No problem, it's not an immediately obvious thing (especially with efficiency being presented as a percentage value), the only reason I know this at all is from having it repeated in a bunch of PSU reviews I've read (and seeing them test at 100% of the rated power). It is indeed a bit odd that if OCP was triggered given the 400W rating, but there are many possible culprits - there could be voltage droop under high transient loads, an overly sensitive OCP setting, tiny transient loads that significantly exceed what is registered by software, or many other factors causing crashes. Recent Intel CPUs can consume
a lot of power during short-term boost, and GPU Boost 4.0 can also produce rather spiky power loads.
Also, I don't quite know if HDPlex accounts for conversion losses in the DC-ATX unit in their 400W rating - they are two separate units after all. 12V DC-DC PSU solutions (PicoPSU, J-Hack, G-Unique) sidestep this through producing 12V directly in the AC-DC unit, but HDPlex and their 19V setups need further conversion to 12V, so there's an inherent extra layer of conversion losses there, unlike pretty much all other PSU setups.
Actually, looking over the specs, this is likely the issue - and in a fun turn of events, this is perhaps the only PSU setup where your way of calculating things is actually correct! A great facepalm moment for me, at that
The HDPlex 400W AC-DC is rated for 19V 21A (=399W), at 94% efficiency (so ~25W heat load in addition to the 399W output), but that's just a plain 19V output. In addition to that the HDPlex 400W DC-ATX is itself rated for "400W (500W Peak) >94% DC-DC Efficiency". Which, unlike other PSUs,
must be subtracted from the output of the AC-DC converter, as that can only output up to 400W total for the input power of the DC-ATX. So with both chained together you are indeed looking at a ~400*0.94=376W peak output to the PC. The only way for HDPlex to avoid this with their two-stage conversion would be to actually rate the AC-DC unit for ~425W output. I would frankly have expected them to do so, but ... apparently not. I guess it's not technically misleading, given that both products can indeed deliver their ratings, they just can't quite do so when paired together. Which, being the most obvious implementation, is a bit dodgy.
In your use case things like motherboard and RAM power draw, as well as CPU VRM losses probably also counted for something. Motherboards and RAM don't consume much power - 15-25W is a reasonable estimate for a modern ITX board with two sticks of RAM at normal voltages - but CPU VRM losses can amount to a decent bit with a boost-heavy Intel CPU. CPU VRMs have ~94-95% efficiency at the high end, so for a ~160W draw registered in software (which might very well be higher in real life, at least for a short spike) that's ~another ~10W of power drawn from the PSU. And at this range, every little bit counts, of course. Taking that into account, that places you very close to 400W total power draw if the 3060 Ti is running
between its measured 214W (average) and 224W (peak) power draw in-game, and definitely above the ~376W the HDPlex DC-ATX can output with a 400W max input.
