zuldan wrote on Yesterday, 22:47:Ouch! What make/model PSU is that?
It's a Codegen "520W" model - rated with 42 amps on 5V (the real limit would be 30 amps, as that's the rating of the 5V rectifier diode). A not so great PSU. However I don't want to go through the efforts of checking ancient CPUs, when going to 12V for the CPU is possible. I halt my overclocking until the 12V mod is done. No need to risk anything.
Living wrote on Today, 02:03:i had to abandon the idea of a fast Socket A computer because of this. The pull from the 5v rail is insane and there is no modern power supply below u$s 150 that can sustain that (i refuse to use 20years+ PSU's)
Well.. fast Socket A is a wild range. If you overclock in extreme ranges, the CPU consumes just a hecking lot of power. Here's 2V VCore + SuperPi 1M on an A7N8X - this is the 5V rail:
We're having 4.86V at the ATX connector with a current draw of 34A - which results in 165,24W of power going into that board
The attachment power_hungry.jpeg is no longer available
Ozzuneoj wrote on Yesterday, 22:11:I doubt I would ever do anything like that, but I find the idea totally fascinating. I'd love to see pictures and read about how you do it.
How thoroughly have setups like this been tested? Are there any long-term negatives to doing this aside from the usual stress of overclocking on the CPU?
I wouldn't see any serious issues with that. They've been proven during overclock to run with 12V for the VRM just fine.
STM specifically states that L6917B (used both on the A7N8X and K7S8X) is capable to run with 12V just fine. You'll get a drop in efficiency though.
The attachment Screenshot from 2026-07-06 06-48-56.png is no longer available