Okay, my Epox board's BIOS raises the voltage in +0.05v increments. In fact, the settings don't actually read "1.40v," "1.45v," 1.50v," etc., but rather "+0.00v," "+0.05v," "+0.10v," etc. I never actually checked to see whether or not "+0.00v" amounted to 1.40v, as it should. I seem to remember seeing 1.47v during the boot cycle, but that was a while ago. Anyway, kicking it down from +0.10 to +0.05 did drop the voltage somewhat.
+0.10v gives a true voltage of 1.60-1.63v, depending on CPU load.
+0.05v gives a true voltage of 1.53-1.57v, depending on CPU load.
I was able to knock it down to +0.05v and still boot Windows, but attempting to go all the way down to +0.00v was unsuccessful. So, my Sempron 64 2800+ is now running at 1.53-1.57v, 2.4 GHz, with 200 MHz RAM and 900 MHz HT. Thanks for the help, everyone!
Of course, I still need to run Prime95 to test stability. An short test (not adequate for testing stability, but still enough to max out the temp.) seems to indicate the maximum temperature under Prime95-caliber load is 61^C, which, although still not ideal, is a heck of a lot better than the 67^C I was dealing with, before the voltage drop. Also, the new test was run in a cooler room than the old, suggesting an even more dramatic drop in temperature than a simple 6^C.
Thanks again, everyone!