boesch@Shasta.STANFORD.EDU (Brian Boesch) (08/08/87)
------ I have seen ads for several 80386 machines some include sockets for 80287 coprocessors som 80387 coprocessors and some are switchable. I asked a salesman at PC-limited (Their machine only supports 80287) and he says they get 90% of the performance of a 387 for a lot less cost. Is the only difference between the 387 and 287 a few percent of speed up (I assume from a wider buss interface)? Do the 80287 and 80387 take the same commands? and is software compiled for one executable on the other? Thanks in advance. Brian Boesch@shasta.stanford.edu Boesch@shasta.uucp
jay@splut.UUCP (Jay Maynard) (08/13/87)
In article <1927@Shasta.STANFORD.EDU>, boesch@Shasta.STANFORD.EDU (Brian Boesch) writes: > I asked a salesman at PC-limited (Their machine only supports 80287) > and he says they get 90% of the performance of a 387 for a lot less cost. > > Is the only difference between the 387 and 287 a few percent of > speed up (I assume from a wider buss interface)? > > Do the 80287 and 80387 take the same commands? and is software compiled > for one executable on the other? In the August 1987 issue, PC Tech Journal says that its floating point benchmark ran over 6 times faster on a Compaq Deskpro 386 than it did on a PC/AT at 8 MHZ with a 287. Since the AT-series runs the 287 at 2/3 the main processor clock speed (don't ask me why), the 287 in the tested AT ran at 5.33 MHz. That makes the 16 MHz 387 do things about twice as fast as a 287 for equivalent clock speeds. I don't know whether to attribute the speedup from the 287 to the 387 to a wider bus interface or not, but it appears to be markedly faster anyway. Note, though, that you pay a price for the speed: the 387 is going for >$700, while even an 8 MHz 287 is still in the $275 range. -- >splut!<...Jay Maynard, K5ZC | uucp: ...!seismo!soma!uhnix1!sugar!splut!jay "Don't ask ME about Unix... | GEnie: JAYMAYNARD (...e-i-e-i-o!) I speak SNA!" | CI$: 71036,1603 FidoNet: SysOp @106/64 The opinions herein are shared by neither of my cats, much less anyone else.
kanwar@intelca.UUCP (kanwar chadha 75657 @intelca) (08/14/87)
> ------ > > I have seen ads for several 80386 machines some include sockets for 80287 > coprocessors som 80387 coprocessors and some are switchable. > > I asked a salesman at PC-limited (Their machine only supports 80287) > and he says they get 90% of the performance of a 387 for a lot less cost. > > Is the only difference between the 387 and 287 a few percent of > speed up (I assume from a wider buss interface)? No. There is a substantial speed difference between the 80287 and 80387. 80387 is 5 to 8 times faster than the 80287 depending on the instruction mix you have. The 387 speed improvement is mainly due to four factors : (1) The 80387 core was redesigned from scratch to improve the clock counts for floating point instructions, e.g. the 387 executes an ADD instruction in approx. 25 clocks compared to about 100 for the 287. (2) The 387 has added some new instructions like sine/cosine etc to improve the performance in engineering intensive applications. (3) The 387 clock frequency is 16 Mhz and 20 Mhz while the 287 only goes up to 10 Mhz. (4) As you mentioned the 386-387 system operates on 32-bit bus versues 16 bit for the 286-287 system. When you combine all of the above, you can see that the 387 has a big performance advantage over the 287 or the 8087. Actually the 387 has a better price-performance ratio than the 287. Ofcourse the real performance improvement in application software is dependent upon how much floating point intensive the application is. But in a 386 system it makes much more sense to use the 387 rather than the 287 and companies like Compaq are providing upgrades for 387 in their 386 systems. ( Many of these companies introduced 386 systems with 287 sockets because the 387 was not available at that time.)kkkkk> > f Do the 80287 and 80387 take the same commands? and is software compiled > for one executable on the other? > Yes. The 387 is object code compatible with the 287 /8087 and all your software should run unmodified on the 386-387 system. Kanwar Chadha (Intel Corp.) > Thanks in advance. > > Brian > > Boesch@shasta.stanford.edu > Boesch@shasta.uucp *** REPLACE THIS LINE WITH YOUR MESSAGE ***