[SOLVED] Is my GPU dying (or actually 95% dead)? Fail to boot with VGA debug led.

[WeskerEnd]

tl;dr - Motherboard sometimes shows boot error with VGA led (debug LED), and I need to keep power cycling until PC boots properly, and it all seems fine on Windows when it does. Is my GPU dying for sure or could it be something else?

This started happening a few days ago, sometimes I power on my PC, fans/ram/leds/HDs all turn on, monitor detects a signal but stays black (as if it was receiving a full black screen). If I press Power it just turns off instantly, and today I remembered my mobo has these debug LEDs, and I saw that the VGA led is lit when this happens.

I need to keep power cycling a few times until it turns on as expected. When I'm in Windows, EVERYTHING* is perfect. No crashes, my VGA seems fine, minutes of Furmark with no problems *(while I was typing this, minutes after running the tests windows/explorer became weird, Task Manager showed -1% GPU usage, I tried restarting and got the error many times before booting again). As far as this afternoon I was able to game without problems, though it wasn't a very intensive game.

I tried reseating RAM and GPU, and my PCI-E power connector has 12.2V, I checked each black against each yellow, so I guess psu isn't the problem.

My PC (year bought):
-MSI x470 gaming pro (2019)
-Ryzen 2700x (2019)
-EVGA GTX 1060 6GB SC (2017)
-2x8GB @ 3200MHz loki asgard (2019, chinese ram, "cheap")
-corsair CX430 (2015)

Is there any hope or am I doomed because I need a GPU in the current market?

 

[Lutfij]

You have a green labelled CX unit from Corsair, that automagically entails that your issue can be with the PSU...since that shouldn't be considered a PSU in the first place. Have you tried sourcing a donor PSU from a friend or neighbor who owns a reliably built 550W~650W PSU?

 

[WeskerEnd]

You have a green labelled CX unit from Corsair, that automagically entails that your issue can be with the PSU...since that shouldn't be considered a PSU in the first place. Have you tried sourcing a donor PSU from a friend or neighbor who owns a reliably built 550W~650W PSU?

There's no one I know of to do that, but I could look into buying a new one. Didn't think it could be it.

A new PSU would be so much cheaper and useful, I just needed to get a spare one to check, I'll see if I can do something.

Thank you!

 

[WeskerEnd]

Hopefully it's my PSU, I bought a new one, it should be delivered before next Friday and I'll update when I can test it or if anything happens until then. Thank you.

Spoiler: New PSU

 

[Vic 40]

What did you buy?

 

[WeskerEnd]

What did you buy?

XPG Core Reactor, the 850w one because honestly the price difference from the 750w was nothing.

I know I don't need even half this, but it has a 10-year warranty.

 

[Vic 40]

Good choice, just was curious.

 

[WeskerEnd]

since that shouldn't be considered a PSU in the first place

Ok, ouch!

You have a green labelled CX unit from Corsair, that automagically entails that your issue can be with the PSU...since that shouldn't be considered a PSU in the first place. Have you tried sourcing a donor PSU from a friend or neighbor who owns a reliably built 550W~650W PSU?

My new PSU arrived this afternoon, took my time installing it, first boot everything fine so hopefully that was it, just like you said. I'm just happy that it wasn't my GPU, hopefully it will last for as long as it's needed for prices to come down.

Thank you very much, happy holidays and happy new year in advance!

EDIT:
3 photos for posterity:

Spoiler: PSU label rails specs

Spoiler: a lot of cables, nice pouch

Spoiler: strong protective foam

 

[WeskerEnd]

I guess it was the GPU after all... Computer became unusable these last few days. Sudden restarts, MANY BSoD's, to the point I was scared my CPU/MOBO were also dead.

Today I was able to purchase an old and used GeForce 7100 GS, and it's been working flawlessly, no errors during boot or windows, no restarts, nothing weird.

Spoiler: GeForce 7100 GS

At idle it was at 80ºC, cleaned it, changed thermal paste, used zip ties to put a fan on it, the maximum I've seen since then was 61ºC.

Spoiler: Simple fan mod with zipties

I can't even open Epic Store Launcher with it, but hey, at least I can browse and stuff. As long as the system remains stable I'm happy.

EDIT: Huh, system suddenly turned off. Booted fine, everything seems normal.

 

[WeskerEnd]

My GPU is fine, my friend used it for the whole week without problems. I bought a Ryzen 5600G for the integrated graphics and after playing games with it for 1 hour or so, I got a sudden restart, W10 log says it was a power loss.

This can only mean my mobo is bad, I guess?

 

[Vic 40]

Would like you to do this so we can be sure you don't have a BSOD,

https://www.tomshardware.com/how-to/windows-disable-restart-after-bsod

Still looking at that somewhat unknown ram you use. Otherwise likely the motherboard yes.

 

[WeskerEnd]

I got another SSD and installed W10 on it. Only drivers and Steam installed to make sure it wasn't windows or drivers or anything like that. After playing Rocket League for a while, got this graphics error. Then again after rebooting and trying again.

Spoiler: 2 graphics errors

I removed the second RAM stick and it didn't happen for the couple minutes I tried to repeat it. Changed the sticks, still didn't happen.

Later that day my friend brought back my GTX 1060. Put back my old SSD, kept only 1 stick of RAM, and everything went fine, no problems at all.
Today I swapped the RAM sticks while still using the same slot, so far no problems.

If nothing changes, tomorrow I'll put this stick into the second RAM slot.

This is so confusing... at least it's been working fine now. I'm willing to buy another RAM kit, just wanted to find some signs that it is really bad. Individually, right now, at slot 1, they both seem perfect.

 

[Vic 40]

Ever did a memtest86 on the modules? Do each on it's own, if fine do it with both together. Makes sure you do a minimum of 4 passes.
MemTest86 - Official Site of the x86 Memory Testing Tool

The above looks like a gpu failure btw, maybe memory on that is flawed. But wierd that it works with just only one stick of ram.

If memtest gives faults with either a single stick or both together is it best to get some well known ram.

 

[WeskerEnd]

Ever did a memtest86 on the modules? Do each on it's own, if fine do it with both together. Makes sure you do a minimum of 4 passes.
MemTest86 - Official Site of the x86 Memory Testing Tool

The above looks like a gpu failure btw, maybe memory on that is flawed. But wierd that it works with just only one stick of ram.

If memtest gives faults with either a single stick or both together is it best to get some well known ram.

Yes, I did 7 passes of memtest86 (on the free version you can only do a maximum of 4 at once). Also did 3 passes of Windows' built-in memory test with the extended option, no errors at all, all of them done with both modules installed, didn't try one at a time.

That GPU failure was using the iGPU of the new 5600G, which I know would also indicate a memory problem.

Today I'm still testing using only 1 stick of RAM, but on the second slot, since I didn't get any problems with either sticks at the first slot.

If all goes well, tomorrow I'll install both of them and see if dual channel triggers anything. So far, everything seems fine. I don't have the strength nor willpower to install my 2700x and see if the problems come back, I'm exhausted when it comes to my PC, just want things to work.

 

[Vic 40]

Noone you can lend ram from?

 

[WeskerEnd]

Noone you can lend ram from?

The same friend who tested my 1060. He doesn't have spares, so it would stop him from using his PC.

I'm now using both sticks, if nothing happens I'll just stop testing stuff. Things are working and that's what I want, more than actually finding out what was wrong.

 

[Vic 40]

He could use/test with your sticks.

 

[WeskerEnd]

After days without problems, I got 3 crashes yesterday with a freeze frame. Happened once during gaming, restarted manually, freeze frame crash during W10 login, restarted manually, freeze frame while browsing event viewer. Gave up, came back later, opened it up, didn't see anything unusual, turned it on and was able to play games for a long time again.

Anyway, I just ordered a new RAM kit, these ones... if RAM is not to blame, only mobo will be left because everything else has been changed/tested, basically this is a new PC.

Spoiler: New RAM screenshot for posterity

It sucks, but all these purchases combined wouldn't get me a new GPU if mine had died, so it could've been worse. While a new simple mobo is cheaper than this RAM I've ordered, I need optical audio output to use my headset, and there isn't a cheap mobo with it. Hopefully the RAM is all I need.

I'll update after I receive it and test it extensively.

 

[Vic 40]

Hope the new ram will fix it.

 

[WeskerEnd]

Still waiting for my new RAM, but wanted to add something. I had no problems at all yesterday (Friday) which is the day I most play games on my PC while listening to music, chatting on Discord and so on.

Today I got the "freeze frame" crash again, twice so far... this time I noticed a high pitch whine coming from somewhere inside my case as soon as it crashed. I can't say for sure I've ever heard coil whine, but I'd say it sounded very similar, maybe with a little lower pitch note.

Does this help or mean anything? I'm back to using 1 stick of RAM to see if it will happen from now 'till my new RAM is delivered.

 

[Vic 40]

Would be helpfull if you could figure out where the whine comes from. Could be gpu or psu so narrowing it down would help.

 

[WeskerEnd]

Got my new RAM today. Because the clean W10 install for the iGPU was giving crashes regularly, I removed my GPU and tried that again. Everything's perfect! Played Rocket League and Fortnite for 2+ hours, no errors/crashes. Also got a big boost to FPS, this new RAM is rated 3600MHz, the old one was rated 3200MHz, I'm pretty sure this boost is not because of those extra 400MT/s. These 2 games are perfectly playable 1080p/60FPS with decent quality and no need for resolution scaling on the 5600g.

Now I'll go back to my regular NVMe SSD and just use it as always. Hopefully that's the end of it and I don't get any crashes/freezes.

Never buying chinese RAM again, that's for sure.

EDIT: Just to make sure, I'm using 1:1 RAM/fClock right? Uncore frequency and DRAM frequency?

Spoiler: CPU-Z Screenshot

Spoiler: New RAM box front and back

 

[joan16v]

I agree that the problem maybe is the PSU, try another of more Watts. If not, it may be the GPU, maybe time to put it on the oven (not joking).

 

[Vic 40]

Never buying chinese RAM again, that's for sure.

That is a good idea.😁

EDIT: Just to make sure, I'm using 1:1 RAM/fClock right? Uncore frequency and DRAM frequency?

Spoiler: CPU-Z Screenshot

Yes you do. Still at 2T which may impact performance a little, but is likely because of the 3800mhz speed. Lowering that to 1T might need lowering of ram speed as well and tightening timings. Maybe just adding voltage might help, but for all intents and purposes does your ram run at rated and good speed.

 

[WeskerEnd]

It's been a whole month with the new RAM kit, zero problems. I still can't believe bad RAM could be this troublesome, I always expected clear indicators when it did go bad, and considered RAM to be the simplest/cheapest/most durable component... oh boy how the tables have turned.

If anyone reads this be warned, do not buy unknown RAM; if you do, make sure to blame it for ANY errors/problems you get FIRST, and worry about the rest after making sure a new RAM doesn't fix it. GPU "died"? Try a new RAM. Seriously. "Been there done that".

My bad RAM had no errors at all through many and many test passes (Windows and memtest86), made the motherboard behave as if I had a bad GPU (errors/drivers made it look like so, plus the motherboard debug error led) and CPU (I bent 2 pins more than one year ago so it was believable to me, also showed debug error led), caused random BSoD, caused my PC to suddenly restart or shutdown (which made me buy a new PSU).

Thank you for this amazing forum, for your time and knowledge helping me through this. You truly made it easier. Special thanks to @Vic 40 for not giving up on me and insisting on the RAM, you knew it all along.

This is my final update to let anyone know that it's finally solved for real. My DM is open to any questions.