MP 1,1-5,1 GTX 680 Kernel Panic…

[naught@home]

Hi. First post, but I've been lurking for decades… ?

I have a cMP 3,1 (2 x 3.2 GHz Quad-Core Intel Xeon) with 8GB or RAM. I recently got my mitts on a flashed eVGA GTX 680 Classified 4GB card with two 8-pin connectors (04G-P4-3688-KR). The card is in PCIe slot 1 and connected to the board with a pair of mini6-pin to 8-pin connectors. I'm running the latest dosdude1 patched High Sierra (build 17G14019) with the latest nVidia web drivers and CUDA 10.2 driver.

The card boots up fine. I have the EFI boot screen and am able to boot into the Startup manager with no problem. I've been running fine with the case open for the last few days with no problem and now it is time to put the side panel back on the cMP. The problem is that as soon as I touch the cable (push it into the case to put the panel on) the screen goes black and all the fans start to roar. The only relief is a forced shutdown. If I try to restart immediately the progress bar gets a little past halfway (to the point, I think, that the video drivers should load), then stops and boot up refuses to progress to the end.

If I shut down the cMP, unplug the computer and the monitor from the mains, then disconnect and reconnect the two 8-pin connectors on the 680, plug the cMP and the monitor back into the mains the next boot will be just fine. But if I touch the cable again, same thing; black screen and roaring fans. Once booted to the Desktop I see a 'your computer has restarted because of a kernel panic' or something to that effect. I've copied and pasted the report to a text file and will paste it into a subsequent post if someone thinks there is a need. Right now I don't want to create a long gnubee post that no one is going to read.

Two questions:
1) Why can't I move the cable to put the side panel on without causing a kernel panic?

2) If question one can't be resolved, can I run my 3,1 with the side panel off without causing any show stopping difficulties? The 680 seems to be fine as long as I don't touch the cable even with all CPU cores maxed doing a Blender render.

If it makes any difference, the monitor the 680 is powering is a 23" Apple Cinema HD display, but NOT the aluminum one. Mine is the very old (2002) polycarbonate one. Yes, it still works great after eighteen years. And yes, I need more RAM.

Thanks for any insight given, and sorry for the long post.

 

[ActionableMango]

The problem is that as soon as I touch the cable (push it into the case to put the panel on) the screen goes black and all the fans start to roar.

Simply touching the cable shouldn't cause anything at all to happen. This sounds like a physical problem. Either a bad cable, bad pin adapter, bad power connector on the card, or bad power connector on the backplane board. There is a marginal path somewhere, and touching the cable triggers either an open connection or shorting to ground.

First step would be two remove the GPU power cables. Examine all of the following for damage, corrosion, and loose connections:

• Your two GPU power cables
• Your two 6-pin to 8-pin adapters
• The two 8-pin GPU power connectors
• The two 6-pin AUX PCIe power connectors on backplane board

 

[naught@home]

Thanks for your interest, ActionableMango.

The cables could be bad. The Amazon man brought them to me the other day, so they're brand new, but that doesn't mean one or both weren't borked from the get go.

I will pop the card out and examine cables and connectors as per your suggestion and report back. Any suggestions for cleaning pins? I have isopropyl alcohol and Q-tips.

Another question: Assuming one/both of the mini6-pin to 8-pin connectors is bad, are there any alternate ways to cable the GPU?

I have at hand, a pair of mini6-pin to 6-pin cables and a pair of V-shaped cables, each of those has two 6-pin connectors at the wing ends and an 8-pin connector at the point. Plus I have the two, now suspect, mini6-pin to 8-pin connectors.

I tried

1. a pair of mini6-pin to 6-pin cables plugged into the 8-pin sockets: No boot, though I didn't really expect it to.
2. a pair of mini6-pin to 6-pin cables plugged into the 6-pin ends of the V-shaped cable with the 8-pin connector plugged into the rightmost 8-pin socket on the GPU: No boot. Same thing when plugged into the leftmost 8-pin socket. After some research I came to realize that all sixteen pins on this particular GTX 680 Classified 4GB GPU need to be populated for it to run properly. So then I tried…
3. a pair of mini6-pin to 6-pin cables each plugged into one of 6-pin connectors at the wing end of each of the V-shaped cables with both 8-pin connectors plugged into the GPU and the second 6-pin connectors on the V-cable flapping in the wind. No boot joy there either. Further research led me to the conclusion that the unpopulated 6-pin sockets on each V-cable causes the GPU to sense that it's not grounded properly and hence it refuses to load the drivers and send a signal to the monitor.

Does anyone have any other ideas? I appreciate all insight…​

​

 

[Nguyen Duc Hieu]

My idea is not new: Go for the Pixlar Mod. Drawing power directly from the PSU is the best way to power your greedy VGA card.
But, another thought: Maybe the problem resides in the two 8-pin on the VGA card, bad connections, broken solders, etc... Should check that first. If it is fixed, then no Pixlar Mod is needed.