[SOLVED] How easy is it to mess up the PCIE x16 Graphics card slot? How easy is it to mess up a mobo period?

unclegrumpskin

Honorable
Dec 5, 2014
15
0
10,510
I have an MSI X470 Gaming Plus Mobo and I upgraded my gfx card the other day. When I was taking out the old one (a GTX 1060), I forgot to unlatch the slot lock switch thing, and I was pulling it out with it stuck in only the far right side, and the rest of it came out with ease. I didn't pull at the stuck part very hard, but I was pulling at it for a bit of time before I remembered I didn't unlock the latch. I remembered, unlocked it, got it out, installed the new card (RTX 2060), tested, benchmarked, etc, and it seems fine, I think...

But what if I'm wrong? Can a slot be only 5% broken and still function, but potentially not be working the way it should? Could the same be said about every piece of PC hardware? Just how fragile is the motherboard anyway? Now I'm paranoid about everything and spent a ton of time researching stuff like frame times in games and started measuring hitching and stuttering with MSI Afterburner's frame time measuring graph. The results are different from game to game, since some engines run differently, but it seems mostly consistent. Mostly. So I'm worried that the few little frame time stutters aren't the result of the game, but of my hardware being mildly broken. My temps are fine though, they're better than before.

Also, are you only supposed to pull out the gfx card completely vertically? When I pulled it out and it was stuck, the left side that came out fine came out to about a 50-70 degree angle before I remembered to unlock the latch. Will that screw up the pins and connectors?

Here's basically what happened:

hk3mEr1.png
 
Solution
Woah, take a breath. You are good. That board has those reinforced metal slots first of all. If there was some kind of damage to the lanes then the pci-e would do a fault tolerant thing and configure it to work with fewer frames. If you want to check if it's doing that then gpu-z can check how many lanes are in use. You WILL get inconsistent frame stutters running benchmarks. It's common practice to run the benchmark multiple times and take the average if you want to report it. Unless you can do chaos theory there will be some degree of entropy that can't be accounted for with what all the electrons are doing in your system no matter how close each run is to being in identical conditions.

A sudden gust of air in the room can put cause...

froggx

Honorable
Sep 6, 2017
78
33
10,570
Woah, take a breath. You are good. That board has those reinforced metal slots first of all. If there was some kind of damage to the lanes then the pci-e would do a fault tolerant thing and configure it to work with fewer frames. If you want to check if it's doing that then gpu-z can check how many lanes are in use. You WILL get inconsistent frame stutters running benchmarks. It's common practice to run the benchmark multiple times and take the average if you want to report it. Unless you can do chaos theory there will be some degree of entropy that can't be accounted for with what all the electrons are doing in your system no matter how close each run is to being in identical conditions.

A sudden gust of air in the room can put cause a slight change in what your intake fan pulls in just enough to push a temp. sensor beyond hysteresis for some fraction of a second and cause a vdroop that would never be noticed except this one is at just the right time to make a 0.00000000001v drop on your 12v headed to the gpu and a relatively insubstantial bit gets flipped somewhere that won't break anything by any means but it sneaks into a the framebuffer and adds a single texel that won't ever be noticed visually however some buffer somewhere is perfectly full and can't take the data so it gets pushed the distance of like 2 electrons on the die in perfect synchrony with the benchmark data polling for performance which then gets logged incorrectly because something imaginary happened in some shift register from whatever like 3 pages ago and you that as a variance in the run.

Ideally when you remove a graphics card it will just pull straight up. Keep in mind that with computer hardware you can do everything perfectly and still have something break, so it's not worth thinking about too much. I like to think of PCI-E cards as "1) did it come out without making a horrible cracking noise? 2) if I glance at the now empty slot do I find myself saying 'holy crap how is that even possible?'" If 1) yes and 2) no, then it's most likely good.

I have some examples even from when I have totally done all kinds of magic to the slots and not broken them or done something super trivial and ruined my day. I constantly forget to undo that little latch and get my cards at all kinds of angles. One of my cases has a bulge going outward on the back from where I whacked it with the butt of a screwdriver to "correct the case's clearance defect."

In a completely separate situation with another comp all I did was move it across the room and while I'm plugging it all back in I hook up the monitor and feel a static electricity shock in my fingers, hear the monitor degauss (I still have a couple CRTs) and get hit with sweet smell of cooked computer parts. The GPU, CPU, motherboard and 2 sticks of RAM all died of ESD right then and there.

Another time I forgot to hit that bracket and got frustrated so next thing I know I'm standing up holding the GPU and told it all kinds of mean things and notice "Hey this thing seems way heavier" and it turns out I've lifted the entire case off the ground from the GPU caught on the bracket and my brain pulled a BSOD for about 4 seconds then the entire case drops back to the ground and I'm left holding the GPU. I knew there was all kinds of hell that I had just unleashed so I did what everyone does when they see a bad car crash and say "Oh crap that's horrible" but they're also trying to get a better view. I get the view I need of the PC apocalypse and all I notice is that the PCI-E slot is slightly janky like, tilting slightly sidewise not too bad, everything else looks fine and that only has me freaking out worse cause that can only mean that secret damage is hiding in there and it's gonna just lurk until I'm in the middle of compiling something major then erase everything, but I put that out of my mind and took a deep breath because it was time to do some standard trouble shooting aka reach blindly in the case and twisted the slot back to where it was supposed be, fully expecting the slot to come clean off the board. I get it pretty much looking normal and wanted to get the bad news of irreparable damage out of the way so I put the GPU in the slot, or at least try, but the cooler is big enough it's more of a groping blindly thing and I'm jamming it into the motherboard itself instead of seating it in the slot thanks to the kind of adrenaline flood that had hit me that most people only experience when they have to lift a car off their kid or have a knife fight with bear. A couple more tries and I seat it, hook the monitor in, and this thing flies through POST and boots up fine, I was so angry.

Something to keep in mind is that with a PCI-E slot, people are shoving graphics cards in that have a massive chunk of copper on them, some fans and other things, then asking the slot, along with a screw or 2 to support it. Some physics and you've got Force = Mass x Acceleration. Acceleration is gravity, 9.8m/s^2, because, well, Earth is like that. Mass varies, but in the end you've got a not insignificant amount of force being applied. Being that humans like to take things and say "How can I make this more extreme?" the natural choice is to mount the motherboard vertically and now you've got the card and the copper applying both linear and torsional forces on like 3 axes due to the screw in the case holding up one corner while the other corner is unsupported. All that has different amounts of force acting along the PCI-E slot. The only thing that can be done is to build the hell out of PCI-E slots that will be holding GPUs. With all that going on though the back of the card needs a little extra security, hence that bracket, which is crammed in more or less the hardest to see place it can be. Now the PCI-E slot has to be built up even stronger to idiot proof it. You seemed really stressed but you went to work on a slot that is designed to survive people trying to pull it with higher or lower levels of success while the bracket still secures it. The card itself is meant for the electrical connectors to mate securely enough to do the electric stuff it does. The connectors are relatively large and the one's on the GPU don't bend around like CPU pins or anything. For good measure, I even did some surprise field testing on these slots and in my experience they hold up very well.
 
Last edited:
Solution

unclegrumpskin

Honorable
Dec 5, 2014
15
0
10,510
Woah, thanks for the very comprehensive response! I did check GPU-Z and it does seem to be using the x16 3.0 version of PCIE so that's good. And forgive me for being paranoid, as this card was very expensive and I want to make sure I didn't break anything and that it's working as intended. I've been building computers for myself on my own for the past 6-8 years or so, and it's easy for me to look back at my previous perceived building mistakes as supreme goof-ups that are hampering my performance. And this is quite an expensive hobby, so anything performing less than it should feels like a bitter pill to swallow for me. If I do start feeling paranoid again, I'll just video record my performance in different games and see what people think in a different thread. Thanks again.