Dead PCI-E slot or PSU?

Soldato
Joined
16 Mar 2004
Posts
13,474
Location
UK
So recently acquired an RX 580 4GB from a sell/buy High Street kinda place and everything was hunky dory with the machine, specs are 9600k, B360N WiFi motherboard with 16GB RAM and running on EVGA G3 550W PSU.

Also picked up R7 370 a few days later for a friend's machine and when it arrived thought I would quickly test it in my system. Booted into BIOS but 9 times out of 10 it would lock up the system before getting into Windows. One time I got as far as installing the AMD driver, it just gave me a 173 error saying no card found. Looked pretty buggered so wrote it off as faulty and went to put the RX 580 back inside.

That's where problems began, as the card wouldn't display any image and wouldn't even show BIOS, green LED is on to say power is connected to it and fans spin, but nada. System is still booting and just going back to iGPU and works fine. I ran some Prime95 and even fired up games that Intel UHD 630 can handle like World of Tanks and everything is rock solid.

Figured might be an issue with RX 580 too, maybe old mining card that was on its last legs etc so set up a return for both of them. Just as a precaution, managed to borrow a 750 ti from someone (tested this exact card previously in this machine) and to my surprise having same issue, green LED on for power, fans spin but zero display, not even into BIOS, machine still works via iGPU though.

I cleared my CMOS, switched over the GPU cable (modular PSU) and tried second input on the power supply as well. BIOS is set to check PCI-E first too before switching to iGPU and I can't see either card in the Device Manager. Motherboard is ITX so don't have any other PCI-E slots to try. Checked for any loose bolts in case there's a short somewhere and inspected the slot, don't see any damage of any kind on it or dust inside.

Confusing part though is that if I put the R7 370 back in after all this, it boots into BIOS just fine but does its system crashing thing. RX 580 and 750 ti don't display any image whatsoever though so not sure if it can be a completely dead PCI-E slot. At the same time if my PSU somehow got fried, wouldn't the whole thing fry and simply not work?

Person I borrowed 750 ti from only has a 400W PSU so can't test RX580 in there and frankly at this point I would be very cautious about sticking R7 370 in there or any other machine. I will be returning the 750 ti (which was working as of this afternoon) just to see that somehow didn't get buggered now but I'm frankly a bit confused as to what to do at this point or what's at fault.

To make matters interesting, I picked up an R9 Fury from MM which should be arriving in a few days but guessing there's little change of getting that going at this point.
 
Soldato
Joined
6 Jun 2008
Posts
11,616
Location
Finland
There seems to have some variance in working with UEFI or legacy BIOS mode.
So there's change of needing specific mode. (like UEFI only/no CSM or having CSM enabled)

750 Ti again is such old that suspect it needs CSM enabled for legacy BIOS compatibility.
 
Associate
Joined
6 Dec 2007
Posts
1,368
Location
Cambridge
I'm pretty sure the 750 ti is uefi compatible. Testing it in another slot should rule out a bad slot or slot power delivery issue.

You might also want to reset your bios, then leave your computer unplugged and pull the cmos battery overnight. If something deprooted is screwed, that may clear it.
 
Soldato
OP
Joined
16 Mar 2004
Posts
13,474
Location
UK
You not got a second pcie x16 slot to try it in? try toggle csm/eufi in bios

There seems to have some variance in working with UEFI or legacy BIOS mode.
So there's change of needing specific mode. (like UEFI only/no CSM or having CSM enabled)

750 Ti again is such old that suspect it needs CSM enabled for legacy BIOS compatibility.
I'm pretty sure the 750 ti is uefi compatible. Testing it in another slot should rule out a bad slot or slot power delivery issue.

You might also want to reset your bios, then leave your computer unplugged and pull the cmos battery overnight. If something deprooted is screwed, that may clear it.

Board is ITX so only one slot to try and tried enabling and disabling CSM with no results.

To update, made the unwanted journey today to test the cards and my PSU in the machine I got 750 ti from.

750 ti works just fine in that machine with my EVGA power supply, ran Furmark without breaking a sweat.

RX580 is actually alive and well and working just fine in that machine with my PSU, ran Furmark with no issues.

R7 370 doesn't crash when in Windows, but crashed both times during driver install so definitely writing that off as faulty card.

Back to my machine now, took everything out the case to eliminate potential problems with a short somewhere and once again, zero sign of life from RX580 and 750 ti which I have confirmed as working fine on my PSU in a different machine.

Bar above advice to take the battery out, I think I'm looking at a faulty PCI-E slot as I can't think what else this can be. RX580 worked in there just fine until I put this R7 370 in, although this would be a first case I heard of a card killing a slot on a motherboard.

Only odd point is that out of three cards, the faulty R7 370 is only one to boot into BIOS before dying during load in Windows.
 
Soldato
OP
Joined
16 Mar 2004
Posts
13,474
Location
UK
I've inspected the board and don't see any kind of damage to the port or tab on any of the three cards. PC working fine otherwise on iGPU under load as well, don't have any other PCI cards I could try apart from the GFX cards.

580 is the card that's working after all, R7 370 is one that seems faulty on both machines at this point and problems began after I put that in. Other machine was fine though after I confirmed that it's faulty, other cards ran without issues afterwards.
 
Soldato
OP
Joined
16 Mar 2004
Posts
13,474
Location
UK
Well the battery didn't do anything but last thing I could try. I even stuck the faulty R7 370 back in before sending it back and it's not even recognising that anymore so guess it's a dead PCE-E slot.

Just either a very odd coincidence that it died after I put in that graphics card or this is a very rare case of a faulty card killing a slot.
 
Soldato
OP
Joined
16 Mar 2004
Posts
13,474
Location
UK
Well I received my new motherboard (Asus Z390i) and at this point I'm starting to think voodoo magic is involved because I'm having exactly the same issue, won't recognise any graphics card and won't display any image whatsoever even in BIOS. Plug cable into motherboard HDMI and PC works fine.

I once again took the PC and GPU to a friend's house (just 750 ti at this point) and card is working fine. I used my PSU and card works fine in a different machine. I use different PSU on my machine, exact same problem, no display whatsoever, but works fine with on-board graphics.

At this point, only components I haven't replaced are CPU and RAM, but machine is running any stress test I throw at it with zero stability issues from prime95 to Memtest to Furmark (with iGPU). Only theory I have is that something is potentially faulty with the CPU at this point, either the way it communicates with PCI-E or something forcing the integrated GPU to stay on all the time?

I have been clearing CMOS, selecting PCI-E as primary source and trying to disable iGPU altogether but all with same results.
 
Associate
Joined
6 Dec 2007
Posts
1,368
Location
Cambridge
I'm really struggling to see how the CPU or ram could be causing it, but at this point I'm stumped.

Going back to basics, you've definitely ruled out your PSU and motherboard (unless 2 different brands have the same problem). I don't suppose you could test it with a CPU that doesn't have an IGP? If you run with the IGP, does Windows see the 580?

Have you tried displayport output?

I did find this video walkthrough - https://www.youtube.com/watch?v=wmIOuruxGfY. It might be worth double checking that internal graphics is set to disabled in the chipset too.
 
Soldato
OP
Joined
16 Mar 2004
Posts
13,474
Location
UK
I'm really struggling to see how the CPU or ram could be causing it, but at this point I'm stumped.

Going back to basics, you've definitely ruled out your PSU and motherboard (unless 2 different brands have the same problem). I don't suppose you could test it with a CPU that doesn't have an IGP? If you run with the IGP, does Windows see the 580?

Have you tried displayport output?

I did find this video walkthrough - https://www.youtube.com/watch?v=wmIOuruxGfY. It might be worth double checking that internal graphics is set to disabled in the chipset too.

I've tried DP, HDMI and DVI connections. I've tried disabling onboard iGPU too, I set graphics preference to PEG and after restart that setting just resets itself to AUTO. There's even an option in BIOS to see your graphics card details and it simply says nothing is detected.

PSU powers same cards in a different (Haswell based) machine and PSU from that machine ends up with same, no display problem. I do have another PSU on the way this week so I'll be able to try 3 different power supplies on this machine at this point.

As for motherboard, I can only assume it's working but it sure would be very odd if I ended up with two motherboards with exact very rare problem to begin with. As I'm running out of options and money at this point to keep throwing at the problem, I might see if a local PC repair place will test the board for me with their components without charging me an arm and a leg for it.

Suppose other component I haven't mentioned is my M.2 drive but also fail to see how that would cause this. All else fails, anyone know a good computer exorcist?
 
Associate
Joined
13 Sep 2010
Posts
1,970
If you haven't already, I'd also test with just one RAM stick (try both, in both slots!) in case one has gone bad - if both appear to work I'd also run a memtest just to rule that out completely
 
Associate
Joined
6 Dec 2007
Posts
1,368
Location
Cambridge
Have you tried testing it outside your case? Assemble it on s piece of cardboard and install the bare minimum - so cpu and cooler, 1 stick of ram, gpu, psu. No sata/nvme drives etc.

If you can find a repair shop, it might be worth seeing if they've got a full sized mobo you can test your components in.
 
Soldato
OP
Joined
16 Mar 2004
Posts
13,474
Location
UK
If you haven't already, I'd also test with just one RAM stick (try both, in both slots!) in case one has gone bad - if both appear to work I'd also run a memtest just to rule that out completely

Have you tried testing it outside your case? Assemble it on s piece of cardboard and install the bare minimum - so cpu and cooler, 1 stick of ram, gpu, psu. No sata/nvme drives etc.

If you can find a repair shop, it might be worth seeing if they've got a full sized mobo you can test your components in.

One stick of RAM at a time didn't seem to make any difference and I tested Memtest overnight, 3 full passes. Did 30 minutes of Cinebench with no issues and Prime95 isn't giving any trouble, although funny enough as I went to start the PC to reply on here, it rebooted itself before getting into Windows and then actually BSOD in Windows after with kernel thread priority floor violation code. First time I've seen a blue screen of death on this machine.

I've already tried testing it out the case too to eliminate any shorts and I decided to get a new 32GB kit of RAM, have a beefier PSU coming from MM and even able to loan a 9600k so as it stands, this weekend I'll have enough components to put 2 PCs together so should be able to eliminate the problem!
 
Soldato
OP
Joined
16 Mar 2004
Posts
13,474
Location
UK
Once you eliminate the impossible, whatever remains, no matter how improbable, must be the truth.

All I can think of at the moment, new PSU arrived along with CPU so thought I'd give it a go since RAM now done half a dozen memtest86 passes so really didn't look like the culprit.

New (now third) PSU didn't make a difference in the machine so stuck in the identical CPU (9600k) and low and behold, graphics card is recognised. Just to make sure I'm not going mad, I took out the old motherboard I thought was faulty and it also works fine too. Put in the now presumed faulty CPU in there and once again, GPU isn't getting recognised.

I'm just stumped how a processor can be faulty in such a manner, it's still rock solid in any stress test and I can even do light gaming on it. Either something went haywire inside that defaults and forces the iGPU on at all times or it can't communicate with the PCI-E slot anymore for some reason.

I got it second hand and does look like someone tried to lap it so can't even make out any serial numbers, but either way would be SOT trying to RMA it when I'm not the original owner.
 
Soldato
Joined
24 May 2006
Posts
3,828
Location
Surrey - UK
Maybe a bit of gunk/compound on one of the CPU pins, or one ever so slightly bent out of place throwing off the contact when it's in the board? Can't really hurt to inspect and look closely at them now, although why that would be a sudden new issue where it wasn't before is baffling, perhaps there's a minor break(or bad contact) in a pin you can't see (thinking out loud).
 
Soldato
Joined
6 Jun 2008
Posts
11,616
Location
Finland
I got it second hand and does look like someone tried to lap it so can't even make out any serial numbers, but either way would be SOT trying to RMA it when I'm not the original owner.
In that case suspect it might have been fed overvolts or something like that damaging PCIe bus controller.
 
Soldato
OP
Joined
16 Mar 2004
Posts
13,474
Location
UK
Maybe a bit of gunk/compound on one of the CPU pins, or one ever so slightly bent out of place throwing off the contact when it's in the board? Can't really hurt to inspect and look closely at them now, although why that would be a sudden new issue where it wasn't before is baffling, perhaps there's a minor break(or bad contact) in a pin you can't see (thinking out loud).

Actually had a bent CPU earlier this year that was giving me a lot of trouble so one of the first things I checked. All in good nick and these Intel things don't have any pins, those are on the motherboard.

In that case suspect it might have been fed overvolts or something like that damaging PCIe bus controller.

Kinda my theory now too that it got damaged by overclocking by previous owner(s), had the chip since Feb and only got my hands on a GPU last month and like I said it worked fine until all of a sudden it didnt.

Just odd that it still works perfectly otherwise but guess it could be failing at any point in the future now.
 
Man of Honour
Joined
25 Oct 2002
Posts
31,707
Location
Hampshire
Glad you got to the bottom of, I hate scenarios like this where you eliminate the easy stuff like PSU, GPU etc and then have to move on to swapping the mobo, only to find it still doesn't work with a new mobo, so you've just expeded loads of effort without fixing the issue.

CPU would've been the last thing I would have checked as well.
 
Back
Top Bottom