Troubleshooting my 1080, broken in my main but works on my 2nd rig.

[DarkBrews]

Hello,

 

I am looking for some guidance on how to better troubleshoot a dying or faulty GPU.

 

Mine has a modified water block but there are no traces of leaks, and even if it did it would leak on the shroud of the video card and not on the PCB;  will post pictures if needed but the block is pretty isolated from the main PCB.

 

The Card had become unstable far after I mounted the water cooler but when the computer hard reset the first time my speakers stopped working all of the sudden as well.

After some repositioning and testing with my 2nd rig both the GPU and the speakers started working again but it would hard reset in certain games.

So I tested my main rig with 2 other video cards and the GTX 1080 would a 2nd system and no crash after 7-8 hours of OCCT, Furmark,3D Mark, Battlefield (which was crashy on my main).

Even overclocked it to +200core +500vram, no crashes however started flashing some black frames from time to time. Assuming it's an HDMI issue I lowered the colored depth from 12(apparently!?) to 8 with 4:2:2 chroma and it works perfect.

 

Main rig has an i7 6700k with a Cooler Master G750M which was replaced by an EVGA 750G2

My secondary rig i7 3930k and a Corsair budget 600W power supply which I tested the 1080 with

 

I also tested my main rig with an RX 480 and GTX 970 and it worked perfect

 

So I suspected the power supply and swapped it and the main rig was working flawlessly again....until 4 months after the speakers stopped working again and I pretty much went through whole thing again. Same results

 

 

By now I am about 80% sure that the GTX 1080 is dying but I think the power supply would also make sense as well giving the speaker coincidence and the fact that it worked with a new power supply for a while, are my speakers PSU killers?

 

[DarkBrews]

Update: purchased a 1080ti to replace my 1080, but the 1080 is as stable as a rock now in my 2nd rig  

[Bombastinator]

My guess is possibly neither.  A power supply swap does a few things besides changing the power supply.  It kills power to the system, it forces you to open the case and move things around.  It forces reattachment of connectors, it causes the system to cool down while the thing is changed.  There is also mains power to consider.  Are you using a UPS?

[DarkBrews]

4 hours ago, Bombastinator said:

My guess is possibly neither.  A power supply swap does a few things besides changing the power supply.  It kills power to the system, it forces you to open the case and move things around.  It forces reattachment of connectors, it causes the system to cool down while the thing is changed.  There is also mains power to consider.  Are you using a UPS?

I am not using a UPS, what are your thoughts? I have stress tested the 1080 in the 2ndary system for 16 hours, no reset, no artefacts, nothing

[Bombastinator]

1 hour ago, DarkBrews said:

I am not using a UPS, what are your thoughts? I have stress tested the 1080 in the 2ndary system for 16 hours, no reset, no artefacts, nothing

I’m going to have to retract.  

I don’t know much about 1080s specifically.  It’s not like the process was trouble free, though you did get it working.  My opinion may not be worth much. That 12bit-8bit color thing throws me right into very deep water.  I apparently missed that part in the original post or I would haven’t said anything.  Bad reading comprehension on my part.

 

Why the 12-8 thing makes stuff weird:

Back in the day, the standard were 8, 16, and 24 bit color.  Using 1, 2, or 3 sets of 8 bit numbers.  They were sometimes referred to as “hundreds, thousands, and millions” of colors.

Pro grade cards could use 10 bit numbers, making a lot more available colors, (and quadrupling prices) but it was a software thing and the pro grade cards were effectively the same hardware as the consumer grade cards.  They were referred to as 8 or 10 bit.  There was no 12 bit.  If 12 bit is the new 10 bit what you did is set it to consumer grade from pro grade.  One didn’t used to be able to even do that on anything but pro cards, so something impossible to my clearly limited understanding apparently happened.  It could also be though that you set the thing to 256 color (hundreds).  If so the screen won’t be able to show any picture that doesn’t look like it came off an old old console game.  If you did that your stability test is not a stability test.

 

What I was thinking:

given that my opinion on the 1080 is near useless, if we assume the 1080 is good, which I can’t do now that I’ve read everything more correctly, what I was wondering about was heat and marginal connections.  Heat can do things to marginal solder points. So can movement.  There is also possible connector issues.  Reattaching connectors can improve marginal connections at those connector point if there was one.

 

i asked about the UPS because having one would eliminate mains power spikes as a potential damage source.

[DarkBrews]

1 hour ago, Bombastinator said:

I’m going to have to retract.  

I don’t know much about 1080s specifically.  It’s not like the process was trouble free, though you did get it working.  My opinion may not be worth much. That 12bit-8bit color thing throws me right into very deep water.  I apparently missed that part in the original post or I would haven’t said anything.  Bad reading comprehension on my part.

 

Why the 12-8 thing makes stuff weird:

Back in the day, the standard were 8, 16, and 24 bit color.  Using 1, 2, or 3 sets of 8 bit numbers.  They were sometimes referred to as “hundreds, thousands, and millions” of colors.

Pro grade cards could use 10 bit numbers, making a lot more available colors, (and quadrupling prices) but it was a software thing and the pro grade cards were effectively the same hardware as the consumer grade cards.  They were referred to as 8 or 10 bit.  There was no 12 bit.  If 12 bit is the new 10 bit what you did is set it to consumer grade from pro grade.  One didn’t used to be able to even do that on anything but pro cards, so something impossible to my clearly limited understanding apparently happened.  It could also be though that you set the thing to 256 color (hundreds).  If so the screen won’t be able to show any picture that doesn’t look like it came off an old old console game.  If you did that your stability test is not a stability test.

 

What I was thinking:

given that my opinion on the 1080 is near useless, if we assume the 1080 is good, which I can’t do now that I’ve read everything more correctly, what I was wondering about was heat and marginal connections.  Heat can do things to marginal solder points. So can movement.  There is also possible connector issues.  Reattaching connectors can improve marginal connections at those connector point if there was one.

 

i asked about the UPS because having one would eliminate mains power spikes as a potential damage source.

Awesome explanation, I agree the 12 bpc was a weird thing but the TV I was testing with is cheap and weird so I thought it's because of that. Do you have a recommendation on how to confirm if this is the GPU is dying?

Heat is not really an issue, the video card tops at 65 C but most of the times I have a ventilator blowing on the thing and it keeps in the low 50s.

[DarkBrews]

I reproduced the flicker with another video card on the TV and also I forgot to mention that initially the board had a lot of bent pins when I got it, for free :3. So I am thinking that it might actually be the bent pins as I seem to sometimes reproduce the crash wiggling the cooler while a stress test is running. Any thoughts? The CPU is 6700k

[DarkBrews]

Wiggled the pins on my board a bit more, no more crashes.

At the same time my computer would not work if set horizontally but that is fine for now until I find a convenient replacement.

 

offtopic: Also bought a GTX 1080ti Aorus from ebay and it turns out that was broken in a very weird way... It had performances similar enough to a 1080, 10more FPS on average but 5 or 10 lower minimum frame rates which made me check the videocard with GPU-Z.

 

As a result the GPU only connected to 2x PCIe lanes, this I checked on 3 different systems (actually on one it connected to x8 at some point).

I return the 1080ti but it's so easy to miss on checking something like this as it had no other problems than low minimums.

[Bombastinator]

1 hour ago, DarkBrews said:

Wiggled the pins on my board a bit more, no more crashes.

At the same time my computer would not work if set horizontally but that is fine for now until I find a convenient replacement.

 

offtopic: Also bought a GTX 1080ti Aorus from ebay and it turns out that was broken in a very weird way... It had performances similar enough to a 1080, 10more FPS on average but 5 or 10 lower minimum frame rates which made me check the videocard with GPU-Z.

 

As a result the GPU only connected to 2x PCIe lanes, this I checked on 3 different systems (actually on one it connected to x8 at some point).

I return the 1080ti but it's so easy to miss on checking something like this as it had no other problems than low minimums.

Considering the strange behavior of the hardware it’s not impossible the issue is with your board and not the card.  The whole wiggling bent pins and not able to work flat thing imply a serious lack of good health.