yes it may not directly explained, but you can also see the molex wattage so with it in riser that has 75watts via molex
Apparently you did not understand my question.
I mean, my GPU consumes 120W on its peak consumption. Therefore, I need to power everything up in such a manner that at least 120W can be safely supplied to it.
Let's say, I supplied 75W to the riser via Molex.
How much do I
STILL need to supply to the GPU directly? Is another 75W enough (so the total supply of the riser via Molex and of the GPU via 6-pin is 150W which is more than its peak consumption)? Or do I need to supply at least 120W
directly to the GPU alone and the riser's supply doesn't count?
In other words, is the supply directly from the PSU to the GPU the only one that counts, or does the mediated supply from the PSU to the riser to the GPU count as well and it's the total supply that matters?
I guess we don't understand what you are asking, but I'll try...
You are not in control of Molex power to risers or distribution between molex/PCIE, but I can say it's very little wattage demanded by the riser and PCIE card bus. You absolutely 100% have to use the PCIE 6+2 to power the cards, that's where they get their power from, they won't run without it. I have powered 4 Titan Xp risers from one very thin molex bus and it doesn't get warm. There is definitely not 75w being supplied to each riser. All you need to worry about is total power to the card, through PCIE 6+2 pin power, that's what's giving the card 90% of it's juice. If your PSU has a single 12v rail, you don't have to worry about how it's distributed, just keep it all under 90% of the total PSU capacity, and leave 150W for the mobo/cpu. It won't use that much all the time, but it could spike to that. Remember that cards can spike and some mining programs may push the GPU's harder than others, so try to use aggressive numbers when calculating, not just TDP. I had 4 Titan XP's alone on one mobo that kept tripping out a AXi1200w Corsair. Technically they total consumption should have been 1000w for the GPU's and 100w for the mobo/cpu, but I wasn't counting on the cards exceeding the TDP by almost 50w each on certain algorithms and there was nothing left for the mobo.
An example:
Let's say you have 7 equal GPU cards that have a TDP of say 120w, that means they can peak around 150w, (even if you don't push them you need to leave room). You would safely need 900w to power all six, plus 150w for the mobo. This would mean a good 1200 watt would be the absolute minimum. As long as it has a single 12V rails, you don't care how it's distributed either via the PCIE or the Molex, you just know that each card uses a max peak of 150W total. Now let's say you have two 650W PSU's with a 54 amp rail. You could put 3 cards and the mobo on one 650w psu safely and the remaining 4 cards on another 650w psu. This would leave 50w of headroom to spare on each psu.
I hope this helps.