set gate value the greater the resistance and the hotter they get. Capacitors hold small amount of current and make up small voltage drops. This is called a regulated smoothing circuit. It is designed to smooth out changes of the line voltage, this is called a ripple voltage of maybe 5% maximum. If the 12v dropped to 5v nothing on a motherboard would be able to hold it at 12v.

If you wish to understand a gate reg look up the word zener.

Joe42 · 1 May 2006 at 20:38

Why would they need to produce voltage?
They don't need to hold it at 12v. They regulate the 12v down to 1.4 or whatever the cpu requires, same for the chipset, ram. Nothing in a pc boosts voltages, its reduced down to different levels.
Those regulators that regulate 12v down to 1.4 will work quite hapily if the 12v falls down to 5v or even lower depending on how efficient they are.
The proposals to increase efficientcy had them working off an unregulated 5v rail from the psu as both the regulators and the psu seem to be more efficient that way.

If the cpu is regulated off the 12v rail then presumably the 3.3 is for the rest of the logic on the board, so fluctuations in that might cause some wierd things to fail or fry.

fornowagain · 1 May 2006 at 21:04

Joe42 said:
Why would they need to produce voltage?
They don't need to hold it at 12v. They regulate the 12v down to 1.4 or whatever the cpu requires, same for the chipset, ram. Nothing in a pc boosts voltages, its reduced down to different levels.
Those regulators that regulate 12v down to 1.4 will work quite hapily if the 12v falls down to 5v or even lower depending on how efficient they are.
The proposals to increase efficientcy had them working off an unregulated 5v rail from the psu as both the regulators and the psu seem to be more efficient that way.

You said produce not me. Admittedly its been a few years since I had to design low current, most of mine is in Megawatts but come on. Without a schematic, lets see, I'll wing it. There's going to be a series of steps taking voltage down and increasing current. The first step will be to hold incoming voltage as a reference acoss the board, every single PCB since the dawn of time does the same. Then voltages are dropped in ratio and coil choked for interference then regulated. Why would they design an entire board that could run on 5v but feed it 12v. And then why would they try to hold voltages within 5% if it could handle 120%. No board designed for an input voltage of 12v is going to be able to handle even a fraction of the voltage variation you describe.

Joe42 · 1 May 2006 at 21:17

Bit of confusion about the produce thing. I said they produce the cpu voltage (around 1.4) from the 3.3, but they actually use the 12v as you pointed out. Its still lower than the supply though, i never said they raise the voltage.
As you say most things can't cope with large variations in the voltage but apparently these voltage regualtors can. Why shouldn't they be able to?

I admit i'm not an expert on this, but i do know someone who designs powersupplys for military aircraft for a living and i pick a lot up from him.

He picked up a presentation from someone doing a phd on pc power efficientcy, and one of the proposed designs used a psu with an unregulated voltage rail somewhere below 12, can't remember if it was 11 or 5, and apparently it would not require a redesign of the voltage regulators.

I might go and see if i can find that presentation, there was probably a web address on it somewhere.

Edit: I've got bits of that presentation but not the slide i was looking for. I'm pretty sure he used unregulated 11v and 5v and used extra voltaged regulators to get the 3.3 from the 5.

fornowagain · 1 May 2006 at 21:25

Joe42 said:
Why shouldn't they be able to?

For one, heat. Depends on the package and like you say they can clip a lot of volts. But it's going to be controlled within a small range for stabilty. I think they'd select for its running temps with a certain maximum/idle wattage, input voltage, clipped tolerance and spikes/transients/RF. It's just a gate it passes or holds within a range, they (ones I've seen) look like TO-220 types, three pins. Nothing special, but its good for a range before it dies in a big black mess. They'd control it tightly because of all the timing circuits and large input variations make a lot of transients.

Its an interesting subject, well for me anyway, but it's a bit to much like work

Ask your mate, sounds like he'd have detailed answers, not my area and its a big one.