GTX 770 fried on boot when tried to use DVI-I to VGA

Jandin

Prominent
Mar 27, 2017
2
0
520
Hello all! First, I had this fully working system:

Intel i7 3770
Gigabyte GTX 770 Windforce 2GB, connected to two HDMI monitors, using the HDMI and the DVI-D ports
generic mobo
generic 2x4GB RAM
250GB SSD, 2TB HDD
Seventeam 850W ATX v2.3 V-Force Series ST-850Z-AF

Then I moved and had to leave my monitors. Got two used DVI-D/VGA BenQ monitors and decided to upgrade my mobo with a used Intel DH67BL (SATA3, USB3 and more RAM slots). The "new" mobo wasn't 3rd gen ready but was 3rd gen capable (never heard of AA#? neither had I), so I bought a cheap i3 to upgrade the BIOS exactly as instructed in the Intel website. That was a bit of a pain but it all ended up nicely and everything was working perfectly with the "new" mobo and the i3 and one monitor in the 770's HDMI port.

Then I put my i7 back in and plugged in the second monitor with a VGA cable and an adaptor in the 770's DVI-I port (which I had never used before). As soon as I turned the thing on, fans went crazy and something smelled burnt.

Turned off the PSU, took everything off and started assembling it back, and things were working via integrated graphics, dual monitor even (mobo also has DVI-I and HDMI ports). I left the GPU out until everything else was "working".

Two strange things happened though: after a normal Windows 10 boot, I went into the BIOS and it showed Intel ME Firmware version of 0.0, which I promptly flashed back to 8.x. This hasn't happened again. Also, my Windows 10 couldn't sleep or turn off anymore, resetting instead.

Finally, I put the 770 back in using its HDMI port and the mobo's DVI-I because the 770 wasn't displaying video. After a CMOS clear, the 770 started working again and even the reset-only behavior went back to normal!

I though "well, if it can handle that once, why not twice, so I confirm the problem is the 770's DVI-I?" (before, the processor swap happened at the same time, and I changed how the system was sitting on the table. Call me incremental guy from now on). Confirmed, it fried again right on boot.

However, now the BIOS won't detect the 770 anymore, even after CMOS clear. Tested it in a working build and it didn't recognize it either. Tested a working GPU in my build and it got recognized. So I suppose the 770 didn't handle the second fry (well, that was dumb), but I still don't understand how the DVI-I port did that kind of damage. Mobo's auxiliary 4 pin connector was in place, the monitor, VGA cable and adaptor didn't fry the mobo's DVI-I (even though it is a bit shaky and flickery). Could a factory defect in the 770 that remained undetected have done that? Would contacting Gigabyte do anything? (warranty is long gone)

Any thoughts are appreciated.

[EXTRAS: the mobo and the aftermath]

Now I'm using the integrated graphics and the reset-only behavior is back. I've tried turning off network wakes, reset on crashes, uninstalled some drives and modified that registry entry, all without success. Funny thing, I found a guy who was having this exact problem with this exact mobo, but when using the PCIe instead of when it's empty. And I also found people having fan control problems with BIOS versions 156+, whereas for me fans are problematic with -155 (they start normal and scale up to pretty noisy in a few minutes with the system idle), even though with -155 I've managed to have a few successful sleeps and shutdowns.

For the kicks, I tested booting Ubuntu with the 156 BIOS (I have tri-boot). It was fine, but shutdown gave me HDMI green screen, DVI-I black screen, fans 100%, a little heart attack and a reflex of my finger to the PSU switch.

Windows 10 has also been resetting randomly, but it's hard to troubleshoot this with all the fiddling I've been doing.
 
Solution
It turned out the PSU was causing the reset-only behavior. It was also responsible for the shaking and flickering of the mobo's analog output, and possibly the random resets too.

The voltages were ok, but I read somewhere that even with good voltages, bad currents can cause this. A physical difference I noticed is that whenever I turned the faulty PSU off, the mobo's standby led turned off a few seconds later. With working PSUs, the led just dims (and maybe fades over a long time, haven't checked that). Also, if the led is dimmed and the faulty 24-pin plug is connected, it instantly turns off.

With a bad PSU in the equation, I can think up some explanation like: the 770's analog output circuits required PSU correctness in ways that...
It turned out the PSU was causing the reset-only behavior. It was also responsible for the shaking and flickering of the mobo's analog output, and possibly the random resets too.

The voltages were ok, but I read somewhere that even with good voltages, bad currents can cause this. A physical difference I noticed is that whenever I turned the faulty PSU off, the mobo's standby led turned off a few seconds later. With working PSUs, the led just dims (and maybe fades over a long time, haven't checked that). Also, if the led is dimmed and the faulty 24-pin plug is connected, it instantly turns off.

With a bad PSU in the equation, I can think up some explanation like: the 770's analog output circuits required PSU correctness in ways that weren't necessary for the digital output, and the PCIe power connectors had more problems than the ones supplying the mobo's analog output. 770's analog fries, mobo's analog only shakes and flickers.

I got a very good PSU now, and might try to revive the 770 once more, not touching either DVI ports ever again. However, even if it works, putting it under stress would be scary, so this looks like a good time to upgrade to a 1080. Still, what a sad way to say goodbye to an old friend, killed by a cheap PSU and uninformed decisions.
 
Solution