That's easy. . . the NVME drive is drawing ALOT more power than the SD card. I assume that you're powering your Rpi thru the USB-C port on the Rpi? The Rpi-5 is already drawing alot of power and if it dips down below a certain voltage for a millisecond, then "things" can happen. It may not be enough of a brown-out to cause a total reboot of the Rpi, but it may be enough to cause individual chips or sub-systems to reboot.
.
Everything can be working along just fine for hours and then something happens where all 4 cores of the Rpi fire up to 100%, and the Ethernet system is working at 100%, and . . . and. . . brownout.
.
I went thru the exact same issue when I was setting up my Rpi-5/NVME system. The solution I came up with is to power the Rpi directly thru the 5VDC GPIO pins with a power supply that has alot of ass behind it maintain voltage no matter how many chips fire up.
.
I am currently getting ready to test an Rpi Hat that I've integrated a 5VDC@6A power supply feeding directly into the GPIO pins that should be able to handle anything the system throws at it, but that won't be ready for public consumption for a couple more months.
.
In the meantime, order one of these: a.co/d/1iQcW01 and power your Rpi directly to the 5VDC GPIO pins. The Rpi is specced to handle up to 5.2VDC on the GPIO pins, so hopefully you'll get one that gives you 5.1VDC. I've used several of these during my testing and it was the solution to all the little glitches I was having. Solid as a rock.
.

Read More...