Boinc benchmark do not show real value: since there is point where memory become to slow for transfer data, and calculation become slower. Try yourself different combination , but current one 12 workunit per cycle is definitely the worst choice you can make. Same will be with 6 workunit per cycle. Try 2 cores per workunit ( with HT off , of-course)
314187728^131072+1 GENERALIZED FERMAT :)
93*10^1029523-1 REPDIGIT PRIME
31*332^367560+1 CRUS PRIME
Proud member of team Aggie The Pew. Go Aggie!