Please Help Track Down (Temporary) Non-Boot (with VGA LED on)

[flexy]

Ok, try to make this short.

I am not a newbie when it comes to PCs. My system however is old.

Asus Z87-Pro, EVGA 970 SC, Toughpower 800W PSU, 16GB Ram, Win10 x64.

The PC worked for years without issues, but in the last few days I noticed there is a problem with sometimes the PC not coming up from a cold start in the morning. What happens is that the screen stays black, and the bright VGA LED near the PCIE slot on the board lights up.

After a while, for (still) unknown reasons, after many switches on/off, cold and warm starts, the PC might finally boot. It can also sporadically happen that the PC boots, but the system then only sees the GTX 970 as some weird VGA card w/ error 43, with GPU Z unable to get any information, like memory, sensors etc. EXCEPT the bios information that it's a GTX 970. Then I boot again, and the system finally recognizes the GPU and the system works entirely normal.

* What I already did:

Of course, re-seating the card, 5x+ times. Also tried on other slot.
Today I opened the card entirely and did a new thermal paste job as well, just out of boredom. (Very simple on those old EVGA cards)

Right now, after messing with it for hours today and several attempts to boot, now I am working in Windows as usual and everything is ok, except there is always a chance that the system won't recognize the card again at the next boot.

"Common sense" of course tells me it's the GPU slowing going, but I still can't explain the behaviour that ultimately after several attempts to boot, PC on/off etc. it at some point works. If the GPU would be shot, well logic says it would be shot and not work after some time.

Any idea? Maybe weak soldering spots on the GPU? Hair-line fractures or something?

Ty!

 

[C1]

In the ole days, a component such as a capacitor could easily come down with an intermittent such as a short or open circuit situation.

Typically when something like what is being reported happens to me it ends up being a bad connection, so I pull the system down and do a cleaning along with replacing the CMOS battery if there is one. (Some boards used capacitors or sealed battery packages soldered on to the MB.)

When vacuuming out the system, all major power connectors including PCI cards are removed and cleaned with DEOXIT before re-seating.

In your case, since the system seems to have no issues after warm up, then there is something you might try. When the system fails on a cold boot (eg, after sitting powered off overnight), then shut the power off and use a hair dryer to heat up the GPU then try re-boot. (No need to remove the GPU, just heat it up.)

If the system seems to prefer that the GPU is warm, then there probably is a problem with the GPU such as a bad component on the GPU board or maybe even a defective solder joint. (Solder joint cracks can be really tricky, almost impossible to verify visually. I try to handle verification of these through attempting flexure of the circuit card; eg, use a shim to temporarily warp the board slightly.)

 

[VirtualLarry]

Got a BNIB MSI Z97 PC Mate ATX Haswell-socket waiting for you, just send a PM, and some reasonable amount of Paypal.

Edit: Current market price is $189.99 for USED.
https://www.newegg.com/Product/Product.aspx?Item=9SIAFVV79R2349

 

[flexy]

Why I am asking, to determine whether it MIGHT be the board or the GPU. Heck, it could even be the PSU.
(Years ago I already had to fix the PSU because a cap was bulging, resulting in StandBy power not working and therefore no boot. Since the PSU is like 9000 years old, chances are it might not provide adequate voltage at times. But just guessing here of course)

C1, yeah I considered all these things, including the idea with the hair dryer. (It looks very much like it that the system starts easier when it's warm).

Problem, as I see it, it could really be everything.

If I can't track it down in the next days (I am sure I can't since it's one of "those" issues....), I will likely have to do exactly that, take everything out and clean/dust and re-assemble etc. (And do some more extended testing with the built-in GPU respective spare GPUs, to help me track it down more. This should establish whether the problem is with the GPU)

 

[Ketchup]

In your CPU has integrated graphics, try a few cold starts without the video card and see if the issue returns.

Another thing to try would be booting the computer with the video card in an alternate pci-express slot.

 

[flexy]

Ok I just did a few restarts/reboot with IGPU, no problem. (And earlier tested the GPU on the other slot, but same problem with booting).
So I guess it's the GPU.

 

[Ketchup]

Most likely. That test should narrow it down to the power supply or the video card. If the system never fails under load, I have a feeling it's something in the video card.

 

[flexy]

Although it's 99.9% the card (since it boots fine with the igpu) I haven't bothered swapping w/ my wife's GTX970. Maybe I should when I have time. (There is a slight chance it has to do with the PSU 3VSB voltage, like I had in the past). This COULD result in startup issues (aka PC not booting), and as I observe when it happens also not powering keyboard lights etc. And WHEN it finally starts, everything (ie. voltages) would look normal, since the standby voltage AFAIK is only used to get the system up (among other things) but not during normal use of the PC. (On the other hand, the last time when that cap blew on the 3VSB rail, the PSU shut off and there was no power, no fans and nothing. This time, fans are spinning. Means PSU is definitely providing power. I also measured yesterday with a multimeter when I was playing around, and voltages looked fine)

And no, if it boots (like today after approx 20-30 attempts) and it finally "gets" the card, I can play and run benchmarks like normal..