🚀 go-pugleaf

In article <3312@utcsri.UUCP> henderso@utcsri.UUCP (Mark Henderson) writes:
>I heard recently that the 80287 runs at a clock speed of only 5.3 MHz when
>in an IBM AT and also that one can purchase a small board which fits into the
>80287 socket on the AT, to drive it at 8 MHz (assuming of course one has an
>80287-8). Where can one purchase such a board? Any information would be
>appreciated. If you have them, please include prices.

Depending on what you are using, putting a 287 may not get you
anywhere.  In fact, it may slow down your system!  
 
You may want to check the IBM special issue (10/85) of Byte, page 66 on.
the benchmarks that they ran did not give the increase in proformence
that you might expect.  In at least one case it slowed it down.

They used different clocks for both the 286 and 287;
286: 6 & 9 MHz
287: 0, 4, & 8 MHz
 
You might check byte for the speed up board for the 287, as I have seen
them in byte.

greg waldref

....tektronix!psu-cs!psueea!waldref

> I heard recently that the 80287 runs at a clock speed of only 5.3 MHz when
> in an IBM AT and also that one can purchase a small board which fits into the
> 80287 socket on the AT, to drive it at 8 MHz (assuming of course one has an
> 80287-8). Where can one purchase such a board? Any information would be
> appreciated. If you have them, please include prices.
> 
The 80287 runs at 1/3 of the system clock while the processor runs
at 1/2.  On 6 and 8 mHz machines this gives speeds of 4 and 5.33 MHz
respectively for the coprocessor clock.

The 80287 is designed to be able to run asynchronously from the 80286
clock however. You can attach an 80287 running at 10 MHz simply by
plugging it into the socket provided BUT giving it a separate clock
signal. You can get plug in modules that do this from Microway and
Hauppauge with speeds from 6 to 10 MHz. All they do is interpose
a socket which breaks the clock supply from the motherboard and then
provide their own clock. You must have the correct (approx 2:1)
clock duty cycle though.

Mike Gingell ...decvax!mcnc!ecsvax!mjg

> In article <3312@utcsri.UUCP> henderso@utcsri.UUCP (Mark Henderson) writes:
> >I heard recently that the 80287 runs at a clock speed of only 5.3 MHz when
> >in an IBM AT and also that one can purchase a small board which fits into the
> >80287 socket on the AT, to drive it at 8 MHz (assuming of course one has an
> >80287-8). Where can one purchase such a board? Any information would be
> >appreciated. If you have them, please include prices.
> 
> Depending on what you are using, putting a 287 may not get you
> anywhere.  In fact, it may slow down your system!  
>  
> You may want to check the IBM special issue (10/85) of Byte, page 66 on.
> the benchmarks that they ran did not give the increase in proformence
> that you might expect.  In at least one case it slowed it down.
> 
> They used different clocks for both the 286 and 287;
> 286: 6 & 9 MHz
> 287: 0, 4, & 8 MHz
>  
> You might check byte for the speed up board for the 287, as I have seen
> them in byte.
> 
> greg waldref
> 
> ....tektronix!psu-cs!psueea!waldref

As I understand it, the faster your 286 goes the faster your 287 should go. 
Best bet would be to match the speed of your 286 with a 287 of comparable or
faster speed. Microway, Inc. in the Boston-area makes MANY 287-type speed up
products.


-- 

-- Ed.
    
Net  :  {noao!ihnp4!yale!}!hsi!tankus
Snail:  Health Systems Int'l, 100 Broadway, New Haven, CT 06511
Bell :  (203) 562-2101

In article <253@psueea.UUCP> waldref@psueea.UUCP (Greg Waldref) writes:
>Depending on what you are using, putting a 287 may not get you
>anywhere.  In fact, it may slow down your system!
>     
>You may want to check the IBM special issue (10/85) of Byte, page 66 on.
>the benchmarks that they ran did not give the increase in proformence
>that you might expect.  In at least one case it slowed it down.

As I recall, the impression I got when I first read the article mentioned
above was different than the impression I got upon the second reading.
Now that I have reread the article quickly, I am not convinced the author
was comparing apples with apples when he stated the 80287 can degrade a
PC's performance.  Sure, it is possible to construe cases in which this is
true: run the 80286 at 10 MHz and the 80287 at 2 MHz, or write some software
that uses the 80287 to move floating-point numbers from one memory location
to another without change in format, as opposed to using the 80286 to do the
same.  *I* (though maybe not you) think the main point being made was that
the 80287 might provide marginal speedup at best in some cases, and if one is
not real careful to do things right, it may actually hinder performance.  (I
guess my only quibble with Mr. Waldref's article is the last sentence quoted
above.)
-- 
Samuel A. Figueroa, Dept. of CS, Univ. of KY, Lexington, KY  40506-0027
UUCP: cbosgd!ukma!sambo    CSNET: sambo@uky.csnet

	"Micro-ln is great, if only people would start using it."

On 80287 performance:  I just benchmarked some code that we wrote in Lattice
C 3.1 with and without the 80287 chip.  The actual execution time was the same
both ways, although the exact split between the two sections of the program
was different.

                           Byte shuffling      Calculation
With 80287                     6 sec              5 sec
Without 80287                  5 sec              6 sec

This was run on an ATronics 12 MHz AT with the 8MHz 287.

-- 
Bill Roper, ihnp4!chinet!roper