GPU failing? Or is it PSU? HEEEEELP!!!!

[sand_man]

Please help me understand what is going on with my PC?!!!!

This is a long and boring story so please bear with me...

So about 2 weeks ago I decided to give my pc a service. My setup, Sabertooth Z77 mobo, i7-3770k cpu, Corsair AX850 psu, 16gb G.skill ram, Noctua DH14 air cooler, Raedon HD7970 gpu, multiple hardrives connected, 6 in total including 256gb ssd as primary. 3 monitor setup 1920x1080. No over clock! Zero gaming!!

So the service entailed blowing all dust out of the machine using wife's hairdryer. Remove DH14, wash dust off radiator grills and fans, re apply thermal paste (MX-2). Remove HD7970, disassemble, wash dust of fans and radiator grills, re apply thermal paste (MX-2). Reassemble and reattach both air-cooler and gpu.

Boot up machine and all is good, temps are excellent, rig is quiet and performing optimally.

Roughly 2 days later the fun starts.

I notice on 2 screens out of 3 there's a single horizontal line running across the screen, the line is around 10mm thick and is moving around the centre of the screen, sometimes accompanied by a smaller brother and has a relative that causes the screen to jump and flicker too. The first time I notice this I wasn't at my desk our even working on my pc, it's quite late at night and I choose to ignore it and by morning everything was running smoothly again and I'm happy (the rig runs 24/7 so I didn't switch it off till morning).

3 days later the lines return while I'm busy on my machine and they won't go away so trouble shooting must begin.

Using the AMD settings software I check for any noticeable anomalies but find none. I decide to use the global overdrive feature to see how the screens behave by underclocking/overclocking. Any deviation from the default settings causes more severe flickering and the line is joined by others. I revert to default settings and it is fairly stable but still evident. Any user activity on the effected screens exasperates the issue.

Time to uninstall and reinstall the AMD drivers and software. I use the AMD clean up utility to uninstall all AMD related software and drivers and reinstall the latest software 18.3.3. Boot up and flickering and lines are gone and haven't been seen since, however.....!!!!!

The flickering has been replaced by a new anomaly.

When I open Google Chrome or iTunes or Photoshop the mouse lags, activity on the pc lags and 4-5 seconds later the apps open and performance returns to normal. It doesn't seem to effect any other programs. Opening IE, Outlook, 12 video files via VLC simultaneously, as many windows as you want, IDM, uTorrent, CCleaner, settings, control panel, Sonos app... no lag or stutter whatsoever, all renders and performs as it should. Once an instance of Chrome is already open, opening additional tabs or opening additional new windows, does so without any issue whatsoever but close all windows and open a new window and it lags for 4-5 seconds before rendering and returning to normal!!

I then decided to configure my 3 monitor setup as a single monitor using AMD eyefinty settings. The lag is gone!!! Hallelujah!!!.... and replaced by tearing.... severe tearing. Just moving a window across the screens, half of the window is still on screen 1 when the other half is on screen 3! Okay, well not that bad but you get the picture!!

So wtf is going on do you think guys?!!

I've tested memory, all good, I've run the rig in safe mode, seems to be fine then, selective mode with startup deactivated = tearing. I ran FurMark gpu benchmark tool but killed it about 2 minutes in cause I don't like to stress test components to that extent but the card and cooling fans were running at optimum. I've adjusted refresh rates (only 2 options 50/60Hz), no difference. I've left the rig off for an hour to eliminate a cooling issue, booted up cold, no difference. CPU temps btw are perfect (monitored with core temp).

I've checked power connectors running to the GPU and PSU, I've checked cables for dust and debris at the connection points and I've done the same with the card and it's pci slot.

Disconnecting a monitor helps, 2-3 seconds lag, disconnecting 2 monitors and everything works as it should! So I suspect a failing PSU?!!! I don't have a spare PSU or GPU lying around with which to test though. I've installed CPUID to monitor but have no idea what I'm looking for re voltage scores.

As a last resort I may do a fresh install of Windows 10-64bit but I don't believe this is a software issue. I've uninstalled/reinstalled the 3 offending programs without success.

Thoughts?? theories?? questions??!

Thanks for listening...

 

[sand_man]

There a lot of things I've tried. As I remember what, I'll update the thread accordingly. There is no BIOS update for my mobo but I did reset to default, although no changes have been made in the BIOS.

 

[CamiKaze]

Does it look anything like this?
https://mybroadband.co.za/vb/showthread.php/936023-GPU-busy-dying?highlight=

 

[ODTech]

There a lot of things I've tried. As I remember what, I'll update the thread accordingly. There is no BIOS update for my mobo but I did reset to default, although no changes have been made in the BIOS.

Artifacts and lines on the screen usualy means gpu memory has a issue.
Did you reasemble it as it was before? Some cards's has heatsinks or those thermal pads on the memory and power components.

If it was power it won't be so specific as lines on the screen unless the specific rail feeding the gpu has a issue. PSU problems usualy shows itself with the pc freezing or bsod's.

 

[sand_man]

Does it look anything like this?
https://mybroadband.co.za/vb/showthread.php/936023-GPU-busy-dying?highlight=

Not nearly as bad. The lines were horizontal vs vertical and less and a LOT thinner!!!

Artifacts and lines on the screen usualy means gpu memory has a issue.
Did you reasemble it as it was before? Some cards's has heatsinks or those thermal pads on the memory and power components.

If it was power it won't be so specific as lines on the screen unless the specific rail feeding the gpu has a issue. PSU problems usualy shows itself with the pc freezing or bsod's.

Yes, should have added that to my OP. I've been looking online for thermal tape. I'm flabbergasted at how difficult it is to source. Sent my SO to Matrix Killarney yesterday but they don't stock and couldn't advise where to find it! Suggested a hardware shop?!!!

I suspected it could be a VRAM issue but I assumed running a cold boot would eliminate a heating issue!! I did think that I may have damaged a chip just enough to cause this issue. Separating the radiator grill/fans from the main board is potentially hazardous. This is the 2nd time I've done it and the first time required a fair amount of effort. The card was reassembled exactly as it should be but those pads are obviously not as adhesive as they were before and one pad had a piece break off, about 10% of the pad. The others minimal, maybe 2-3% lost on account of disassembling!

I need to source thermal tape and reapply. Been trying to find... Didn't see any on Takealot or Wootware...

 

[sand_man]

Does it look anything like this?
https://mybroadband.co.za/vb/showthread.php/936023-GPU-busy-dying?highlight=

NO, taking a closer look at that thread, that looks severe. I don't think it's dying like that card died....

 

[sand_man]

https://www.pricecheck.co.za/offers...ear+Warranty+Product+Overviewarctic+Thermal+P

Okay, found thermal tape, will order and apply..

 

[DMNknight]

You can use thermal paste on the memory heatsinks and apply a >very< small dab of super glue on the outside to keep it in place.

From your symptoms it indicates GPU memory and the damage has probably already been done and probably by the hair dryer.
Ever noticed how hair becomes all floaty and crackles when brushed after being dried that way?
That's called electrostatic discharge.

As a rule, air blasting PCB's should be avoided, as the dust particulates can be forced into the very small spaces between components and provide you with a not so handy short circuit.

Air blasting is fine for heatsinks and fans. Use alcohol and a brush to clean a pcb.
You might get lucky and dislodge/dissolve the particulate matter by using the brush and alcohol method if it's not static electricity that caused the issue.

 

[sand_man]

You can use thermal paste on the memory heatsinks and apply a >very< small dab of super glue on the outside to keep it in place.

From your symptoms it indicates GPU memory and the damage has probably already been done and probably by the hair dryer.
Ever noticed how hair becomes all floaty and crackles when brushed after being dried that way?
That's called electrostatic discharge.

As a rule, air blasting PCB's should be avoided, as the dust particulates can be forced into the very small spaces between components and provide you with a not so handy short circuit.

Air blasting is fine for heatsinks and fans. Use alcohol and a brush to clean a pcb.
You might get lucky and dislodge/dissolve the particulate matter by using the brush and alcohol method if it's not static electricity that caused the issue.

Very informative post! Thanks!

I wouldn't back myself using thermal paste and super glue on the heatsinks...

To be clear I didn't use the hairdryer on the card itself and when using it, never blowing hot.

I did use a brush with rubbing alcohol on the outside of the GPU chip to remove excess thermal paste that had leaked over. This was post issue though at a 2nd attempt at remedying.

Are you suggesting I could repeat the process on the VRAM chips in an attempt at removing particulate matter? Mmmmmm... when I get thermal tape I'll try it...

The static electricity theory is plausible. I was barefoot when working on my rig. Does that count?

 

[DMNknight]

Do as follows:

- apply thermal paste and apply the heatsink
- take a pin and get enough superglue on there to apply a small drop between the chip and heatsink on the outside, where the 2 meet.
- apply on 2 or 4 sides or even on the corners, as you feel comfy.

The drops are so small that they are removed easily when you need to and the heat sinks are not heavy enough to break it either.
There's also not enough to pull pieces of the chip off either

With hair dryers it's not so much the heater but the makeup of it. Anywhere where you get surface friction, you stand a chance of getting static buildup. Moving air with a plastic fan, through a plastic nozzle is likely to invite such electrical activity.

However, be that as it may.

Why I say the symptoms are probably memory, is because the symptoms decrease when you remove monitors and the only resource that really gets knobbed when adding monitors is memory.
The 4-5 second delay when opening certain apps probably mean they are consuming a certain portion of consecutive memory that has a faulty section in it and it needs to move this errored data to another "sector". Smaller applications probably have less issue.
This is doubly backed up by the fact that everything works fine as long as you have those apps open. The issue has already been "bypassed" and any other updates to that worker process work 100%.

It could be that your apps are on a certain hard to read portion of the hard drive? That's also plausible... but doesn't explain the stripes... which memory does.
Other poster mentioned it already, PSU faults usually result in BSODS or sudden reboots, not stripes on screen.
Underpowered GPU will simply shut it down as the motherboard will try and protect itself from being overdrawn (if I understand the why's of it correctly)

Chances are slim, but yes, try and clean it with alcohol. Just google it and you will find a few handy youtube videos about the process.

 

[sand_man]

Awesome post, thanks for the advice!!

Still not comfortable with the super glue solution! Less so know after your detailed explanation!

I concur with your damaged memory theory! Makes a lot of sense now!! I also agree the likelihood of reversing the damage is slim... I'll wait for the thermal tape, clean and apply then...

Look, the machine is definitely usable in its current state, just a little annoying.

 

[animal531]

Sometimes when reseating cards you can get some stray dust/hair in the slots. The same when blowing with air. Remove the card, clean the slot/card and reseat it, you can do the same with your memory.

Since the problem started after you explicitly did something, the cause was most likely you.

 

[Jet-Fighter7700]

Sometimes when reseating cards you can get some stray dust/hair in the slots. The same when blowing with air. Remove the card, clean the slot/card and reseat it, you can do the same with your memory.

Since the problem started after you explicitly did something, the cause was most likely you.

+1 also check the connector itself of the GPU, sometimes the coating becomes a bit brittle and flakes off,
can cause issues with monitors not showing up correctly and weird lines on the screen,

just a suggestion to maybe check those? seeing as you dont do this often, might have peeled more than you intended.

 

[sand_man]

Sometimes when reseating cards you can get some stray dust/hair in the slots. The same when blowing with air. Remove the card, clean the slot/card and reseat it, you can do the same with your memory.

Since the problem started after you explicitly did something, the cause was most likely you.

+1 also check the connector itself of the GPU, sometimes the coating becomes a bit brittle and flakes off,
can cause issues with monitors not showing up correctly and weird lines on the screen,

just a suggestion to maybe check those? seeing as you dont do this often, might have peeled more than you intended.

Yeah, I've done all of the above. Checked the power connectors, checked the fan connectors. Not sure how to deep clean the slot though? Need some compressed air probably.

Ordered thermal tape, just waiting on that and I'll re service the card.

I'm not all that optimistic though, in sorting this out. Have since ordered a GTX1060 from Wootware....

 

[sand_man]

Tried the card in a different PCI slot with the same results so I think it's safe to assume a dirty PCI slot is not the issue....

 

[sand_man]

GTX1060 in and system is purring again... Will have another bash at repairing the 7970 next week and revert...