Question Is my GPU dying?

[Sad_Face]

Hello,

I got my MSI 4070Ti Ventus 3x in January and everything was perfect. But in June I had my first system lock - PC was idle on desktop (I wasn't around), screen went black without any reason and fans started to spin at max. Restarted my system and everything was fine for month and then it started. Sometimes I had two locks in hour while playing (it doesn't matter if it was some AAA RTX title or small abandonware game that required 8MB VRAM). I've removed GPU from case, reconnected all cables making sure that they're locked tight, removed GPU drivers using removal tool and installed new ones as admin and set power usage to max performance, used OCCT GPU stress test for 70 minutes. No issues for another month. But it started all over again. Random locks, sometimes fans are at max, sometimes not, sometimes system seems to work (disk led is blinking), but with black screen. Once after lock my system started with black screen and some blue -ish dots on it ( https://freeimage.host/i/JHZ9GhG I can't insert image from link...). I had two locks in 10 minutes today on idle, now my PC is running fine since three hours. Should I RMA GPU or am I missing something that cousing problems? I've ordered custom 16 pin connector just in case.

My specs:
Mobo: Gigabyte B550 Aorus Elite V2 (bios version F13g)
CPU: Ryzen 5800X - no OC
PSU: Be Quiet Straight Power 11 850W
GPU: MSI RTX4070Ti Ventus 3X
Windows 10 64-bit

Regards

 

[lojikal]

Please confirm which fans were spinning at, "MAX". How do you know it's the GPU fans?

 

[Sad_Face]

Please confirm which fans were spinning at, "MAX". How do you know it's the GPU fans?

GPU fans. My CPU fan is Be Quiet and it's... queit even on high rpm and with case fans I would feel airflow if they were running full speed. GPU is loudest thing on my system even if it's going half rpm or less.

 

[seogoat]

Lock ups can be a faulty ram module. If you have two sticks of ram try testing one at a time, at least to rule that out.

 

[Sad_Face]

Lock ups can be a faulty ram module. If you have two sticks of ram try testing one at a time, at least to rule that out.

Tbh I would rather replace my ram modules (70$) with new ones than RMA my GPU...
I'm using Patriot Viper 4 3200 MHz 2x16GB, have them since 3 or more years. Sometimes I have two locks in an hour and sometimes no locks in two weeks or month. So it's total lottery and I can get 'lucky' when running on one stick.
Can you suggest some RAM testing tool that I could leave working on loop for few hours?

 

[Avro Arrow]

I can get 'lucky' when running on one stick.

That's what she said!

(Sorry, I couldn't resist! 🤣)

 

[Order 66]

Memtest86 will take at least a few hours to run depending upon how much memory you have.

 

[Sad_Face]

Memtest86 will take at least a few hours to run depending upon how much memory you have.

Tested on two sticks:

IMG 20230919 221418

Image IMG 20230919 221418 hosted in Freeimage.host

freeimage.host

IMG 20230919 221433

Image IMG 20230919 221433 hosted in Freeimage.host

freeimage.host

I guess all I can do now is to change original 16 pin connector to custom one and if locks will still occur I will use my old GTX 970. If I will not have any locks until November I will RMA my 4070Ti.

I'm still open for suggestions tho.

 

[seogoat]

Still may be intermittent ram problem. Run with 1 stick until problem shows itself, then switch to other stick until problem shows up. if problem occurs on both sticks separately then you can rule out the ram. If problem never happens on other stick then it's ram. Memory can still be unstable and not show any errors on memtest.

Also like you said if you switch back to the old gpu and it's fine then it does point to that new gpu causing a problem.

 

[Sad_Face]

Still may be intermittent ram problem. Run with 1 stick until problem shows itself, then switch to other stick until problem shows up. if problem occurs on both sticks separately then you can rule out the ram. If problem never happens on other stick then it's ram. Memory can still be unstable and not show any errors on memtest.

Also like you said if you switch back to the old gpu and it's fine then it does point to that new gpu causing a problem.

I'll go with custom cable for few day and change my GPU, it's easier this way I think. Main problem is that those lock don't occur in same situations or daily so I need longer tests. I'll will use GTX 970 until mid October or until first lock.

 

[Sad_Face]

It's like 30 or 40 days after changing original 16 pin to some $30 Amazon one. I've done RE4: Separate Ways in one go, few older games, few total retro on monitor and TV, movies and one 26 straight download run. Not even single reboot or system lock.

I think I found my issue or at least it looks like it.