Question NVMe SSD Randomly Disconnects During Heavy Load (Games / Stress Test)

[sojoya]

Hey all,

I need help diagnosing a persistent issue with my PC. The problem is that my NVMe SSD (low budget SSD called RX7 1TB) randomly disappears from the system during heavy load — usually when launching or a few minutes after playing demanding games stored on the D: drive, or when running FurMark + CrystalDiskMark together to imitate the worst gaming case scenario. When this happens, Windows either crashes with BSODs like "Kernel Inpage Error" or "Unexpected Store Exception", or it just freeze. The drive reappears after powering off for a while. If I don't play any games on that D drive the system works flawlessly for example just browsing or very light game.

System Specs:​

• CPU: Ryzen 5 5600
• GPU: MSI RX 6700 XT (12GB)
• Motherboard: Gigabyte AB350M Gaming 3 (latest BIOS, Version F53g) just updated the bios and the issue persists
• RAM: Adata XPG Gammix 32GB DDR4 3200MHz
• Boot Drive: SATA SSD 128 GB (Windows installed here)
• Game Drive: RX7 NVMe 1TB (D Drive
• PSU: Corsair TX750M Gold
• Case: Gamemax Spark M-ATX

Evidence & Testing:​

• Ran OCCT Combined test (CPU + RAM, GPU Adaptive, and VRAM) also trigger the same issue
• Problem only happens when GPU + NVMe are both under load. This doesn't happen if I run Furmark or CrystalDiskMark separately
• HWiNFO logs show +3.3V rail dips to 3.06–3.10V right before drive disappears. Under idle/load-light scenarios, 3.3V stays above 3.20V.
• Windows Event Viewerlogs confirm repeated:
  • stornvme Event ID 129 – Reset to device \Device\RaidPort1
  • disk Event ID 51 – Paging operation failure on D :
  • ntfs Event ID 50/140 – Delayed write failures
• BIOS NVMe self-test passes; no SMART issues.
• Undervolting GPU (to 1110mV, power limit -6%) reduces total power but doesn't stop crash during stress.
• Pure CrystalDiskMark or real gaming (UE5 title) doesn’t trigger crash if GPU is lightly loaded.
• The weird thing is, the drive is stuck at 40 degree celsius temperature no matter how intensive the work is, suspecting something wrong with the sensor or controller

What I suspect:​

• +3.3V rail instability under full system load is causing NVMe controller to lose link.
• Could be PSU sag, or motherboard's 3.3V rail delivery to the M.2 slot is weak.

Looking for advice on:​

• Is there a way to make sure this is only the NVME drive problem
• Or is this a PSU problem or motherboard issue?
• Is using a SATA SSD for games instead a safe long-term workaround?
• Any other BIOS tweaks I can try (C-states, spread spectrum, etc.)?

Thanks — I can provide HWiNFO CSV and Event Viewer XML if needed.

(Edit: Additional spec info)

 

[Lutfij]

(low budget SSD called RX7 1TB)
I think that's your issue. To add, Adata never made an SSD with teh RX7 naming scheme.

The drive reappears after powering off for a while.
This would indicate a thermal issue. Perhaps the heat from the GPU on the SSD is causing it's performance to degrade. Make and model of your case and the temps for your SSD when the issue crops up?

Motherboard: Gigabyte AB350M Gaming 3 (latest BIOS)
For the sake of relevance, please state the BIOS version for your motherboard.

PSU: Corsair TX750M Gold
How old is the PSU?

 

[sojoya]

(low budget SSD called RX7 1TB)
I think that's your issue. To add, Adata never made an SSD with teh RX7 naming scheme.

The drive reappears after powering off for a while.
This would indicate a thermal issue. Perhaps the heat from the GPU on the SSD is causing it's performance to degrade. Make and model of your case and the temps for your SSD when the issue crops up?

Motherboard: Gigabyte AB350M Gaming 3 (latest BIOS)
For the sake of relevance, please state the BIOS version for your motherboard.

PSU: Corsair TX750M Gold
How old is the PSU?

Thanks for the swift reply, yep it's not an Adata SSD as you mentioned, I've added the additional specs

- Case is Gamemax Spark M-ATX
- SSD Temp doesn't change at all and dead stuck at 40 degree celsius (on HDSentinel, CrystalDisk, HWInfo, etc)
- BIOS, is Version F53g just updated the bios from F51h last night and the issue persists
- The heat assumption should affect when I tested FurMark only without any drive activity as the M.2 sits directly below the GPU, but the drive was still detected even when GPU reach 86 degree celsius
- PSU is an old one, I think it's from 2017 or so

 

[LordVile]

Does the drive have a heat sink?

 

[sojoya]

Does the drive have a heat sink?

Can confirm, it does not have a heatsink

 

[JayGau]

Probably the cheap drive's controller can't sustain the heat (from itself or other components like the GPU) and fails when gaming.

Drives are probably the worst thing to cheap out after the PSU. There are so many horrible SSDs out there that cost peanuts but can't run anything without crashing, or fail after a few months. You should back up everything you want to keep from this drive and buy a decent one. A good 1 TB NVMe from a recognized brand (Samsung, WD, Crucial) cost about 60-70$ USD. There's no reason to go with the crap.