I was just made aware that the Rev. C2 version of the AMD Phenom II X4 945 can be 125W TDP (most are 95W TDP), so that part is correct. So it does not account for the 95W Rev. C2 or C3 version of the CPU...
I went through the trouble of setting up a test server on a MiniPC (Intel NUC NUC6AYH), and it does run there. Though, I forgot to run lscpu for the instruction flags... so I searched for it online, and that NUC's CPU does indeed have the flag.
I then actually compiled the source on an even older Dell Latitude D830 and executed it. Same illegal instruction (SIGILL) as on the aging server. It does not have the flag.
It does run on my main laptop that has a high-end Sandy Bridge CPU, which has the flag.
--
From the AMD server and main laptop logs:
MariaDB 11.4: InnoDB: Using generic crc32 instructions
MariaDB 11.8: InnoDB: Using crc32 + pclmulqdq instructions
However, I found an old .tar image of the server 'www' data from September 6, 2023, taking up ~548 GiB on the www-data RAID setup. (The final one is .xz'ed, taking up ~449 GiB on the NAS.) Thought I purged it years ago, lol.
There's also some old HP laptop disk image on there from 2014 that I have never removed, taking up ~109 GiB; since copied to the NAS years later.
Another ~800 GiB used up there is a bunch of Teknoparrot ROMs just sitting there as well.
Very old versions of MySQL sit there; no longer can use them since the move to MariaDB.
There's more decade [plus]-old cruft on that setup that existed ever since the server used a 4x 250GB RAID0+1 setup in the early 2010s. Removing a lot of it would free up probably ~2TiB of space.
Somehow, one of the PHP -FPM configs wouldn't stick (I changed the listen user/group entry to what httpd is using and it was stuck on default nobody...). I had to edit it locally and not from my laptop. 🤨
php 7.4.33: fpm now consistently runs on Unix sockets instead of TCP/IP:Port, meaning connections within PHP are even faster. This was mainly due to merging out-of-date configurations that I thought were up-to-date. [Was originally CGI for the longest -> FastCGI for testing -> FPM TCP/IP:Port -> FPM Unix Sockets]
Other notes: MySQL/MariaDB did run on TCP/IP:Port due to Unix Sockets breaking at random historically, but that had been stabilized for at least a year.
As a result, I have finally upped the bandwidth limit from 786KB/1MB (rest of main site) and 393KB/512KB (certain download sections) to ~5MB and ~2.5MB respectively.
Other thing was Intel Core i3: 12100 -> 13100... but two pins located at the bottom-center of its socket on the board in question were quite bent. I managed to fix them perfectly, though.