PSU Voltage Measurements Under Load vs Idle

[Stardance]

While a new computer has been running Prime95, CPUID Hardware Monitor reports the -12V voltage between -9.54V minimum and -9.09V maximum.

Thermaltake tech support says that "software" measurements are unreliable, and voltages should be read with a meter to verify whether they are correct.

I can use a multimeter (which I have) and the motherboard manual has a diagram with the pinouts. But I have a question:

Will being under load (such as Prime95) affect the value of the multimeter voltage readings?

Or should the voltage reading be the same value as the voltage which the line is supposed to be carrying?

I don't want to RMA the PSU unless it really is defective!

 

[hunter315]

Software reading is unreliable, and nothing really uses the -12V rail anyway. Right now HWMonitor says that my 12V rail is running at 4.29V, and that my -5V rail is running at -1.09V(the -5V rail hasnt been included in power supplies for a few years now, motherboards dont even have a contact where it would connect) so i dont trust HW Monitors readings much


Voltage will change with load levels, this is why reviewers test with a variety of load levels, the PSU will try to keep it pretty consistent but it will vary so test at idle and at full load as those will usually be your two extremes, as the current draw increases the losses in the PSU cables and inside the PSU increases so the PSU boosts the voltage a bit inside so its right when it gets down to the board.

 

[Stardance]

Thanks for your answer. I've been wondering about the -12V "rail" because the ATX connector is the only one with a pinout for it (according to the Gigabyte GA-890FXA-UD5 manual).

In another thread here, someone stated that a "tolerance" of 10% was acceptable for the -12V feed. The monitor readings are below that much, but Prime95 is running with the set of tests that are expected to maximize heat generation, which is to maximize electricity consumption.

 

[hunter315]

Honestly the -12V rail is the one i would be least concerned about, i think its still used in the PCI connector for legacy support any maybe 9 pin serial connectors for biasing, but everything else runs off of 12V, 5V or 3.3V, on most systems even if the -12V rail hit 0V nothing would notice so dont worry about its tolerance too much, focus on the +12V, 5V and 3.3V ones.

 

[Stardance]

As far as I know, the +12V, 5V, and 3.3V rails don't have a maximum that differs much from the minimum, if at all, although none of the three are exactly their nominal voltage. For example, the 3.3V feed is reported to be 3.32 as current, minimum, and maximum. The +12V and 5V feeds deviate a little, too.

I suppose that I will check them with the multimeter when I examine the -12V feed, too. Maybe the ASUS EAH5670 video display card uses the -12V line, or is that one old enough to be "legacy" yet? 🙂

 

[ko888]

Since there is a COM/serial port header on that motherboard that is where the -12V would feed to. If you don't use that port then you don't need to worry about it. This header isn't used unless you specifically ordered the optional COM port cable.

Graphics cards don't use -12V.

Generally, supply voltages must be within ±5% of their nominal values at all times. The little-used negative supply voltages, however, have a ±10% tolerance. The supply voltages and their valid ranges are as follows:

+5 VDC : +4.75 V to +5.25 V
−5 VDC : –4.50 V to –5.50 V
+12 VDC : +11.40 V to +12.60 V
−12 VDC : –10.8 V to –13.2 V
+3.3 VDC : +3.135 V to +3.465 V
+5 VSB : +4.75 V to +5.25 V

 

[RetiredChief]

Software Voltage measurements are reliable as long as you can verify that they are reading the correct bits!! I normally verify the +5 and +12 on a one time basis using the molex connector (Red = +5 and yellow or orange = +12V Two blk wires are rtns). The problem is that it is unknown what the software does when it reads an invalid Bit. Most software revertss to displaying a rail when it incounters an invalid value which is way outside normal values. Example displaying 3 Volt for the +5 or 3 -> 9 volts for a +12 Volt - The computer will NOT boot, so they are known invaild readings.

As pointed out the -5 and -12 although still provided are only used in OLD addon cards/serial port components and the software may not even provide a reading, just a default garbage value.

Yes you can measure this on the 20/24 Pin connector, and since nothing is really using it, it will be an "idle" value. Hard to get at, I use a straigth pin to measure @ the 20/24 pin connector with the Neg meter lead connected to the black Molex pin, SO unless you need it, not worth it.

Side comment on the tolerence of the -5, although the spec is 10%, if infact it is used (required) and goes to any TTL lodgic the tolerence should be the same as +5V. The is the min/max voltages that can be used with TTL logic chips.

Editted:
In my case HWMonitor showed no voltage for +12 But what look like good values for the -12V - Discovered that the +12 V was really showing up as a neg value, but the value was correct - they still have not correct this for my MB. But the reading is accurate and reliable, I just have to ignore the (-) in front of the value and know that it is for the +12V rail.

 

[Onus]

What PSU do you have?

 

[hunter315]

The -12V still exists in PCI slots, but does not exist in PCI-E slots, so unless you have a PCI card in the very bottom slot on your board absolutely nothing is using it, and even if you do, odds are its still not using it.

I forgot this at the beginning, easiest way to know that your readings arent venturing out of spec, if your system is stable and not giving you any issues then its fine, dont go looking for problems with no evidence because you may find something that isnt normal but also doesnt cause any problem so there is no reason to worry about it.

 

[Stardance]

Since there is a COM/serial port header on that motherboard that is where the -12V would feed to. If you don't use that port then you don't need to worry about it. This header isn't used unless you specifically ordered the optional COM port cable.

Graphics cards don't use -12V.

Generally, supply voltages must be within ±5% of their nominal values at all times. The little-used negative supply voltages, however, have a ±10% tolerance. The supply voltages and their valid ranges are as follows:

+5 VDC : +4.75 V to +5.25 V
−5 VDC : –4.50 V to –5.50 V
+12 VDC : +11.40 V to +12.60 V
−12 VDC : –10.8 V to –13.2 V
+3.3 VDC : +3.135 V to +3.465 V
+5 VSB : +4.75 V to +5.25 V

Thank you for your answer. RetiredChief is correct about the tolerance for -5V being 5% instead of 10%, certainly if those feeds are used in any hardware to power a TTL logic chip(s). And I believe that Hunter315 is correct about the use of -12V for some PCI cards and adapters (including COM ports). There is probably a diagram of PCI pin assignments and voltages somewhere on Wikipedia. It's been years since I looked at any expansion slot specs. 🙂

For what it is worth: at CPU idle, according to the Radio Shack multimeter readings on the 24-pin ATX molex, -12V was -12.01V and steady as a rock, while CPUID Hardware Monitor's readings from the ITE IT 87 sensors reported it as -9.68 V and fluctuating by 0.5 volt or more. Similarly, the multimeter readings for +12V, +5V and +3.3V were at their respective nominal values, give or take less than about 0.25 V at most, and were not changing as often or by as much as the Monitor's respective displayed values.

Putting the PSU "under load" by running the GIMPS Prime95 Torture Test did not make much difference in the multimeter readings during the time that I observed them. They tended to be less steady, but the fluctuations were less than about 0.2 V for the +12V line and less than about 0.1 V for the +5V and +3.3 V lines, respectively. The -12V line remained at -12.01V on the multimeter while the ITE IT 87 chip and/or CPUID Hardware Monitor had the value rapidly increasing to about -6.9V.

So I've concluded that I don't have any clear evidence that the PSU is defective or malfunctioning. If it ever did malfunction, then it did not affect the results of any of the numerous difficult mathematical calculations performed by the Prime95 Torture Test during the 73+ hours that the program ran. As far as I can determine, the machine is good to go. 🙂

 

[Stardance]

What PSU do you have?

The PSU is a Thermaltake TR-450W.

Components were chosen that minimize electrical consumption (thus, waste heat), and that work under relatively high temperatures. It remains to be seen how well I met those twin goals, but June will be here eventually. 🙂

 

[delluser1]

the numerous difficult mathematical calculations performed by the Prime95 Torture Test during the 73+ hours that the program ran.

Consecutively ?

 

[Stardance]

Consecutively ?

Yes, 73+ hours consecutively (of course). Some people run Prime95 all day for seven consecutive days before they decide that the computer is trustworthy.

There were a couple of false starts. Prime95 would run for about three hours, then the entire computer would shut down. However, it did not seem to be a "crash". Neither Prime95 nor Windows 7 recorded any error messages. After thinking about it for a while I realized that the default Power Management plan was the cause. Evidently, if someone is not using the mouse and/or pressing a key(s) on the keyboard from time-to-time, then Windows 7 concludes that "no one is using the computer". So it would close all running applications and put the computer into a "sleep" mode. I had to change that criteria to "never" so that Prime95 could run 24 hours per day until I decided to stop the test.

 

[delluser1]

Yes, 73+ hours consecutively (of course). Some people run Prime95 all day for seven consecutive days before they decide that the computer is trustworthy.

Of course
:lol: :lol: :lol:

 

[hunter315]

Thats a good way to hurt a system...


For a mechanical analogy, go buy yourself a new car and run it right at the red line for 73 hours and see how much it likes it.

 

[Stardance]

Thats a good way to hurt a system...


For a mechanical analogy, go buy yourself a new car and run it right at the red line for 73 hours and see how much it likes it.

Your analogy is false. A car is essentially a system of mechanical devices, not a system of electronic devices such as a computer. The only "mechanical" components in my computer have electric motors: the hard disk drive, the optical disk drive, and the fans. The Prime95 Torture Test, as such, doesn't "test" any of them.

How will using a new computer continuously for 72 hours damage it? The traditional 72-hour "burn in" period was adopted from experience and statistics. For most electronic devices, if a device is going to fail, then it will fail during the first 72 hours that it is in use, else it is unlikely to fail until it approaches the end of its expected service life. (Failure during the first 72 hours is sometimes informally called "infant mortality" <a grim joke>.)

If there wasn't something wrong with it, then it wouldn't fail. The "Prime95 Torture Test" just uses the computer to do exactly what it is supposed to do -- compute!! For most "personal computers", the software that is running and the hardware that it uses spend almost all of their operating time just waiting for something to happen. Ordinarily, they're waiting for the user to press a keyboard key(s), to move the mouse, to click a mouse button, or to speak into a microphone. That is, they're waiting for input, for data to process.

So I believe that it is reasonable to run the Prime95 Torture Test for the burn-in period, and many overclockers use it to test the stability of their configuration of the CPU, memory controller(s), and system memory. Note that the Torture Test does not test video display cards, hard disk drives, optical disc drives, network interface devices, or any other system components.

The Great Internet Mersenne Prime Search (GIMPS) uses the ordinary computers of volunteers. The Torture Test was originally developed to identify computers which are likely to perform the mathematical calculations and report that a number is a Mersenne prime, but that result will not be confirmed by performing the same calculations for the same number with one or more other computers.

If you want the details, then visit the GIMPS website, download the software, extract the files and read the ASCII text files about Options and about Overclocking, respectively:

http://www.mersenne.org/freesoft/#newusers

You aren't required to "join" GIMPS in order to download the software package, only if you want to join the search for Mersenne prime numbers. 🙂

 

[delluser1]

I just want to let you in on a little secret;
Most of us overclock and use Prime for stability testing.
That last post sounds like it came from someone either marketing for Prime or very new to the game.

 

[Stardance]

I began using Prime95 for 72-hour burn-ins not long after the GIMPS started. I don't overclock and don't have the time to learn whether I had the inclination. My remarks were posted in reply to the ones which I quoted from hunter315, and were not meant to be insulting to anyone.

That said, the discussion in this thread has become rather far off-topic and I would just as soon end it on this note.