Introducing NeXT - The Wonders of NEXTSTEP and OpenStep
Linked by Eugenia Loli on Tue 15th Jul 2003 18:26 UTC
OpenStep, GNUstep Every so often I have this urge (maybe more of an itch) to spend hours and hours on the web trying to find information about old, obsolete computers of the past. I am intrigued by the XEROX Alto and Star ('70s-'82), the Apple Lisa ('83) and, of course, CRAYs ('75-ish). These were revolutionary machines indeed, they wrote golden pages in the history of computing. In the end of the 1980s, a new innovative product was ready to ship, created by a bunch of people coming from Apple: The NeXT platform.
E-mail Print r 0   · Read More · 86 Comment(s) http://osne.ws/34a
Permalink for comment
To read all comments associated with this story, please click here.
Re: Intel i860
by whaaa on Wed 16th Jul 2003 22:56 UTC

"All RISC machines are hard to program. I'm not sure the i860 was any harder than any of the others at the time (about 1990 to 1993). If I remember correctly, the chip had a 40 MFLOPS theoretical peak performance (64-bit floating point), and you could achieve a respectable 36 MFLOPS actual performance on matrix-matrix products. For the time, this was about as good as it got for microprocessors -- it would take another 4 or 5 years for the pentium performance to catch up."

Not exactly the instruction dependencies to keep the machine fed and happy were horrible for the i860, you could have far more elegant systems like MIPS offering the same o better performance during the same time frame. The "actual performance" was achieved on handcoded programs, compilers themselves were rather hard to optimize, hence the lack of interest in the marketplace. Plus the lack of x86 compatibility made vendors look elsewhere for RISC machines.
But for specific well tuned code the 860 was a good chip, hence its relative acceptance as a coprocessor.