I set my JP1 & JP2 back to stock settings and guess what?
boots up and benches sandra, runs q3 timedemos fine...
at 4:5 2-3-3-5 1T (exact spec) ! woo

I'm going to try for 2-2-2-5 but if it won't do it (or 2-2-2-6 1T on auto, that's a very good idea mechBgon

) then I'm happy with spec.
a few numbers for ya over the 2-3-3-5 2T ones,
sandra
int 2021
float 1941
Q3 @ 1024 had an avg of 186.7 fps but keep in mind this is on the new 29.42's, not the older 28.32's I had to use in my past benches.
float goes up by about 30 points going from 2T to 1T, not that significant but the greater 'victory' is getting it to run at full spec after these weird jumpers. now for more tests... and honestly the bigger issue holding me back is NeverWinter Nights

can't pull myself away... now to see if I can push this baby further (the Corsair guy in their forums seems to think these should do 2-2-2 even tho it's not official)
edit: well, 2-2-2 still doesn't want to work, be it 2-2-2-5, 2-2-2-6, 1T or not. I'm happy with how it is...for reference, I did get 2-2-2-5 1T to post in windows, but when I ran Sandra my mouse died and 3dmark poofed on execution, as well as a bluescreen when I tried to reboot...
all for these silly numbers:
int 2044
float 1958
once I bumped it back up to 2-3-3-5 1T (STABLE), I immediately got:
int 2026
float 1945
it's just not worth the hassle, these chips are tested to run at their verified high speeds and nothing higher

Samsung 6ns would work better, but you still wouldn't see much gain without overclocking and/or unlocking your athlon to really utilize 166/333. As you can see, these are barely above the PC2100 scores, if even that. real-world gain isn't much... video drivers do me more good

hope that helps