🚀 go-pugleaf

In article <1992Aug19.132542.2667@crd.ge.com>, davidsen@ariel.crd.GE.COM (william E Davidsen) writes:
|> In article <16rb6gINNt38@iraul1.ira.uka.de>, S_JUFFA@iravcl.ira.uka.de (|S| Norbert Juffa) writes:
|> 
|> | The cache is definitely the main source of the 486DLC's performance
|> | gain over the 80386. This can be easily deduced from the performance
|> | of Intel's RapidCAD chip which is basically a 486 without the cache
|> | and with a 386 pin out. While the RapidCAD offers roughly twice the
|> | floating point performance of the Intel 387DX, my benchmarks showed
|> | it to be at most 25% faster than the 80386. Typical performance gain
|> | for integer applications is more like 15% though. This definitely
|> 
|>   Whoa! Is this chip OEM only or available to real people. I have a app
|> running on old (1986) 386-16s with 287 math chips. If the rapidCAD chip
|> is a drop in I could probably get an order of magnitude boost in the FP
|> performance.
|> 
|>   Motherboard upgrade is not available for this application.

The RapidCAD part is actually two devices -- one which must go into the 
386 socket and one which must go into the 387 socket. I presume that you 
don't have a 387 socket on your old motherboard so this is probably not 
an option. BTW, I have seen it advertised so I'd assume that it is available 
to real people. BTW, as I understand it, the current 287 rev has the current 
387 core and thus performs much better than the 287 did, so you might see 
a performance gain by moving to current 287s (esp. since I think that 
RapidCAD or 387s are not an option for you). 

The device which goes in the 387 socket from the RapidCAD chipset only has 
a "couple of hundred gates" which are used to ensure that certain signals 
relating to FP exceptions are handled correctly. The FP unit is on board 
the "386" part of the RapidCAD chipset. 

Jon Mellott
High Speed Digital Architecture Laboratory
University of Florida
(jon@alpha.ee.ufl.edu)