New build powers on once, then requires 24 pin MB contact to be removed in order to start again

bvhaartman

Junior Member
Oct 25, 2016
3
0
6
Hi,

In an ambition to refresh my gaming PC and hopefully make it quieter I did a rebuild 2 weeks ago. I have run into an issue that I am unable to resolve and are looking for recommendations on how to proceed further. As full background info, I should perhaps mention that the RAM on my old build stopped working "sort of".

Old build:
Lian-Li PC-07 chassis
Corsair AX850 Gold CPU
Asus Sabertooth X58 MB
Core i7 950 CPU running at stock
12GB (3x4GB) Corsair XMS 1600 RAM running @ X.M.P. settings
MSI GTX970 Gaming GPU
Crucial MX-100 512GB SSD
Samsung 1TB 7.2k SATA HDD
Noctua NH-D14 CPU cooler with only the 140mm fan centrally mounted
ASUS ROG SWIFT monitor (2560x1440 GSYNC)

I use the PC for mostly general purpose activities and gaming. Most played titles are not extremely challenging from a graphics perspective: World of Tanks, Fallout 4, World of Warships. During the summer I started experiencing crashes and total freezes especially in World of Tanks, and the failures led to the BIOS falling back to default options (overclocking failed). I tried using "Auto" as default for all settings including memory but this led to powercycling during startup so apparently there was some error reading the SPD of the memory. Using X.M.P everything was normal except for the crashes while gaming. I tried removing and reinstalling the memory but after this Windows only reports 4GB being present whereas BIOS and CPU-Z sees all 3 sticks. I saw this as an excuse to upgrade the PC with the following parts:

New build:
Fractal Design Define R5 with extra 140mm fan
Corsair AX850AX Gold CPU (reuse)
Asus Z170 Pro Gaming ATX MB
Core i5 6600k CPU
32GB (2x16GB) Corsair Vengenace Black LPX DDR4 3200Hz CL16
MSI GTX970 Gaming GPU (reuse)
Crucial MX-100 512GB SSD (reuse)
Samsung 1TB 7.2k SATA HDD (reuse)
Noctua NH-U12 CPU cooler
Same monitor as above

I assembled everything and the PC booted to BIOS normally, I left all settings on auto and proceeded to do a repair of the existing Win 10 install on the SSD. This proceeded normally with the usual 2-3 reboots along the way. I porceeded to reinstall the SW I had lost during the Windows repair, reinstalling Office, restoring backups in outlook etc. Redirected My Documents, Pictures, Music, Movies etc to the D:/ (HDD) drive. The evening was running late so after this I shut down the PC. The next morning it was completely dead, I thought I might have a bad connection somewhere so I unplugged and replugged everything on both PSU and MB side, and the PC started normally, tried shutting down and again no response. Cleared CMOS, removed battery and power for 15 mins, no change. Isolated which cable would need to be unplugged to it being the 24 pin ATX. Disassembled PC and tried breadboarding with the following findings:

With no GPU or storage, touching the pins for the power button causes the PC to boot and MB speaker issues one short beep (everything OK) touching the pins for 3 seconds causes a shutdown. A new short touch of the pins makes the PC boot up again with an OK from the MB speaker. I can repeat this several times with no issue.

Adding the GPU I begin to have some issues PC sometime boots, sometimes not, I can resolve this various ways, changing the PCI-E cables from one connector to another on the PSU worked once, switching which PCI-E cable uses 8 and which uses 6 pins to the GPU works sometimes but not consistently. I have since reassembled the PC but feel I'm still on square one. PC boots once but requires pulling the ATX 24-pin to be able to restart.

Some other observations:

As I have the sidepanels off, i can just barely hear a quiet clicking noise from the PSU when the PC runs.
The fan on the PSU runs after shut down (not sure for how long...), I guess this is normal.
It is impossible to wake the PC from sleep mode, power button blinks but just goes off if I push it, no response to any keyboard input...
If I manage to restart by pulling the ATX plug, the ethernet port fails to be able to wake my Powerline "wall wart" from its powersaving mode.
I have run 1hr of OCCT standard + graphics load test to see if this would show any crashes etc but everything completed normally.

Any and all suggestions on how to proceed from here would be extremely welcome!
 

bvhaartman

Junior Member
Oct 25, 2016
3
0
6
Hi,

Thanks for the tips. Actually just now I thought I'd try replacing the PCI-E cable as it had some kinks and twists from the previous build. That did not help, however I found my old GTX470 GPU and thought I'd give that a try. With that installed powercycling works fine. I can't check if "everything" works as my monitor only has DP and the 470 dvi and mini hdmi and I have no adapters. I'll get an adapter tomorrow and verify the 470 is causing no problems. If so I guess I have a good reason to get a new GPU. PSU seems not to be the root cause. Maybe the freezes in the old build were also GPU related. Sad to see a GPU give up after just over 2 years. It was my first MSI product so my image of the brand will be tainted.

Thanks again!
 

UsandThem

Elite Member
May 4, 2000
16,068
7,380
146
Glad to see it seems you narrowed down the issue.

I know it sucks when you have a component die early, but MSI is one of the better manufacturers of video cards. All brands, even EVGA, will have cards die sooner than they should, as they are complex pieces of technology. Once a little time passes, the irritation won't be as fresh as it right now. How do I know? I said the same thing back in the 90's after having a dud motherboard from MSI. However, after being on here for many, many years, I have seen that you will have duds from every manufacturer.....even Intel at times.

What separates a good company from a bad one is how they take care of you when this happens. Luckily, most companies offer a 3-year warranty on video cards in the U.S. You'll have to pay to send it to them, and it might not be overly fast, but they should fix your card or send you another one.

Good luck!
 
Last edited:

bvhaartman

Junior Member
Oct 25, 2016
3
0
6
Soooo... Thinking my GTX970 had given up the ghost, I bought a new GTX 1070. Once installed the PC displays exactly the same issues as before. It turns out the problem is the Display Port cable which apparently was an uncertified version sending back +5V on pin 20. Curious that the X58 platform would still boot and run whereas the Z170 will not even POST. Will test the GTX 970 in another computer this weekend since I did not feel like tearing apart my new build now that I finally have it running. Would the non certified DP cable have damaged the GPU?