Upgraded PSU and 1 GPU now overheats instantely...

[adaseb]

Very puzzled by this and searched and didn't find anyone with this issue.

Basically decided to upgrade the PSU on one of my rigs, was running fine before. GPU temps were ~65-70C or so after a few minutes. So I switched PSUs and when I powered it up I noticed that one of the GPUs (1 out of 3) instantely overheats and starts to throttle, basically when mining starts it goes to 80C and temps keep rising. So I figured the GPU must of finally died and put in another GPU, and guess what? That GPU also overheats instantely. So I put in the original GPU in a different rig and it didn't over heat. So it has to be the PSU.

However I don't see how this is possible since the other 2 gpus are not overheating. Was wondering if anyone has experined this issue before.

[1714021465]

1714021465

[1714021465]

1714021465

[Globalminingcorp]

Probably the thermalpaste has hardened overtime. When you kept running the rig there was no problem and the paste made a good connection betwen the cip and the cooling block but when you shut it down, the cards cooled off. Now the paste is not making a good connection between the chip and the cooling block. You can open the card the put new thermal paste, its not easy especially on some brands of videocards but its worth a try.

[adaseb]

Probably the thermalpaste has hardened overtime. When you kept running the rig there was no problem and the paste made a good connection betwen the cip and the cooling block but when you shut it down, the cards cooled off. Now the paste is not making a good connection between the chip and the cooling block. You can open the card the put new thermal paste, its not easy especially on some brands of videocards but its worth a try.

No its not the thermal paste because the GPU was running fine in another rig. It didn't overheat in the other rig only the one with the new PSU. Seems there is some weird issue with the PSU somewhere.

Originally I assumed it was the GPU and might of been the thermal paste but its also overheating another GPU and not overheating if I move that GPU to another rig. Seems the PSU is the culprit somehow however I don't know why.

[arielbit]

higher voltage? check with voltmeter or msi afterburner?

[JayDDee]

Make sure airflow hasn't been disrupted and fans are working properly. I know its obvious but needs to be said.
Seems pretty obvious the problem is the PSU but you can do some more troubleshooting to confirm it.

Swap slots, power connectors, both, with another GPU in the rig.

And the final obvious suggestion is to reinstsall the old PSU assuming it was working fine.

[adaseb]

higher voltage? check with voltmeter or msi afterburner?

Yes I am starting to think this might be the issue. I will disassemble and measure the PCIe voltage and see if its +12V.

Make sure airflow hasn't been disrupted and fans are working properly. I know its obvious but needs to be said.
Seems pretty obvious the problem is the PSU but you can do some more troubleshooting to confirm it.

Swap slots, power connectors, both, with another GPU in the rig.

And the final obvious suggestion is to reinstsall the old PSU assuming it was working fine.

I never removed the GPU, all I did was switch the PCIe connector to the new PSU, it wasn't moved at all.

[Metroid]

What its current voltage? To actually be overheating to a point of throttling, voltage must be very high.

[arielbit]

higher voltage? check with voltmeter or msi afterburner?

Yes I am starting to think this might be the issue. I will disassemble and measure the PCIe voltage and see if its +12V.

might wanna check the voltage too while running.

if it is a high voltage thing maybe undervolting in settings the overheating card might solve the issue.

[adaseb]

So measured the voltages its 12.25Volts without a load. Same as the other rails. Then decided to start mining and measured again at the GPU terminals and the voltage is 12.15V. So it seems the voltages are fine.

It seems that as soon as it starts to mine, right after the DAG loading, the temps jump from 45C to 80C and starts climbing, when I terminate mining, it instantely drops back down to 45C or so. Seems like there is some sensor issue on the GPU itself but only with that PSU.

Very puzzled.

[Metroid]

So measured the voltages its 12.25Volts without a load. Same as the other rails. Then decided to start mining and measured again at the GPU terminals and the voltage is 12.15V. So it seems the voltages are fine.

It seems that as soon as it starts to mine, right after the DAG loading, the temps jump from 45C to 80C and starts climbing, when I terminate mining, it instantely drops back down to 45C or so. Seems like there is some sensor issue on the GPU itself but only with that PSU.

Very puzzled.

Does it happen to other gpus on just this one?

[adaseb]

So measured the voltages its 12.25Volts without a load. Same as the other rails. Then decided to start mining and measured again at the GPU terminals and the voltage is 12.15V. So it seems the voltages are fine.

It seems that as soon as it starts to mine, right after the DAG loading, the temps jump from 45C to 80C and starts climbing, when I terminate mining, it instantely drops back down to 45C or so. Seems like there is some sensor issue on the GPU itself but only with that PSU.

Very puzzled.

Does it happen to other gpus on just this one?

Only on that GPU. I measured the motherboard voltages and they were like 3.4V and  5.1V and 12.2V.

So I decided to switch to a 3rd PSU and with that PSU the problem went away. So it seems my most expensive PSU is the culprit. Anyone had this issue before with the Lepa 1600W PSU?

Comparing the power usage, it seems with the faulty PSU the consumption was about 15Watts higher, so obviously the temps might be correct. Seems the GPU was shorting out somehow. Puzzling.

[JayDDee]

Comparing the power usage, it seems with the faulty PSU the consumption was about 15Watts higher, so obviously the temps might be correct. Seems the GPU was shorting out somehow. Puzzling.

Yes puzzling. How can the PSU cause higher GPU temperature under same load without increasing voltage?
And why only one GPU?

[dikobraz123]

Considering all things mentioned earlier, I would guesstimate that you have problem with ripple voltage on that particular rail. Try to measure ripple voltage using a true RMS AC voltmeter on each rail (AC voltmeter will measure only the ripple component of DC output) - if you measure different value for suspicious rail - you have found your problem. Probably bad capacitor on the rail output filter.

[adaseb]

Considering all things mentioned earlier, I would guesstimate that you have problem with ripple voltage on that particular rail. Try to measure ripple voltage using a true RMS AC voltmeter on each rail (AC voltmeter will measure only the ripple component of DC output) - if you measure different value for suspicious rail - you have found your problem. Probably bad capacitor on the rail output filter.

I don't have a RMS AC voltmeter but I will investigate further and see if I can replicate the same result on the other same rail with other connectors and see if switching to a different rail changes anything. I tried a different PCIe connector but it might of been from the same rail.

[philipma1957]

So measured the voltages its 12.25Volts without a load. Same as the other rails. Then decided to start mining and measured again at the GPU terminals and the voltage is 12.15V. So it seems the voltages are fine.

It seems that as soon as it starts to mine, right after the DAG loading, the temps jump from 45C to 80C and starts climbing, when I terminate mining, it instantely drops back down to 45C or so. Seems like there is some sensor issue on the GPU itself but only with that PSU.

Very puzzled.

Does it happen to other gpus on just this one?

Only on that GPU. I measured the motherboard voltages and they were like 3.4V and  5.1V and 12.2V.

So I decided to switch to a 3rd PSU and with that PSU the problem went away. So it seems my most expensive PSU is the culprit. Anyone had this issue before with the Lepa 1600W PSU?

Comparing the power usage, it seems with the faulty PSU the consumption was about 15Watts higher, so obviously the temps might be correct. Seems the GPU was shorting out somehow. Puzzling.

 separate rails bro. it is a terrible psu for mining.  and somewhere some place say 2014 or 2013 there are threads complaining about it.

https://www.newegg.com/p/1HU-0042-000D6?Description=lepa&cm_re=lepa-_-9SIAHT8BFS3218-_-Product


...

This LEPA G1600 Power Supply comes rated with 1600-Watts output, peak power up to 1700-Watts. Surpassing 80 Plus Gold level (G1600-MA is 80 Plus Gold certified. ) Full Modular Cabling Design makes system upgrade and replacement easy. High Density PCB Design makes G1600 only 18cm in depth, providing more space to dissipate the heat inside the chassis. 135mm ball bearing fan with thermal programmed speed controller delivers silent, cool operation. Honeycomb-shaped ventilation design increases airflow and keeps power supply cool. Solid State Capacitors and 105 degree C Japanese Grade A Capacitors provide great reliability.

DC-to-DC Converter Design ensures system stability. Full Bridge and Zero Voltage Switch (ZVS) topologies provide superior performance and reliability. Dual Sided PCB design synergizes PCB layout. Six powerful +12V rails with high maximum loads offer great compatibility with heavy-duty graphics cards. Ability to run up to 4-Way GPU (SLI / Crossfire) and Dual CPUs together effortlessly. Compliance with EU ErP Lot 6 standards (lower than 1W power consumption at standby mode). Over Power, Over Voltage, Under Voltage, Over Current, Over Temperature, Short-circuit protection provide great safety to your system. ATX12V version 2.3 and EPS12V version 2.92 compliant.


Features:

to Worldwide, Six Months Quality Warranty
Interface: Plug and Play travel easily, Hi-Speed USB 2.0
Support Windows98/ SE /ME / 2000 / XP / Vista/Win 7/Win 8
Support to burn CD Only, not DVD-RW
Read Speed: 24x (CD) / 8x (DVD)
Connectors:
1x

Package Quantity: 1 ...

one of the six  rails is fucking you badly.

If you can return it do so.

[arielbit]

multiple rails are bad for mining troubleshooting and maintenance, it adds another layer of "things that might be fucking with you" lol.

i had a 1500w thermaltake psu (multiple rail) back then, can't mine if 4 gpus are plugged (doesn't matter how low the power draw of my 280x's), works with 3 cards.