Fatal error: Please Power Down And Connect The Pcie Power Cable ...

[[from wife's laptop]]

This may seem a bit lengthy but I’m attempting to provide sufficient info for troubleshooting on the first go.

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 [B]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 but the timing seems 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. 14hrs).

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 the 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 (using HWinfo) 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,

First off, thank you for responding.

am I to presume your My Computer specs are not up to date

No, the specs are current. Your explanation of AMD64, however, caused me to recall (can't check since my machine is now totally DOA) that the Win 10 update title included "AMD64". I suspect now that due to MS previously "updating" my Nvidia driver with an old version I keyed in on the AMD thinking another display driver snafu had occurred. Given your explanation it seems likely the update I referenced had nothing to do with the display driver.

CHECK very carefully the pins on the card for the PCIe slot and the connection for the 8 pin power

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). Whilst my custom build is my design including all capability/compatibility research and component acquisition, 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.

I understand the reasoning underlying double-checking GPU seating/cable connections or power strip interference but am a bit skeptical because:

1. the card was installed roughly 5 months ago and performed flawlessly until the most recent two weeks
2. degradation of functionality appears to have been swift and irreversible -- this morning (our time, I'm in BE) the problem was more pronounced with [1st try] the "no video signal" message appearing and the display going to standby before the American Megatrends BIOS logo could be displayed and [2nd try] POSTing but losing video signal almost immediately after I'd entered BIOS (the "Please power down etc" message was not displayed in either attempt)

It seems (tho I'm very willing to be educated here) if it were a connections/interference root cause wouldn't performance degradation would be more intermittent over a more extended time-frame rather than a fairly a abrupt decline? Doesn't that sharp decline indicate component failure?

As another bit of info the system's PSU is semi-modular & I've read that those can be unreliable.

Looking forward to continued dialogue ...

I know you appreciate that I was not aware of your problem in examining the inside of the case and the card

Totally, no way you could have known unless I made you aware.

Please confirm you mean that the computer has now NO indications of power - even if there is no display eg power led - HDD led etc on tower.

Power is reaching the machine. I can intermittently get into BIOS but before I can examine anything the screen goes blank/black and displays the "no video input" message ...

I'm currently attempting to identify/engage somebody to help me do the "examine card seating/cable connections" activity you suggested.

Thank you for the extensive & detailed instructions; they helped progress the troubleshooting.

A preliminary result is in; tentatively it looks like a defective GPU. The card has been pulled and the system is now functional.
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 I've not replaced the defective unit with a different dedicated GPU; I'm just using the i7 6700's onboard graphics (which are quite noticeably 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.