Safety Question When Using Chinese PSU Tester

B

Boris_yo

Limp Gawd
Joined
Oct 22, 2011
Messages
224
Hello,

I have semi-modular Corsair power supply and this is the tester I have bought:

-1109306013-866513075.jpg

Right now only 24-pin power cable is disconnected from motherboard. All the other power cables, except floppy are connected.

I want to test 24-pin cable but am concerned whether I should disconnect all the other power cables from motherboard.

I thought when PSU is switched on, all the other power cables are powered on as well and it's better to disconnect them from motherboard as well.

Thanks
 
I would definitely disconnect the PSU entirely. Powering only some parts of a complete system could have unexpected effects. Although it shouldn't cause damage, I wouldn't take the risk without good reason.
 
yeah it wont cause problems with components but might misread due to draw on the psu. why you testing it?
 
Those chinsey "PSU Testers" are a joke. They'll only tell you what the voltage of the PSU is in an unloaded state, nothing more, and that's nowhere near a PSU test.

They don't do load testing or ripple testing, they're effectively useless. Money would be better spent on a good multimeter.

Also, don't force a PSU on with a paperclip when connected to components that don't want to normally power up, you're going to fry something if you already haven't done so.

I would suspect either your motherboard or GPU are dead. Some component probably shorted out, which is why the system won't power on.
 
GiGaBiTe said:
would suspect either your motherboard or GPU are dead. Some component probably shorted out, which is why the system won't power on.
Click to expand...
About a year ago when it happened first time, I tested motherboard with CPU, GPU, RAM outside of chassis. PSU worked fine, with fan working and stable voltage readings that I tested with Chinese multimeter but I could not power on the system. Left it for a few days and just before I placed an order for a new PSU everything powered on to my surprise.

Assuming PSU is fine, I don't see how GPU shorts out the system. If I send motherboard to retailer's testing lab the system "decides" to power on like it was first time and I will be out of time and shipping fees.
 
Did you ever confirm that the 24pin connector was making good contact and didn't have a loose or broken connector?
 
Boris_yo said:
About a year ago when it happened first time, I tested motherboard with CPU, GPU, RAM outside of chassis. PSU worked fine, with fan working and stable voltage readings that I tested with Chinese multimeter but I could not power on the system. Left it for a few days and just before I placed an order for a new PSU everything powered on to my surprise.
Click to expand...

You either have an intermittent short, loose connection, or some component somewhere has a failed solder joint. You could try putting pressure on the board in various spots with your finger and see if the board turns on. Don't go "hulk smash" on it, just enough to barely flex the board.

Boris_yo said:
Assuming PSU is fine, I don't see how GPU shorts out the system.
Click to expand...

Easy. If one of the mosfets shorts out, it will prevent the system from powering up. The GPU itself can also short out, as can the memory chips.
 
hititnquitit said:
Did you ever confirm that the 24pin connector was making good contact and didn't have a loose or broken connector?
Click to expand...

I tried playing with it, plugging it out and in but system didn't turn on. It might be shortened out. If it is, I won't be able to confirm it. Will check today with PSU tester and see if voltage instabilities happen when I wiggle 24-pin cable.

GiGaBiTe said:
You could try putting pressure on the board in various spots with your finger and see if the board turns on. Don't go "hulk smash" on it, just enough to barely flex the board.
Click to expand...

Should I do that with everything outside of chassis? What if chassis shortens out system? Would be strange since it ran for many months without issue, working 3 hours per week on average though.

GiGaBiTe said:
If one of the mosfets shorts out, it will prevent the system from powering up. The GPU itself can also short out, as can the memory chips.
Click to expand...
Do you know how long does motherboard take to start working after short? If I recall correctly I tested with CPU and 1 memory stick in different slots last year and despite that I could not power on system.
 
Boris_yo said:
Should I do that with everything outside of chassis? What if chassis shortens out system? Would be strange since it ran for many months without issue, working 3 hours per week on average though.
Click to expand...

Easier to do it in the case since the motherboard would be suspended on standoffs. Just remember not to go hulk smash on it. But even assuming you find it to be a connection issue, there's not much you can do about it.

Boris_yo said:
Do you know how long does motherboard take to start working after short? If I recall correctly I tested with CPU and 1 memory stick in different slots last year and despite that I could not power on system.
Click to expand...

Short circuit protection is done by the power supply, not the motherboard. PSU short circuit protection reset methods vary. Sometimes they'll reset on their own after x amount of time, sometimes you have to remove power completely by unplugging them from the wall. But if the short is still there after the PSU resets, it's just going to trip again.

I have seen marginal components have weird shorts that were temperature based, but it's not that common.
 
GiGaBiTe said:
Easier to do it in the case since the motherboard would be suspended on standoffs.
Click to expand...
Any chance standoffs and screws short out motherboard? Maybe there are some kind of non-conductive washers available online to buy?
 
Boris_yo said:
Any chance standoffs and screws short out motherboard? Maybe there are some kind of non-conductive washers available online to buy?
Click to expand...

Standoff mount holes are supposed to be ground, there's no reason to isolate them unless there's a fault with the board. In my entire 25+ years of system building, I've only ever come across one motherboard that had something shorted to a standoff pad that required the use of insulating washers.
 
Today after 3 days or so not using my PC because it didn't turn on I turned it on and it works. I decided to try wiggling 24-pin cable at the spot where multiple cords come out of sleeve, before connector. On 27th second of the video you can see PC restarts:



I also tried wiggling cable after connecting it to chinese PSU tester but all readings were normal.

20210903_144849.jpg

Is it something in that spot that causes power interruption? Maybe it can't detect power interruption because it is very brief? Maybe this problem happens only when there is a load applied on PSU?

If only my PSU was fully modular I would just buy a different cable but it's not and I can't borrow PSU from anyone. Maybe my chinese multimeter is sensitive enough to detect brief power interruption? I could test pin-by-pin but doing that while wiggling cable will require 3rd hand, which I don't have. I am not a Goro...
 
If it were me, I'd cut the rubberry sleve off to see what the cables look like under it. If there's a janky splice or a nicked cable that gets less well connected when you wiggle it, that could be a big increase in resistance which would lower delivered voltage --- but only with a load applied. If the PSU tester doesn't have a meaningful load, the voltage will show as normal even with a very high resitance/very thin connection.

Of course, you'll never get that sleve back on (or at least not without a ton of work). It could also be wiggling moves the pin in the connector that increases resistance if it's too loose.

A bad connection like I'm thinking you have would also tend to get worse by itself under load. High resistance makes high heat, makes things loosen makes high resistance, until it makes no connection, then it cools down (sometimes reconnecting).
 
Boris_yo said:
GiGaBiTe Hello. What do you think? Can 1 out of 24 pins have intermittent power failure?
Click to expand...

There can be many failures that would cause that. The socket could have problems, the socket pins could have failed solder joints, or the cable could have loose connections to the pigtail to the PSU, or the pins and sockets are corroded/oxidized.

Many of the power pins in the Minifit Jr. plug are redundant, so losing a pin won't necessarily cause a shutdown. You'd have to lose one of the singular wires for a fault to happen, like +12v, -12v or the power good signal. If you lost a ground or one of the +5v or +3.3v wires, nothing would necessarily happen because they're all paralleled to each other. Losing the +12v may technically not cause a shutdown depending on how the board is wired internally. Many boards have the +12v from the Minifit Jr. connector connected to the +12v from the P4/EPS12v connector in some fashion, and it could start pulling power from there.
 
GiGaBiTe and toast0 Hello. Do you know of a bootable USB flash drive for testing system functionality and especially running stress tests?

Like this one but open source: https://www.passmark.com/products/bootable-usb-flash-drive/index.php

I want to run stress tests for PSU in a pre-boot environment with storage devices disconnected to prevent potential issues caused by restarts. I wonder if I should also disconnect HDD and eject m.2 SSD from motherboard to be safe.

Thanks
 
Just make a Linux live USB stick and install Prime95 and some Uniengine benchmarks.

You'll want to make sure your cooling systems are working properly, Prime95 especially can cause potential hardware failure.
 
GiGaBiTe said:
Just make a Linux live USB stick and install Prime95 and some Uniengine benchmarks.
Click to expand...
Downloaded PassMark's BurnInTest evaluation version and created bootable USB. Ran a few times for 15 minutes each, due to evaluation version's limiation. Ran also once in Windows 10. All is fine, provided 15 minutes is enough to test PSU's stability. If it's not, maybe I should try Prime95 with Uniengine?
 
Boris_yo said:
Downloaded PassMark's BurnInTest evaluation version and created bootable USB. Ran a few times for 15 minutes each, due to evaluation version's limiation. Ran also once in Windows 10. All is fine, provided 15 minutes is enough to test PSU's stability. If it's not, maybe I should try Prime95 with Uniengine?
Click to expand...
As long as your running realistic loads, prime and uni should give you a solid result.
 
GiGaBiTe hititnquitit Someone told me that he uses multimeter on molex connector to check PSU under load. Told me that molex and 24 pin cable share same line so if there are no voltage instabilities but I experience issues it could indicate physical issue in 24 pin cable. At least it could help me isolate issue outside of PSU.

What do you think?
 
Boris_yo said:
GiGaBiTe hititnquitit Someone told me that he uses multimeter on molex connector to check PSU under load. Told me that molex and 24 pin cable share same line so if there are no voltage instabilities but I experience issues it could indicate physical issue in 24 pin cable. At least it could help me isolate issue outside of PSU.

What do you think?
Click to expand...

Modern PSUs often have multiple 12v rails, so it's not always certain that the 12v rail on the Minifit Jr. connector is the same as on the AMP, PCIe or EPS12v connector. You'd have to open the unit to determine the topography of the power rails.
 
GiGaBiTe toast0 Hello. Using Kubuntu live bootable USB. Thought I will be able to monitor CPU and GPU temperatures while stress testing but it turned out to be a headache. Tried MangoHud but it requires dependencies to be installed. I install one and then it turns out I am still missing another. Installing another and missing another. Some don't work etc...
 
Boris_yo said:
GiGaBiTe toast0 Hello. Using Kubuntu live bootable USB. Thought I will be able to monitor CPU and GPU temperatures while stress testing but it turned out to be a headache. Tried MangoHud but it requires dependencies to be installed. I install one and then it turns out I am still missing another. Installing another and missing another. Some don't work etc...
Click to expand...

Why are you using such a complicated application for sensor monitoring? Use lm_sensors, you can type "sensors" in a bash terminal and get the temps of everything easily.

If you need a graphical thing, use the XFCE panel sensors plugin.
 
GiGaBiTe said:
If you need a graphical thing, use the XFCE panel sensors plugin.
Click to expand...
Thanks.

Guess it's a plugin that shows stats on top using overlay of some kind?
lm_sensors worked except CPU fan and GPU fan were not detected.
GPU temp was varying between 29 C and 30 C on Unigine Heaven benchmark and Mprime.
Maybe it was not a GPU sensor and belonged to another component because it seems strange
that GPU will stay on that low temp when system is under heavy load...
 
Boris_yo said:
Guess it's a plugin that shows stats on top using overlay of some kind?
Click to expand...

It will show in the XFCE panel. You can orient the panel on the top or bottom of the screen, or have a floating panel if you want.

Boris_yo said:
lm_sensors worked except CPU fan and GPU fan were not detected.
GPU temp was varying between 29 C and 30 C on Unigine Heaven benchmark and Mprime.
Maybe it was not a GPU sensor and belonged to another component because it seems strange
that GPU will stay on that low temp when system is under heavy load...
Click to expand...

You're generally not going to get fan speeds because it requires reading whatever hardware monitoring chip is on the board, and there hundreds, if not thousands of different parts from different manufacturers. lm-sensors requires a driver for these, and in many cases, they're not available because the drivers aren't open source. many of these monitoring chips were reverse engineered by the Linux community, which is why you can see anything at all, like temperatures.

The weird readings you're seeing could be lm-sensors reading a spurious value from an invalid location. You can install the proprietary Nvidia or AMD drivers though and get the temp and fan speeds there though.
 
GiGaBiTe said:
The weird readings you're seeing could be lm-sensors reading a spurious value from an invalid location. You can install the proprietary Nvidia or AMD drivers though and get the temp and fan speeds there though.
Click to expand...

Installed 460 tested driver and also installed nVidia CUDA toolkit I guess, tried nvidia -smi terminal command to get temperature readings but there was error etc. It's not that my default OS is Kubuntu that I need to delve into all that stuff so I think I am done testing.

I didn't leave mprime stress test and Unigine benchmark running for hours though. However running them for 10 minutes didn't show any abnormalities with power supply and motherboard. My only suspicion is in power cable I used with PSU which could have led to original issue.

Thanks for your help comrades.
 
Boris_yo said:
Installed 460 tested driver and also installed nVidia CUDA toolkit I guess, tried nvidia -smi terminal command to get temperature readings but there was error etc.
Click to expand...

Use the control panel and look at the thermal settings.

1633757414584.png
 
You must log in or register to reply here.
Back
Top