At 11:00AM this morning, I decided to get on some benchmarking I had to get done. A problem arose, however. I needed to rebenchmark the GTX 580, but the one in the test machine had an after-market cooler, making it less-than-ideal for temperature testing (in a head-to-head / apples-to-apples scenario). I have been using the other GTX 580 on-hand in my personal PC, and as it's using a stock cooler, the chore in my future was obvious.
What should have been a simple swap was anything but. After the job had been done, I started experiencing GPU issues on not just one machine, but both. Nuts, huh? I was left baffled. I can understand one little thing going awry to cause me some grief, but two PCs at the same time was insane.
After hours of trouble-shooting, I noticed that on the card in the benchmarking PC, a couple little marks were found on the gold contacts on the PCIe connector. It looked like rust, but wiped off just fine with Q-tips and rubbing alcohol. After re-installing the card and testing it out with a one-hour loop of 3DMark 11, it seems the problem has been fixed. For that PC, at least. Back to mine.
I took out the card and installed a total of four different other cards... all which changed the situation none. I reinstalled the GTX 580 again, flustered, booted on the PC and went to the kitchen to make a sandwich. When I returned, there was video on the screen. How, and why? It seems that the GIGABYTE X58A-UD7 motherboard I am using decided to complain about a failed overclock (the PC isn't overclocked), and as such, the reason video didn't show is that the board was trying to get itself to run properly. Had I waited about three minutes on any given boot, I could have saved a lot of time.
But not all was right. I tested out a game quickly, and the graphics driver crashed. I tried another game... the graphics driver crashed. Rinse and repeat across multiple reboots and games. In one instance, I merely loaded up a YouTube video and all of Windows crashed - because the video uses GPU acceleration.
I took the card out and noticed subtle marks on this one also. It became clear quick that something was wrong with that particular PCIe slot in my PC, as it had to have marked up the other card and then this one once I installed it. I cleaned out the with Q-tips once again and tested things out again. Still no luck... same exact issue.
I pulled a rinse & repeat here... except I cleaned the slot a little more thoroughly (I am not even going to go into details, as my method is truly ridiculous). Finally, the card seemed to work well. I loaded up 3DMark 11, started the default suite and things seemed good. Until it crashed about half-way through. "The PCIe slot is busted", I thought. So, I moved the card down to another x16 slot and booted it back up.
Before I ran more testing though, I looked at GPU-Z logs I had made prior and realized that the card had reached 101ºC the last time it ran. Once I looked back in the PC, I realized this was due to my not plugging in the GPU fan... so that 3DMark 11 run was being run with no airflow. Wee.
In the end, that last thorough cleaning of the slot worked and I'm finally back to having a usable PC. Unfortunately, this entire clusterfrog of a hassle sucked down ten straight hours of my day, totally pushing me back on things I've been needing to get done. All because of a PCIe port that decided to toss some random dirt onto each of the GPUs. *sigh*
Oh well, tomorrow will be better ;-)
What should have been a simple swap was anything but. After the job had been done, I started experiencing GPU issues on not just one machine, but both. Nuts, huh? I was left baffled. I can understand one little thing going awry to cause me some grief, but two PCs at the same time was insane.
After hours of trouble-shooting, I noticed that on the card in the benchmarking PC, a couple little marks were found on the gold contacts on the PCIe connector. It looked like rust, but wiped off just fine with Q-tips and rubbing alcohol. After re-installing the card and testing it out with a one-hour loop of 3DMark 11, it seems the problem has been fixed. For that PC, at least. Back to mine.
I took out the card and installed a total of four different other cards... all which changed the situation none. I reinstalled the GTX 580 again, flustered, booted on the PC and went to the kitchen to make a sandwich. When I returned, there was video on the screen. How, and why? It seems that the GIGABYTE X58A-UD7 motherboard I am using decided to complain about a failed overclock (the PC isn't overclocked), and as such, the reason video didn't show is that the board was trying to get itself to run properly. Had I waited about three minutes on any given boot, I could have saved a lot of time.
But not all was right. I tested out a game quickly, and the graphics driver crashed. I tried another game... the graphics driver crashed. Rinse and repeat across multiple reboots and games. In one instance, I merely loaded up a YouTube video and all of Windows crashed - because the video uses GPU acceleration.
I took the card out and noticed subtle marks on this one also. It became clear quick that something was wrong with that particular PCIe slot in my PC, as it had to have marked up the other card and then this one once I installed it. I cleaned out the with Q-tips once again and tested things out again. Still no luck... same exact issue.
I pulled a rinse & repeat here... except I cleaned the slot a little more thoroughly (I am not even going to go into details, as my method is truly ridiculous). Finally, the card seemed to work well. I loaded up 3DMark 11, started the default suite and things seemed good. Until it crashed about half-way through. "The PCIe slot is busted", I thought. So, I moved the card down to another x16 slot and booted it back up.
Before I ran more testing though, I looked at GPU-Z logs I had made prior and realized that the card had reached 101ºC the last time it ran. Once I looked back in the PC, I realized this was due to my not plugging in the GPU fan... so that 3DMark 11 run was being run with no airflow. Wee.
In the end, that last thorough cleaning of the slot worked and I'm finally back to having a usable PC. Unfortunately, this entire clusterfrog of a hassle sucked down ten straight hours of my day, totally pushing me back on things I've been needing to get done. All because of a PCIe port that decided to toss some random dirt onto each of the GPUs. *sigh*
Oh well, tomorrow will be better ;-)
