Fatal Error - Please Power Down And Connect The PCIe Power Cable For This Graphics Card

[In-Trouble]

[[[Writing from wife’s laptop]]]

I know this isn't really hardware troubleshooting forum but I don't really know where else to start & I've found folks here to be quite knowledgeable in the past. If you can direct to more appropriate forums (except reddit/Quora which I've to be fairly uninformative) I'd be very appreciative. Having said that I'll still post my situation in case someone here can & will help (BTW: new registered name since I'm not on my machine, so yep "post #1").


This may seem a bit lengthy but I’m attempting to provide sufficient info for troubleshooting on the first go; system spec:
motherboard: Gigabyte z270 HD3
CPU: i7-6700
GPU: MSI GTX 1660S
RAM: 8x2 GB Corsair Vengeance 2400
PSU: Cooler Master Bronze 650w
OS: Win 10 1909
Age: 2 yrs except GPU which is about 3 months old



PROBLEM DESCRIPTION

I’m experiencing the following fatal error:
PLEASE POWER DOWN AND CONNECT THE PCIe POWER CABLE FOR THIS GRAPHICS CARD (apologies for all caps but that's how it's actually displayed)

The first occurrence was two weeks ago immediately following a Win 10 Nvidia driver “update” (the driver was much older than what was installed but still potentially only coincidental I suppose) while working in (current version) VirtualBox.

I powered down, waited a ½ hr and the machine booted fine and then functioned acceptably for the ensuing week.

Initial research indicated either [A] a hardware problem; poorly seated/incorrectly connected cables, insufficient/dying PSU, dirty interior/PCIe lane, old/dying GPU and corrupted PCIe lane OR display driver conflict.

Since:

• • the GPU is new, high end of mid-range, recently installed and has been functioning flawlessly for 3 mo.
  • the PSU is two yr old and calculations of the aggregate power requirement of my build ranged from 390w to 420w
  • apparently vBox has long been known for causing fatal device driver conflicts (per a fairly reliable source, my brother-in-law)

I focused on resolving the driver possibility.

I used a tool called DDU to purge Nvidia from the system and reinstalled the proper Nvidia drivers; this initially seemed to have solved the problem.

On Mon/Tue last week – I forget which – Win 10 installed an AMD based generic display driver and a new problem emerged when the video signal was again lost shortly after the update install. The only available solution was power down with power button, wait a bit and cold boot. While this could also be coincidental the timing is suspicious.

For the past week a variable amount of time-- occasionally 2-3 minutes but could go for an hour or so -- after starting each morning the video signal would be lost necessitating the physical power down/cold boot. After that it ran fine for the rest of the day (approx. 14 hrs).

The same cycle started today but when trying to cold boot instead of POSTing normally the PLEASE POWER DOWN AND CONNECT THE PCIe POWER CABLE FOR THIS GRAPHICS CARD message was displayed.

At present the machine will boot if left off for 15–30 minutes but in only 2–3 minutes vid signal is lost. My last attempt I entered EUFI/BIOS – reported system temp was 36 C (which is normal) – and even only sitting in BIOS vid signal was lost in less than 5 min.

Yesterday I logged performance for about 9 hrs. All but two indicators (based on my 65 yr old memory) were within expected ranges: all temps running in high 20s/low 30s (or about 8–12 degrees above ambient), CPU & GPU usage averaging 14% and 16% respectively and CPU cores clocking in the green (can’t remember the numbers). The two indicators outside the norm were (again from memory) Performance limiter flags: GT:Residency State Regulation & GT:Running Average Thermal Limit, both of which were “Yes”.

Windows memtest & sfc /scannow were run successfully last week with no errors found.

Any help you can provide will be much appreciated,

 

[Dalai]

Since you said the video signal is also lost when just sitting in the BIOS, I rule out everything related to the OS, including drivers. Did you (temporarily) swap the PSU to rule it out (or to find it to be the culprit)? Sure, it has enough wattage and it's from a reliable brand, but it's possible that it can't deliver enough power on all required rails at all times - especially at boot time when the BIOS checks for the PCIe power cable.

Regards
Dalai

 

[Zach]

I would agree with the above.

Also:

TechPowerUp Forums

Computer hardware discussion forum, focused on tweakers, overclockers, gamers and enthusiasts.

www.techpowerup.com

 

[In-Trouble]

@Dalai & Zach -- completely agree it's now primarily a hardware problem. In my haste to post yesterday I failed to acknowledge that and until the hardware issue is resolved anything else is secondary. Based on your responses I should just eliminate the whole driver debacle discussion from the problem statement; I thought it might have some value in the troubleshooting but apparently not. In case it's important, the problem is more pronounced today with the "no video signal" message and the display going to standby before the America Megatrends BIOS logo could be displayed.

Note regarding an immutable process constraint: I'm physically disabled and changes inside the tower case or to the rear of the case requires engaging competent assistance which takes time with timing delays now exacerbated by the caronavirus. My custom build is my design including all capability/compatibility research and component acquisition but for physical assembly and interior/exterior cabling I'm restricted to a supervisory function. Implications with respect to proper assembly are obvious (and a risk I accept to get the system I want). The situation is simply a process constraint which I accommodate. No offense taken btw --there's no way you could have known.

So at this juncture i'm trying isolate which component is the root cause so a replacement can be ordered. It's intriguing to me that -- given video signal loss which to me indicated a potential GPU problem -- you zero in on lack of enough power. I had viewed a PSU as a very mature technology and therefore highly reliable for a long time if the unit is a reputable brand. For what it's worth I also posted this at TenForum (a Windows 10 OS forum) and got the same advice

Thanx for your time and expertise (and the pointer to TechPowerUp - it looks promising).

 

[Dalai]

Don't worry. Giving too much information is less likely to happen. It's never a bad thing to tell people what has been checked already and some of the thought process. Most of the time there is too little information provided so that asking questions is necessary.

I don't exactly rule out the GPU as the culprit, but given the information you provided, it seems more like a PSU issue. Yes, PSU tech is pretty mature at this point. I don't think the PSU itself is broken (although I don't rule it out either), but it's possible that a plug is poorly seated - and if it's a modular PSU, both ends of the cable(s) need to be checked. Not sure if you checked those already, your OP only mentions it as a possible thing to look at.

It may even be as simple as the graphics card not fully inserted into the PCIe slot. These slots can be annoying at times. Little story on the side: A couple of months ago, I needed to put an additional washer under the motherboard on top of one of the standoffs (next to the PCIe slot) to make sure that the GPU can go all the way in and makes good contact. Admittedly it was an old motherboard that was already bent a little bit; that's what motherboards tend to do over time.

It completely eludes me why Windows would (want to) install an AMD driver on an Intel-based system with an nVidia GPU... But Windows 10 is a whole other story anyway .

Regards
Dalai

PS: I know it takes time to properly check all components, even for a non-disabled person. So take your time.

 

[In-Trouble]

Thank you Dalai.

If it is a connections issue (either GPU seating or power cables (the PSU is "semi-modular")) wouldn't that have surfaced shortly after GPU install (5 mo ago)?

Or can things work loose over time? I'm careful to not bang the tower case but I guess there must be some vibration in the system with the CPU/PSU fans running ...

RE Win 10 - in complete agreement; I only "upgraded" from 7 a mo ago because I'm starting shop for new hardware. Not that there's necessarily a cause-effect happening but that is when my troubles started; I confess it's made me suspicious that Win 10 is fiddling with my motherboard firmware or some esoteric EUFI settings

 

[Dalai]

It's possible for plugs to get loose over time if they haven't been fully inserted into the connector until the latch "clicked". Vibrations are less of a concern in this regard but cables tend to have a bit of weight to them and may "pull" on the connector/plug; this also depends on the cable management of course. And graphics cards are putting weight on the PCIe slot, making the card sag over time, especially heavier ones (although yours isn't on the heavier side).

Regards
Dalai

 

[In-Trouble]

ok - thanks.

 

[ddDerelict]

[[In-trouble's alter-ego]]

Problem turned out to be a defective GPU. Pulled the card and (after downloading/installlng the necessary Intel drivers) I've got surprisingly acceptable graphics performance just with the CPU's on-board graphics capability.

Thanks again for the help and pointers.

Administrivia question: I created the in-trouble account only because I didn't have access to my computer with its attendant saved ids/passwords. Should I delete that id & if I do will this thread disappear? I'd like the thread to remain in case it might helpful for somebody else ...

 

[Dalai]

Thanks for the feedback. Now it's logical that the BIOS won't complain about missing PCIe power cables when there's no dedicated GPU installed. Did you replace it with another one? If so, did the issue go away?

Regards
Dalai

 

[ddDerelict]

I am in the middle of the RMA process (in case somebody from Belgium reads this it was purchased from Alternate.be & they're replacing the "defective unit" with no hassles).

So at present, no, I've not replaced with a different dedicated GPU; I'm just using the i7 6700's onboard graphics (which are quite noticably inferior) as an interim solution. According to the technician who helped me out we won't know if the GPU was truly the root cause until we get the new one installed. His reasoning is that the GPU is the biggest power consumer in the system and without it the power "draw" is low enough that a failing 650w PSU wouldn't be noticed. I did log -- via HWinfo -- system performance for a few hours (10,626 Excel rows) the day before the fatal failure but I'm not technically knowledgeable enough to know how to interpret the data. I had hoped I could find indicators that I could use to determine whether the root cause was GPU or PSU and that I could also use in the future as "warning signals" of imminent component failure but no joy

We also found that the rubber sheathing on the main power cord (between the wall and the PSU) had deteriorated to the point that internal individual wires in the cord were damaged and could not pass the conductivity test (the technician has meter with alligator clips and metal probes for testing that). Whilst replacing the cord didn't solve the problem there was speculation that inconsistent power levels may have contributed to -- if not outright caused -- electronic component failure. According to the tech sharp power fluctuations can be deadly for electronics.

Once we get the replacement GPU installed I will update status.