Well, I did it.
I decided to go ahead and test the GPU with a PSU that is clearly rated too low. Now, keep in mind: The purpose of this test is not to identify whether or not a given GPU would run with a given wattage PSU. Nope - this test is to demonstrate that "Just because
it works, doesn't make it a good idea".
I picked a PSU from my stuff that has all the necessary connectors (CPU 4-pin 12v, motherboard 24-pin, etc) but has no PCIe connectors whatsoever. I intentionally chose the crappiest PSU I could find in the lot: A 300W unit that weighs so little it probably couldn't even actually drive 300 watts for any length of time, and even according to it's own is label, is only capable of 252W total @ 12V. And, to make it really interesting it only has two Molex connectors to connect the adapter I used (two Molex to 8-pin PCIe)...and (you'll love this)
both of them are on one lead going back to the PSU itself, with a single, 20ga wire (not even an 18, mind you) to feed the whole 8-pin connector. Lots of crimped connectors, and mechanical pin matings between the PSU and the 8-pin GPU connector.
Hopefully you can see I really wanted this to be "worst case".
I decided not to use the 1070 after all, because it does actually use more power than a 1060, and to be honest I was concerned for the equipment, given my horrible PSU choices as above. So, I took one of the EVGA 1060 SSC cards I have which has an 8-pin PCIe connector and used that.
And, lo and behold...
"It works!"I even got up the nerve to run a 3DMark11 "P" benchmark on it, to make sure the GPU was loaded; watched the fans come on as usual. during the second frame rate test (they're off at minimal loads) It finished the entire run, with no indication it was struggling at all. Scored precisely where my records indicate it should. For all intent and puprose, it's ready to go, right?
Well...maybe not. I did reach over to the wires feeding the GPU's 8-pin connector. And it was warm, especially right at the point where the first Molex 'daisy chains' to the second, right where I'd expect it to - where all that resistance is going to occur, from pins and mechanically crimped connectors. Not melting insulation warm...but warm enough I'd have concerns about closing it up in a chassis full of other components and their heat. And that was from a few minutes of benchmarking, never mind hours of a game running continuously.
I even took some pictures and made a short video, though I don't know if I'll post them or not because now you apparently have to use a third-party host, which I'm not keen on.
Anyway, there you have it: Even though I can make it work under what are obviously dangerous and stupid conditions, I would never (ever) recommend that anyone do something like this. I was pretty nervous running this test under controlled conditions for even a short period. I can only imagine what it would be like to run a setup like this for a long time - but I'd bet money there's someone out there who's done it, if they're not actually doing it right now.
This should illustrate clearly that just because you can do something, with adapters and so on, to make a computer run, doesn't mean it's a good idea, compliant with spec, or even safe to do so.
Speaking of adapters, this should also clearly demonstrate why I'm not a big fan, and
how easy they make it to do something totally stupid. Hopefully we all know better than to plug too many cords/devices into those 'power strips'...and this is no different at all.
Best regards to all
