davidsen@steinmetz.ge.com (William E. Davidsen Jr) (07/13/88)
Found an interesting number in a benchmarks the other day. I tested
two 386's, one running an 80287 at 10 MHz, the other an 80387 at 20 MHz.
Both were driving in 32 bit mode on the CPU, although it doesn't make a
lot of difference.
I expected the 387 to be twice as fast due to the clock speed, but
found it was 8-10 times faster, depending on the application. This makes
it about 3x a VAX 11/780 with fpa.
Looking at old results, I see virtually no difference in the
8087/80287 performance, with a slight bias toward the 8087, due to
faster data load times as I recall.
From this I conclude that adding a 386/387 add in board is effective,
but a 286/287 add in will only buy the difference in clock rates at the
coprocessor. Since many 287s run at 2/3 speed of the 286, this may be
slower than an old XT, certainly slower than a Turbo XT, if the 8087 is
run a full CPU speed.
Does anyone have results replacing the 68881 with a 68882?
--
bill davidsen (wedu@ge-crd.arpa)
{uunet | philabs | seismo}!steinmetz!crdos1!davidsen
"Stupidity, like virtue, is its own reward" -me