Dual Prime 95 failures? why does a test fail?

[nealh]

I have a question regarding dual(for that matter single ) prime 95 failures...I am trying to help myself and other determine what to tweak for stablity

I am assuming 12hrs testing

Small fft=max stress on cpu
large fft= max heat, power consumption, cpu, ram
blend or custom test=more ram stress, ?more system stress, ? less cpu stress

so if you pass at given oc small fft this means the cpu is ok,
what would passing large fft means...that cpu and what is ok?
clearly a blend test will show more ram issue

so my question is empiric:
if you pass small fft..cpu at that vcore/oc is ok..then test large fft and blend or custom test

if you fail large fft what is the culprit..ram, heat etc(when I prime my Venice which has passed before I stopped 15 hrs each of small fft/large fft ...my temps were 1C different)...if small fft passses but large fft fails and temps look ok..where do you try to adjust setting

my feeling with the blend, I assume it shows a weakness in the ram-cpu function if small fft is ok, I look at the cpu as being ok at the given speed but the ram or memory controller is unhappy?

does the error message prime reports mean anything..all I ever see is execution halted because of hardware failure, round was 0.4 and 0.5 expected or something like that??

any thoughts

 

[Budman]

if i were you i would run http://www.memtest.org/ to rule out the ram before going further.

 

[nealh]

I have personally run the ram at lower voltage than previously needed

memtest 50 passes on test 5, 3hrs on test 7(12 passes or so) & 8(totaled 40 passes or so)

previously ran 6hrs all tests


These are theortically questions as I am slowly max. my opty 170...trying to decide what I need to tweak

My ram does not do as well with my opty as my venice....