JerMe said:If Memtest86 can run for 24 hours error free, but Prime95 crashes within 5 minutes, then the computer is unstable, right?
IME, Memtest86 is good for an initial test to find a FSB level that'll give you a clean boot. e.g. If a stick is "Memtest86 clean" @ 233, it probably won't be prime stable at 233... you'll probably get a clean boot to desktop, but you'll usually have to take 2 - 5 notches off the FSB, depending on the mobo, PSU and other factors, to be prime/3D stable. Some BH5 sticks will also suck at low voltage; but as you jack up the vdimm to or past their sweet spot, they start scaling up well with the vdimm increase.
JerMe said:Stick #2 can only get to 223MHz(Prime95 stable)/227MHz(Memtest86 stable). All at 2-2-2-6, Vdimm 2.9V. Together in dual channel they run as fast as the slowest stick.
What is its top end prime-stable FSB @ 2-2-2-6 using 2.75-2.8? If it's below 217 then it would be a candidate for RMA (assuming you're using an nF2 platform judging by tRAS=11). I'm guessing it won't do 217, 2-2-2-6 w/2.8v if it takes 2.9v for prime-stable @ 223, 2-2-2-11.
Something else to consider before you RMA though.... it appears the tail end of the Mushkin level II stock is not doing as well as earlier yields; you just might get some slower sticks with the RMA replacement.