Question PSU Sparked and started smelling, MCB tripped

[SteelMouse]

Hello guys,

PC Specs:
CPU: Intel i7 5820k
RAM: G Skill Ram 32 gig (Four 8 gig sticks)
MOBO: Asus X99-E WS
GPU: Old : MSI GTX 970
GPU: New: MSI RTX 3080 Gaming X Trio 10G
PSU: Corsair HX 1000i - Entire build date: March 2016, so PSU is around 4 years 7 months old.
HDD: 2 x WD 2 TB, WD Black
SSD: 1 x Samsung 850 EVO + 1 x SanDisk 240gb SSD
CPU Cooler: Liquid cooler: Corsair H100i - has 240mm radiator with two fans in front and two in back as push and pull setup.
Both the front two and back two fans are all Noctua 120mm iPPC 3000 PWM
Fans: Exhaust: 1 x Noctua 140mm iPPc 3000 PWM
2 x Noctua 120mm iPPC 3000 PWM
Intake: Liquid cooler sits in the front of the case, so acts as intake as well
Side: 1 x Noctua 140mm iPPC 3000 PWM
Bottom: 1 x Noctua 120mm iPPC 3000 PWM
Fan Curve: All exhaust + Intake are roughly running around 2000 RPM.
GPU Fan Control: With the new MSI RTX 3080 I had an aggressive fan curve at 60% at 20 deg celcius, 70% at 30 deg celcius,
80% at 40 deg celcius and 90% at 45 deg celcius so on 100% at 50 deg celcius
Peripherals: 1 x Corsair K95 Platinum Mechanical Keyboard
1 x Steel Series Rival 710 Mouse
1 x Sennheiser Game One Headset
1 x Sennheiser normal cheap headset

I installed a new MSI RTX 3080 Gaming X Trio 10G on my PC. Since this card has three 8 pin connectors, I used the existing pcie connection (original, came with the PSU HX 1000i and was attached to the old MSI GTX 970 GPU)
and added two more pcie cables (original cables, I had them stored in the packing box of the HX 1000i). So all in all, the new MSI RTX 3080 had three individual PCIE cables going from 3 PCIE ports in PSU to the graphics card.

Event 1:
After installation, I powered up the PC, everything went ok, I installed the latest Nvidia Drivers, then set up the MSI Afterburner with the above mentioned fan curve for the GPU.
Then I ran Fur Mark test at 1920 x 1080 preset, GPU went from like 30 deg celcius start to 50 deg celcius by the time Fur Mark finished and I got a nice 17,111 score in the Fur Mark at the 1920 x 1080 preset.
I then started a game (World of Tanks, HD client with HD textures, Ultra setting, everything maxed out) and also had Discord streaming of the World of Tanks client.
Game loads up fine and I start a match, once match starts, my computer shuts down, my UPS (APC 1500 VA) makes a loud beep beep beep beep and entire power is off (lights, pc, everything is off).
I switch off the PC's wall socket switch and go check the UPS. The UPS has the red fault light displayed. I check the UPS MCB. It's tripped. I put it back and then UPS comes back online and all the lights are back up.

Event 2:
Current Situation: Wall Socket: OFF + PSU Switch : ON
I turn on the wall socket and the UPS again makes the loud beep beep beep and red fault light, everything is off (all lights are off) + MCB on the UPS is tripped again.
I switched off the wall socket of the PC + also switched off the PSU Switch.
I put the MCB back and the lights are back up.

Event 3:
Current Situation: Wall Socket: OFF+ PSU Switch: OFF
I turn on the wall socket, everything is fine up to now.
I turn on the PSU Switch, now there's a spark in the PSU, everything is dark (lights went off) and I could clearly see the spark in the PSU in the dark room.
And there is burn smell coming from the PC.
Then I disconnected the wall socket and also the wire going to the PSU, PSU switch is OFF.

Analysis:
1. Everything was fine with the PC even with Fur Mark testing, and maybe all the Ultra settings caused the GPU to draw quite a bit of power.
But MSI says recommended PSU is 750W (https://www.msi.com/Graphics-Card/GeForce-RTX-3080-GAMING-X-TRIO-10G/Specification)
and mine is 1000W PSU. I should have been ok with the PSU being able to handle the max needed power of the entire PC loadout and all of it's components.

2. The MCB in the UPS tripped in all three events. During event 2, I didn't even switch on the computer, barely switching on the wall socket (with PSU switch being ON) and power going to the PSU caused the MCB to trip..
Third event was also similar, this time, switch the wall socket on, then the PSU Swtich ON caused the spark and the MCB on the UPS tripped.

Questions:
1. Does the power go to all components without even starting the system?
I have seen my motherboard lights, ethernet lights go up without powering on, maybe some power goes to motherboard, but does power go to other components, GPU / HDD, etc,.?

2. Assuming power goes to all components in the system (without powering on the system, just letting power go to PSU), the entire system would maybe draw 300 W at startup (no idea, a guess
based on the normal power draw after starting up is around 240 W power in and slightly lower power out). Maybe in event 1, all the Ultra settings caused the power draw to be high, but I can't
understand why the MCB be tripped without even the computer being switched on.

3. Would there be a possibility of the PSU sparking damaging my other components: motherboard / CPU / RAM / GPU / HDD / SSD?
I hope nothing is fried, especially the GPU, it's the GPU's first 1 hour in the PC, I would be very happy if the GPU lived.
Does power go to all these components from the PSU even without turning on the system?

4. Past event: Maybe useful for troubleshooting: In the past with the same PC, I had the liquid cooler mounted on the top as exhaust and it leaked down into the GPU.
Then the comp shut down in the middle of a match, then again after being turned on for 5 mins, then the third time didn't even start.
The liquid cooler leaked into the GPU's voltage controller ( I think it's the voltage controller, black colored square looking thing) and the GPU went out.
I replaced the CPU + MOBO after that (the Asus X-99 E WS is the replacement new board).
I also took the PSU to the official corsair service center and they checked and said everything is fine. The had a test rig, hooked it up and said it's fine.
Maybe this past event weakened / semi damaged the PSU and now when the power draw went higher than the usual draw, the PSU died.

5. MSI states they have extra fuses on the custome PCB of MSI RTX 3080 to prevent electrical damage, maybe hopefully that saved the card.
Asus X99-E WS is a workstation board, maybe has decent overvoltage / short circuit protection and didn't get damaged.

6. Any advice and suggestions on how should I go about troubleshooting the PC would be greatly appreciated.

7. Considering the HX 1000i Corsair PSU has 7 year warranty, I can get a new one, but I am thinking of getting a massive Corsair AX 1200i or Corsair AX 1600i.
Kindly suggest some very very good, top end PSU (budget's not a problem, I don't want to ever fry my components, so would like to get the best of the best PSU available).

Thank you for your time, and I really really hope the new GPU is fine, if that thing is gone, that would be a very nasty painful day.

 

[Ralston18]

More information needed.

As I understand your post, you are using a wall switch to turn on/off the computer?

If that is correct, why are you not using the PC's power switch to turn the computer on or off?

Does that wall switch also provide power to other system components: UPS, monitors, router, printers, etc..?

What are the electrical connections/paths?

For example:

Wall switch ----> Wall outlet -----> UPS ----> Surge Protector/Power strip? then ----> power cables to PC, monitor, printer, etc..

Are there any other appliances or devices using the same electrical circuit that the wall outlet is on?

Edit and correct my line diagram as necessary.

Need to gain an understanding of the "big picture": there may be multiple problems.

 

[Grobe]

I can somehow try to answer to #1.

It totally depend of what specific component that are broken. There is usually a transistor (mosfet, bjt or similar) that handles high voltage and current spikes (at the main part). If that goes broken, it typically act like a dead short - like the situation you describe.
In such an situation - there will be a current spike into the main transformer into the PCU, current rush until the core saturates wich again causes the mcb to trip, but also causes short voltage spike on secondary side (i.e. you'll observe some led lit for a short time).

There should be overvoltage protection built in to the PSU - but if the break of the MCB produces a very powerfull voltage spike, you do put the motherboard and other components in risk.

 

[jonnyguru]

The UPS can't handle the inrush current of a 1000W PSU trying to charge the bulk caps.

Solution: Stop turning the PC off with the wall switch. This doesn't make sense.

Also, getting a more massive PSU doesn't solve this problem. It'll only make it worse because it has even bigger bulk caps to charge.

 

[SteelMouse]

Hello guys,

I am not using the wall switch to turn off the PC. I did that during event 1 as a measure of caution (Everything went off all lights + PC, UPS started making beep beep beep, then I rushed to wall socket and turned it off). I normally use the shut down command, computer shuts down, then I switch off the wall switch before I go to bed.

Normal Switch Operation:
UPS ---> Wall Socket (Switched off during bedtime and switch ON in the morning) ---> Surge protector Switch (Usually left ON) ---> PSU Switch (Usually left as ON all the time) ---> Computer Power On button on the NZXT Case (Used to switch ON / OFF the PC in addition to shutdown commands).
Once PC is shut down, I switch off the wall switch before going to bed, that's the usual routine. And the 1500W Surge protector and PSU Switch are left ON all the time.

With this info, maybe kindly review Events 1 to 3.

Here's the electrical connection path:
Mains ---> Divides into Light and Heavy line ---> Light line has the Siemens RCCB + UPS ---> Wall switch --> Wall Outlet ---> 1500W Surge Protector ---> Surge protector has only 2 appliances (PC PSU Power Input + LG 24GM77 Monitor Power Input)

Better diagram:

https://imgur.com/a/1Nz0frU

View: https://imgur.com/a/1Nz0frU


In this add RCCB to the diagram: After the mains divide into Light + Heavy, there's two RCCB's one for Light and one for Heavy. I usually have seen the RCCB's trip but never the UPS's MCB.

Here's my monitor:
https://www.lg.com/uk/monitors/lg-24GM77-B
Looked and it seems to have 27 Watt power consumption.

My APC UPS Model: BI1500SINE-IN
https://download.schneider-electric...82G5CH_R4_EN.pdf&p_Doc_Ref=SPD_GLIN-82G5CH_EN

APC: BI1500SINE-IN:
Power rating on Mains: 1500 VA at 230 Vac

From the above document, this is relevant to all the 3 Events that happened yesterday:
I had the Fault light on constant Red + MCB on the back tripped all three times + there was beep beep beep going on.
" 2 : FAULT : If illuminated, the unit is overloaded, has a shorted output, has overheated, or there is an internal electrical fault. Contact APC Technical Support. "
"
"8: Miniature Circuit Breaker : Miniature Circuit Breaker (MCB) in BI1500SINE-IN will trip if there is an overload while the unit is operating on mains power. Manually reset the MCB.Note: If the unit is overloaded, the circuit breaker will trip. There is no indication of this on the front panel of the unit. "

"Troubleshooting: The Home UPS is beeping continuously : Possible Cause: his can be caused by low battery voltage, failed or dead battery, an output short circuit, over temperature : Corrective Action: Recharge battery. If this does not solve the problem, contact APC Customer Service "

Additional Info: I don't think the batteries were drained yesterday, yesterday there was normal power supply without any outages.
Additional Info: I have never seen the APC UPS do this in my 4 years of using it.

Kindly let me know if you need info, thanks a lot for your insights.

 

[SteelMouse]

Addional info: The UPS powers entire Light line of the home. The heavy line has all the heavy stuff - fridge / water heater.
In addition to the PC + Monitor being the load on the UPS, there's also atleast 2-3 lights, one fan + one outlet that has an 5 socket extension cord: From extension cord, power goes to 2 Wifi Routers + 2 FTTH Input routers, total 4 small devices on the 5 socket extension cord.

 

[SteelMouse]

Additional Info:
View: https://www.reddit.com/r/nvidia/comments/ixklsn/msi_geforce_rtx_3080_ventus_3x_oc_problems/g89r8if/

From one of the comment's in the above thread:
"even with the 16 phase power design, the power stage capability is much larger than the GPU can request on average.
16 stages x 50Amps / stage = 800Amps total. and that @ ~1V VCore voltage, you have a 800W capable power delivery.
The problem comes with instantaneous high clock boosts that can occur and may cause power draw variation from ~100W to ~350W witch can cause slight voltage drops on the supply rail -> crash.
The main advantage of a design with more power stages is the better immunity to this transients."

I think that's what happened to me as well. MSI RTX 3080 was fine during driver installation, FUR Mark test at full HD preset (1920 x 1080), got a nice 17,111 score in it. Then the game (world of tanks), even launched fine, once match started, then boom, my UPS's MCB tripped, UPS goes to Fault red light, and UPS is all beep beep beep continuous beep.

I think that transient power draw nailed something in the PSU, so in the events 2 and 3 (which were attempts to turn on the comp), the PSU gave up in event 3. Looks like if this is the case, then going to get everything fixed and then underclock the whole card to stable safe levels.

Reason I think this might be the case is, the PSU Corsair HX 1000i was rock solid for all these 4 years and handled the MSI GTX 970 just fine. Then this happens with the new card loading up a single game.

Well if that's the case, being on the bleeding edge has a price to pay, sometimes dice doesn't roll in your favor.

 

[jonnyguru]

Reason I think this might be the case is, the PSU Corsair HX 1000i was rock solid for all these 4 years and handled the MSI GTX 970 just fine. Then this happens with the new card loading up a single game.

Well if that's the case, being on the bleeding edge has a price to pay, sometimes dice doesn't roll in your favor.

Corsair HX1000i has been tested with 3080 and even 3090 cards and there are no issues.

 

[SteelMouse]

Corsair HX1000i has been tested with 3080 and even 3090 cards and there are no issues.

Tested under normal operations / normal games normal settings / ultra settings? I guess they did test out all the settings.

Only reason I am kind of leaning on the transient power doing something to the PSU is there have been problems with this RTX launch and boost clocks and crashes and add the fact that this happened exactly the instant the match started. Had it happened anytime else during normal usage, etc,. then maybe some other things went wrong. Happening exactly the moment the GPU is loaded up is strange (anyway just a line of thought I had) and not trying to be assertive about. Or maybe I have a PSU that's on it's way out and this card sent it out faster. Will be troubleshooting in detail over the next few days, component by component and we will have a better picture.

Thank you for your time and that's a very impressive list of work history you got there @jonnyguru . Didn't know I was interacting with a legend. Thanks again for your time.

 

[jonnyguru]

Tested under normal operations / normal games normal settings / ultra settings? I guess they did test out all the settings.

Tested every scenario. Prior to launch, I sent a truck load of PSUs to Nvidia for testing. I also know what PSUs did NOT pass. ;-)