ALL power supply brands are able to fail, in a bathtub shaped curve of failure over time. I've had to RMA even reputable Seasonic and eVGA units. Some early in life (e.g. a eVGA G2 1000W unit that literally "blew up" with a gunshot-like sound on first power-on complete with smoke...) and some later in life as components age (these can be more insidious in presentation).
The first time I had a PSU failure I was not able to recognize it as the culprit until literally excluding ALL other possible components. It was only noticeable as the GPU not performing as it should. Amusingly, I was running the Heaven benchmark and although it could complete it the GPU performance was well below expectations. It was only after swapping the power supply (Seasonic...) that performance was normal. I RMA'd the power supply and sure enough the replacement unit ran the same GPU just fine.
AMD driver OC tools give you a great way to isolate PSU as the culprit if your system otherwise runs normally in 2D mode or light load:
You can adjust the "Max Frequency" slider and set it to something low like 50% which will force the GPU to use WAY less power. So if your PSU is marginal (e.g. 650W) and was working fine with a less power-hungry GPU that should help you isolate it.
Stock voltage from both AMD and nV tend to be higher than needed with a manual tune due to aggressive GPU boost clock behaviors, so it's rarely the GPU itself in my experience. One exception in my experience was Gigabyte 1070 Xtreme OC cards that had been mined on by a previous owner and were no longer stable at max boost and needed a 50MHz core underclock. But they likely had no margin for stability as a super-binned SKU and I saw reports of non-mined cards also experiencing the same degradation over time.
P.S. Hello to the only other forum user I've seen with a 7900XT