[SOLVED] PSU issue? UPS issue?

[KuniD]

Hi all, would appreciate your input on the following problem.

 

I have been having hardware stability issues for a while and am struggling to pin down the cause. The issues usually materialise as:

 

1) Random reboots

2) Random reboots, that then get stuck in a constant cycle of trying to power on but failing to do so besides a brief flicker of fans/lights. I need to either (a) trip my UPS to battery mode (turning off then on wall switch) or (b) kill the switch on the PSU for a minute or so then turn back on to stop this.

3) Reboot triggered by loading a game in Windows bare metal (i.e. Star Wars Squadrons)

4) In some Windows 10 VM's I have noticed screen artefacts

 

On the last point this could be a software issue as it only happens in Windows 10 VM's, not bare metal or MacOS / PopOS VMs. 

 

On point 3 I have removed my GTX 970, but still experienced random reboots running on just a GT 710.

 

I have also left Memtest running in the past without issue.

 

I thought I'd also test how my server behaves when I turn the wall socket off for my UPS. Even though doing so leads to my UPS going into battery mode, it still triggers a reboot of the server. This surprised me as my UPS is a 1500VA line interactive / pure sine wave, with nothing else connected to it besides the server.

 

I'm thinking the issue may be either the UPS or my PSU. I'm wondering whether I have a config issue on the UPS or if I'm overloading my PSU / have a faulty PSU. I have put my specs below, including a rough calculation for power usage

 

 

UPS: APC Spart-UPS 1500VA (SMT1500IC)

PSU: Superflower Leadex 80 Plus Platinum 750w

Motherboard: Asus X99-Deluxe (?w)

CPU: Intel Xeon E5-2678v3 (12C/24T @ 2.5GHz, 120w TDP)

RAM: 8x 8Gb Hynix DDR-2133 ECC (5w per stick?)

GPU 1: Asus GTX 970 STRIX OC 4Gb (148w TDP)

GPU 2: Asus GT 710 2Gb (19w TDP)

NVMe: 2x Samsung Evo 970 1Tb (one via PCIE add in board. 5w under load each?)

SATA: 6x Seagate Archive 8Tb (ST8000AS0002, 7.5w at load each), 1x Samsung 850 1Tb (5w?)

Fans: 4x 140mm fans (4w each?)

USB devices: Logitech webcam, Jabra conference mic, sometimes an iPhone, Bluetooth dongle, sometimes recharging keyboard

 

Estimating 400-450w load? Would have thought a 750w platinum PSU from Superflower should be able to handle this.

Edited December 28, 2020 by KuniD

[JonathanM]

Easy first test, use a plain surge protector instead of the UPS and see how it behaves.

 

Gut feeling says PSU, since normally PSU's can handle a remarkable amount of input voltage fluctuation without sagging the output, at least if they are running with a decent amount of reserve overhead. To have a system be that touchy leads me to think that the PSU or the motherboard power circuits are misbehaving. There is a decent amount of smoothing and filtering done on the motherboard as well, so that's a possible candidate besides the UPS and PSU.

[Ford Prefect]

13 hours ago, KuniD said:

SATA: 6x Seagate Archive 8Tb (ST8000AS0002, 7.5w at load each), 1x Samsung 850 1Tb (5w?)

 

Estimating 400-450w load? Would have thought a 750w platinum PSU from Superflower should be able to handle this.

...during spin-up, the load for these HDDS can easily double up-to triple that amount.

But that still would leave you under the max margin of your PSU.

How old is the PSU (as they can deteriorate over time and loose approx 10% of their nominal maximum).

Confirm that P4/P8/EPS power sockets are still well connected to the MB and GPUs.

 

[KuniD]

On 12/18/2020 at 12:39 AM, jonathanm said:

Easy first test, use a plain surge protector instead of the UPS and see how it behaves.

 

Gut feeling says PSU, since normally PSU's can handle a remarkable amount of input voltage fluctuation without sagging the output, at least if they are running with a decent amount of reserve overhead. To have a system be that touchy leads me to think that the PSU or the motherboard power circuits are misbehaving. There is a decent amount of smoothing and filtering done on the motherboard as well, so that's a possible candidate besides the UPS and PSU.

 

On 12/18/2020 at 11:21 AM, Ford Prefect said:

...during spin-up, the load for these HDDS can easily double up-to triple that amount.

But that still would leave you under the max margin of your PSU.

How old is the PSU (as they can deteriorate over time and loose approx 10% of their nominal maximum).

Confirm that P4/P8/EPS power sockets are still well connected to the MB and GPUs.

 

 

Thank you both - following your replies I decided to get more organised and wrote up a series of elimination tests to isolate what the problem was. In the end I confirmed it was indeed the PSU. I ruled out GPU's ,cabling, CMOS battery, UPS v mains, etc. I borrowed an old Antec PSU from a friend and it ran the system fine without issue.

 

If you can believe it I had 8 days left on my 5 year warranty from Super Flower. I have RMA'd it and am awaiting a replacement.