dgary@ecsvax.UUCP (08/31/84)
<>
A surprising fact surfaced at an IBM PC AT demo the other day:
While programs generally run 3x as fast on the PC AT as on the
PC, programs making heavy use of the floating point chip (8087 on
the PC, 80287 on the PC AT) seem to run at the same speed. For
example, inverting a 30x30 matrix in IBM's APL took about 10
seconds on both machines. This was confirmed by tests of at least
two programs (the APL interpreter and an program written in (I think
it was) assembler. Very curious. My suspicion is that the 8087 and
the 80287 run at about the same speed, and that most of the
time being eaten up in these tests was floating point crunching.
I'm still surprised the faster and wider memory access didn't
speed things up!
D Gary Grady
Duke University Computation Center, Durham, NC 27706
(919) 684-4146
USENET: {decvax,ihnp4,akgua,etc.}!mcnc!ecsvax!dgarygnu@sun.uucp (John Gilmore) (09/07/84)
Indeed, the 8087 and 80287 are the same inside. There's just a different bus interface. Somehow changing the bus interface took them about as long as it took to produce 80286's. They had working samples in April '83 but at the time various people were figuring out how to hook them up to 68000's. For some mysterious reason they decided to withdraw the part until the 80286 was ready. Here's to open systems.
starr@shell.UUCP (Bob Starr) (09/14/84)
Check the specs of the chips... it takes the same number of clock cycles for the FP operations on both chips. The AT is clocked faster than the PC, so the 80x87 should run a tad faster on the AT. Yes, it is a little disappointing.... but real application speeds (in C) of 10-12K flops isn't shabby for a micro!