+5v averaging ~3V?

K

kumquat

Supreme [H]ardness
Joined
Dec 7, 2005
Messages
5,269
I'm looking at the CPUID Hardware Monitor and my +5V is fluctuating between 2V and 5.4V quite a bit. The 3.3V and 12V look stable, but the -12V goes between 4.3 and 13.5.

I just swapped in another PSU (Cooler Master 460-watt) in place of my XClio Stablepower 460W. I'm not 100% sure the 5V was fluctuating that much with the XClio but I know the -12V was.

Is this normal? Faulty sensor of some sort? Or could it be a sign of a bad motherboard? The system has crashed hard on a brand new Win7 installation a few times in the past few days since setting it up. It's not the RAM, HD, or CPU (I have spares of each I've swapped in).
 
you need to check the voltages with a digital volt meter...get a cheap one from radio shack or so for like 20 bucks and pin out the voltages when the software is showing wrong voltages...software is known for wrong readings....
 
a voltmeter will not be fast enought to keep up with the changes....the ONLY way to be sure is to use an oscilloscope to get real time readings of the voltages.

<- used to do component level repair and a scope is pretty much the only way to go when you want accurate information

what are your system specs cause it seems like your system is pulling more than what the regulartor in the PSU can handle.
 
YeuEmMaiMai said:
a voltmeter will not be fast enought to keep up with the changes....the ONLY way to be sure is to use an oscilloscope to get real time readings of the voltages.

<- used to do component level repair and a scope is pretty much the only way to go when you want accurate information

what are your system specs cause it seems like your system is pulling more than what the regulartor in the PSU can handle.
Click to expand...

an analog one will be
 
there is a reason why we didn't use VM to measure voltages analog or otherwise when it came to doing PC repairs......
 
EricFX1984 said:
an analog one will be
Click to expand...
No, it won't be, and an analog meter is pretty much worthless for precision measurements.

An oscilloscope really isn't necessary unless you have reason to believe there is a real problem. A DVM/DMM however is perfectly sufficient to confirm that voltages are where they should be.
 
Zero82z said:
No, it won't be, and an analog meter is pretty much worthless for precision measurements.

An oscilloscope really isn't necessary unless you have reason to believe there is a real problem. A DVM/DMM however is perfectly sufficient to confirm that voltages are where they should be.
Click to expand...

if the voltage is constant then sure a DVM will suffice but when you see voltage fluctuations or the meter cannot get a read you NEED a scope to see what is going on. You can get used scopes for cheap <$350.

OP has ripple on his outputs and the wild voltage swings the BIOS or other software is picking indicates that he needs to take a closer look DMM will not pick up ripple if it is faster than it's sample rate......even with cheap boards I never saw voltages swing like that unless the PSU was the problem.....
 
YeuEmMaiMai said:
if the voltage is constant then sure a DVM will suffice but when you see voltage fluctuations or the meter cannot get a read you NEED a scope to see what is going on. You can get used scopes for cheap <$350.

OP has ripple on his outputs and the wild voltage swings the BIOS or other software is picking indicates that he needs to take a closer look DMM will not pick up ripple if it is faster than it's sample rate......even with cheap boards I never saw voltages swing like that unless the PSU was the problem.....
Click to expand...
The software readings aren't fluctuating because there is ripple in the output, they are fluctuating because they are wrong or because the software is reading them incorrectly. Ripple is too high frequency for the sampling rate of a DMM, but it is also much too high frequency for the motherboard sensors to pick up as well.
 
Never mind hardware monitor, mobo voltage readings through software are usually inaccurate. The only way to verify your voltages is directly with a PSU tester like Coolmax.
 
Zero82z said:
The software readings aren't fluctuating because there is ripple in the output, they are fluctuating because they are wrong or because the software is reading them incorrectly. Ripple is too high frequency for the sampling rate of a DMM, but it is also much too high frequency for the motherboard sensors to pick up as well.
Click to expand...

say what? PSU will definately show excessive ripple when you are drawing more than what the Voltage regulator can handle. you know sorta like using a 300W PSU where a 350W is recommended. there will natually be some ripple but it should way less than .05Vp-p for a decent PSU

I worked on an early athlon system where my friend used a 250W PSU instead of the one recommended and when you went into the BIOS the machine showed Wild fluctuations in the voltage readings, once the PSU was replaced, voltages only fluctuated by .03V in thre BIOS and the machine no longer shut down and rebooted randomly.

so instead of saying that the software is wrong, lets move on to what the OP needs to do and replacing the PSU may be it.

btw Ripple can be at any frequency as it is defined as the AC volage on a DC output.
 
YeuEmMaiMai said:
say what? PSU will definately show excessive ripple when you are drawing more than what the Voltage regulator can handle. you know sorta like using a 300W PSU where a 350W is recommended. there will natually be some ripple but it should way less than .05Vp-p for a decent PSU
Click to expand...
Where did I say that PSUs never output ripple? I'm not a moron. If you're going to nitpick about how a DMM doesn't have a high enough sampling rate to capture ripple then maybe you should think about the fact that the motherboard sensors cannot pick up anything that a DMM cannot. If the sensors show large voltage fluctuations and a good DMM doesn't show anything, then it means the sensors are reading wrong.
YeuEmMaiMai said:
btw Ripple can be at any frequency as it is defined as the AC volage on a DC output.
Click to expand...
Ripple can be at any frequency, but ripple coming from a PSU will be no lower than 60Hz.
 
oh my, its all about sample time and needle damping. Everyone is right. I would think that if he can see it in the software you could see it on a DVM and a fast meter but I dont know the sample time or what kind of software averaging is being used.

My comment to the OP is that -12 is not used anymore and so any reading should be ignored and if the +5 was dropping to +3 volts for real your hard drive data would be crap or other severe issues would be happening.

If the machine is working properly I suspect is it the program not reading the sensors correctly.

You should use a utility provided by the motherboard manuf or try a DVM shove into a spare molex.

if the +5 is going down to +3 some ripple riding on the outputs is the least of his worries.
 
Windows doesn't run unless all the positive voltages are within about 5-10% of specs, so wierd readings mean the software or the monitoring hardware is wrong.

Here's what I got with an ECS K7VTA3 mobo for the +12V rail:

HWmonitor: 10.62V
SpeedFan: +6V - +8V, twice a second
digital meter: 12.11V

kumquat said:
The system has crashed hard on a brand new Win7 installation a few times in the past few days since setting it up. It's not the RAM, HD, or CPU (I have spares of each I've swapped in).
Click to expand...
How do you know the RAM is good? What brand of chips does it use? If you can't see that information, then the RAM was probably made from UTT (UnTesTed) chips. Chip brands do not include Kingston, G.Skil, Mushkin, Corsair, etc.
 
larrymoencurly said:
How do you know the RAM is good? What brand of chips does it use? If you can't see that information, then the RAM was probably made from UTT (UnTesTed) chips. Chip brands do not include Kingston, G.Skil, Mushkin, Corsair, etc.
Click to expand...

An hour of Memtest on both some Corsair XMMS and OCZ Platinum 2x1GB.

I replaced the PSU with a brand new Antec Neo ECO 520C and the voltage readings are still swinging just as they were before. The system appears perfectly stable now though. Time to retire that old XClio i guess.
 
larrymoencurly said:
How do you know the RAM is good? What brand of chips does it use? If you can't see that information, then the RAM was probably made from UTT (UnTesTed) chips. Chip brands do not include Kingston, G.Skil, Mushkin, Corsair, etc.
Click to expand...
kumquat said:
An hour of Memtest on both some Corsair XMMS and OCZ Platinum 2x1GB.
Click to expand...

Memtest is a terrible Windows-based diagnostic that cannot be trusted, and I've had modules run fine for as much as 5 hours before MemTest86 reported an error.

What brands of memory chips are on the Corsair and OCZ modules? This information will be plainly visible on any good quality module.
 
larrymoencurly said:
Memtest is a terrible Windows-based diagnostic that cannot be trusted, and I've had modules run fine for as much as 5 hours before MemTest86 reported an error.

What brands of memory chips are on the Corsair and OCZ modules? This information will be plainly visible on any good quality module.
Click to expand...

Memtest is not Windows based.. And while it can't possibly detect every single memory problem out there, it's usually sufficient for knowing whether instability is related to memory problems, or if you need to look elsewhere.

Most newer Memory modules have heat-spreaders on them, and thus no way to see what the chips under them are. Not that it makes any difference, because even the best chips in the world can have defects, and the worst ones can sometimes do their job.

Finally, Windows will run with voltages beyond the +/- 5% spec.. it may not run long though, and your hardware will suffer after a while from running beyond specs.
 
larrymoencurly said:
Memtest is a terrible Windows-based diagnostic that cannot be trusted, and I've had modules run fine for as much as 5 hours before MemTest86 reported an error.

What brands of memory chips are on the Corsair and OCZ modules? This information will be plainly visible on any good quality module.
Click to expand...

Memtest is not Windows-based. Both sets of modules have glued-on heat spreaders that I'm not about to take off, and the crashing happened the same with either set in.

Replacing the PSU has solved the problem.
 
just to be sure, your not running both mem sets at the same time, correct?

and 3v on a 5 v line during any operation (even with the cpu idle) will cause a hard crash, 5v is the same across the whole system (there is no ' separate rail ' on the 5v lines for different components)... im willing to bet that cpuid is pointed at some weird point in the system and not at the 5v rail if its still doing it. i had a system that read 5v-vsb (which should read ~5v) as a percentage, but output it as voltage... imagine my surprise when it spiked to 60.25v...
 
Thor said:
just to be sure, your not running both mem sets at the same time, correct?
Click to expand...

Correct. I had the OCZ in the system forever, and I swapped out the mainboard for a different model with gigabit ethernet. That's when the crashes started. I had the Corsair as spare and swapped it in for the OCZ to no effect.

I guess this motherboard was slightly more picky about the PSU.
 
ryan_975 said:
Memtest is not Windows based.. And while it can't possibly detect every single memory problem out there, it's usually sufficient for knowing whether instability is related to memory problems, or if you need to look elsewhere.

Most newer Memory modules have heat-spreaders on them, and thus no way to see what the chips under them are. Not that it makes any difference, because even the best chips in the world can have defects, and the worst ones can sometimes do their job.

Finally, Windows will run with voltages beyond the +/- 5% spec.. it may not run long though, and your hardware will suffer after a while from running beyond specs.
Click to expand...

Unlike MemTest+ and MemTest86, MemTest is Windows based, according to its author: "MemTest is a RAM reliability tester that runs under Windows."

Samsung puts heat spreaders on some of its DDR3 modules, but I don't know of any other quality modules that include them.

How can the best chips in the world have defects? By definition, they can't.

I didn't know Windows came with a voltage spec. :D
 
larrymoencurly said:
Unlike MemTest+ and MemTest86, MemTest is Windows based, according to its author: "MemTest is a RAM reliability tester that runs under Windows."

Samsung puts heat spreaders on some of its DDR3 modules, but I don't know of any other quality modules that include them.

How can the best chips in the world have defects? By definition, they can't.

I didn't know Windows came with a voltage spec. :D
Click to expand...

1) I've never heard of that program, and I bet most people around here haven't either. When Memtest is mentioned, it's usually meant that it's the software that is downloaded and ran from a bootable environment outside of any protected-mode operating system. Only an idiot would think a program thats test memory while still inside an OS that runs in protected mode would be reliable.

2) Most DDR3 modules available for sale include heat spreaders. It's usually the "value" products that don't include them.

3) Chips known to be the most reliable in the industry can still have individual specimens that are defective. That's what I meant by "best chips", but somehow I think you already knew that.

4)You're the one that said Windows won't run outside of 5-10% specs. ATX spec for +5v, +12v, and +3.3 is +/-5%.
 
This isn't a DDR3 system, by the way, guys. The RAM sets are Corsair XMMS and OCZ Platinum, both midrange units with heat spreaders.

The RAM is also fine. I'm not sure why people are still talking about it.
 
I begin to suspect a software issue.
Wish I had a lot more informatin on what "crashed hard" really means as in exactly what does the machine do ? Have you got the OS set to collect error information or just reset on a crtical system error ? Anything in event logs ? BSOD ? Just freezes ? What are you doing when the problem occurs, random, just games etc. Does it show up immedately or random or only after the machine has been in use for a couple of hours. That kind of things.

Off hand as a first step if you are still having trouble, I would reset the machine to a no-OC factory default state, go in an make only the bios changes needed to support your hardware/drives and reinstall the OS making sure to delete the existing partitions and start fresh. Then load video driver and reboot. Load motherboard sound/lan etc. drivers rebooting between each. Then windows update. Then windows defender. Then just one or two of your games and see how it goes keeping notes on what was going on when it crashes.

Thats all I got.
 
The problem was solved by swapping out the PSU.

BTW, it was never OCed.

I wasn't really looking for help solving my crash, though the thoughts and efforts are appreciated. I was wondering why the software was showing such widely varying voltages, and the answer was that the software is unreliable.
 
Zero82z said:
Where did I say that PSUs never output ripple? I'm not a moron. If you're going to nitpick about how a DMM doesn't have a high enough sampling rate to capture ripple then maybe you should think about the fact that the motherboard sensors cannot pick up anything that a DMM cannot. If the sensors show large voltage fluctuations and a good DMM doesn't show anything, then it means the sensors are reading wrong.

Ripple can be at any frequency, but ripple coming from a PSU will be no lower than 60Hz.
Click to expand...

now that you're done, you'll see that I did indeed state that the PSU was the problem and replacing it did resolve the issue.....

BTW I was a component level tech for a very long time and NEVER used a VM or DMM to measure voltages on a PSU hooked into a MB becuase they are NOT reliable for doing the troubleshooting needed to resolve the issue or make repairs Chips (Especially CPUs) are sensitive to excessive noise (ripple) on the DC lines and also low power issues .

In this case the MB sensors did tell the OP exactly what the problem was- bad power
 
YeuEmMaiMai said:
BTW I was a component level tech for a very long time and NEVER used a VM or DMM to measure voltages on a PSU hooked into a MB becuase they are NOT reliable for doing the troubleshooting needed to resolve the issue or make repairs Chips (Especially CPUs) are sensitive to excessive noise (ripple) on the DC lines and also low power issues .

In this case the MB sensors did tell the OP exactly what the problem was- bad power
Click to expand...
And we still have no proof that the MB sensors were reporting accurately. And we don't know if the PSU was actually producing excessive ripple, since it was never tested with an oscilloscope (there are any number of other things that could have been wrong with it to produce a problem like this). And you still haven't explained why the motherboard sensors would somehow be able to sample ripple while a multimeter would not, since the motherboard sensors most likely do not have a higher sampling rate than a decent multimeter, and they certainly do not sample quickly enough to be able to capture ripple without aliasing.

So yes, the PSU was the problem. But that does not mean the motherboard sensors were reporting correctly, and it doesn't disprove anything I have said.
 
YeuEmMaiMai said:
In this case the MB sensors did tell the OP exactly what the problem was- bad power
Click to expand...

Don't congratulate yourself prematurely. The sensors are reporting basically the same as they did with the other PSU.

Current displayed +5V range in CPUID HM is 2.12 - 4.89 V.

+12V and +3.3C are pretty stable. -12V has gone up to -4V.
 
ryan_975 said:
1) I've never heard of that program, and I bet most people around here haven't either. When Memtest is mentioned, it's usually meant that it's the software that is downloaded and ran from a bootable environment outside of any protected-mode operating system. Only an idiot would think a program thats test memory while still inside an OS that runs in protected mode would be reliable.

2) Most DDR3 modules available for sale include heat spreaders. It's usually the "value" products that don't include them.

3) Chips known to be the most reliable in the industry can still have individual specimens that are defective. That's what I meant by "best chips", but somehow I think you already knew that.

4)You're the one that said Windows won't run outside of 5-10% specs. ATX spec for +5v, +12v, and +3.3 is +/-5%.
Click to expand...
1. Memtest: http://hcidesign.com/memtest

The only people I know who say "MemTest" refers to MemTest+ or MemTest86 also say "irregardless" and call a memory module a "chip". ;)

2. Except for Crucial modules, what brands of chips are used by most of those DDR3 modules with heat spreaders on them?

3. If we go by our definition of best chips, does that mean UTT chips (the ones more likely to be found under most heat spreaders) are usually better than chips that have their manufacturers' logos or full part numbers on them?

4. That still doesn't mean Windows has a voltage spec.
 
YeuEmMaiMai said:
BTW I was a component level tech for a very long time and NEVER used a VM or DMM to measure voltages on a PSU hooked into a MB becuase they are NOT reliable for doing the troubleshooting needed to resolve the issue or make repairs Chips (Especially CPUs) are sensitive to excessive noise (ripple) on the DC lines and also low power issues.
Click to expand...
DVM's (digital volt meter) and oscilloscopes essentially measure different things. Unless you have seriously high dollar scopes (like we do in the lab) scopes aren't really the tool to measure average DC voltages. By the same token, we have very nice Fluke bench DVM's that are hella accurate....but not for rapidly changing signals.

Our power supply shakedowns are a two step process:
- DC Average voltage: Measured across a bypass cap, at the load (which means right under the chip being supplied with the voltage), with an AC averaging DVM.
- Noise and ripple: Measured at the same test point, with a fast differential probe, less than 3mm test leads, using a fast peak detecting scope.

Bottom line is that for most troubleshooting, a DVM is a great tool. How fast do you expect the average voltage to be moving around, anyway? By the same token, if you think you are seeing all the noise and ripple with a $350.00 scope and a hand held 10X high impedance probe, I have bad news for you....:cool:

Last point - PC voltage measurement systems are usually very accurate for measuring relative changes in voltage (like, was 3.32 now 3.35) but are notoriously inaccurate when it comes to measuring absolute voltages (like, it says 3.35 volts. Is it really?). What you can do is see what the +5 reads with a decent DVM (easy to get to - use a drive connector) and see what it reads with your software monitoring tool. Remember the delta, apply it to your software readings, and you are usually going to be within a few mV.
 
BlindedByScience said:
Last point - PC voltage measurement systems are usually very accurate for measuring relative changes in voltage (like, was 3.32 now 3.35) but are notoriously inaccurate when it comes to measuring absolute voltages (like, it says 3.35 volts. Is it really?). What you can do is see what the +5 reads with a decent DVM (easy to get to - use a drive connector) and see what it reads with your software monitoring tool. Remember the delta, apply it to your software readings, and you are usually going to be within a few mV.
Click to expand...

And what about when you see the software monitor bouncing up and down 2-3 volts?

I think this is clearly a bad MB sensor.
 
kumquat said:
And what about when you see the software monitor bouncing up and down 2-3 volts?

I think this is clearly a bad MB sensor.
Click to expand...
If the voltage is flying around, it's either:

- Bad PSU
- Bad MoBo sensor

Confirm it with a DVM by measuring a drive cable (for +5) and then you'll know.
 
It would be pretty weird for an old PSU and a brand new Antec Eco to both have the exact same problem don't you think? ;)

I might toss the MM on it, but I doubt it. I'm pretty satisfied.
 
There are things that could drag the voltage around, but you're right - its unlikely (but not impossible) that two supplies in a row are bad.

A quick check with your meter will tell all. Like I said, +5 is available on a drive cable so it's a quick check. If it were me, I'd get out the meter and solve it once and for all because I worry about stuff like that......:p
 
BlindedByScience said:
Our power supply shakedowns are a two step process:
- DC Average voltage: Measured across a bypass cap, at the load (which means right under the chip being supplied with the voltage), with an AC averaging DVM.
- Noise and ripple: Measured at the same test point, with a fast differential probe, less than 3mm test leads, using a fast peak detecting scope.
Click to expand...
How does the differential probe work in that case? Is it to prevent seeing noise from other sources? All I have is a pair of single-end probes and a low frequency scope with differential input mode.
 
larrymoencurly said:
How does the differential probe work in that case? Is it to prevent seeing noise from other sources? All I have is a pair of single-end probes and a low frequency scope with differential input mode.
Click to expand...
You nailed it - the ultra short test leads and the differential probes are an attempt to measure the intended signal, not the radiated noise from other components or in our case (expansion cards) radiated noise from the host platform.

Our scopes and probes allow us to measure noise out to several gigahertz. Why? Among other things, harmonics of internal PLL's can be very high frequency noise, and knowing real noise envelope is critical when you're tuning a power plane and trying to squeeze the last bit of performance out of a product.

You can make differential measurements with your single ended probes, but you're going to be bandwidth limited due to lead length (adds series inductance) and scope bandwidth. Nevertheless, it could be a valuable technique.
 
You must log in or register to reply here.
Back
Top