chau@caen.engin.umich.edu (Hin Fai Chau) (04/09/91)
I have similar questions but it is for 80387 33 MHz math coprocessor, so I combine them with the previous article as follows: 1) How compatible are Cyrix, Weitek, and any others to Intel? What other brands are available? What technology are they based on, CMOS or NMOS? 2) What are the relative performance differences between them in terms of speed and accuracy? Any known hardware bugs in those chips? 3) Are 80387 generally run at the same speed as the main CPU? (I was told that the 2/3 rule is generally true for 86 and 286 machines) 4) What are the relative costs between chips from different manufacturers? 5) Do I need a different coprocessor socket for Weitek 80387 chips? 6) Any clear winners? Any good mail-order places to get them? Any comments or suggestions will be greatly appreciated. Thanks in advance! Hin-Fai Chau chau@caen.engin.umich.edu
edm@hpfcmdd.hp.com (Ed Moore) (04/11/91)
The 2/12/91 issue of PC Magazine reviewed the Intel 80387, Cyrix 83D87 and EMC87, and IIT 3C87. They summarized, "Four good performers, but no stand-out." This is a real-world review, using Lotus 1-2-3, Excel, Quattro Pro, AutoCAD and DesignCAD.
id8rld06@serss0.fiu.edu (Michael N Johnston) (04/13/91)
>> 4) Do I need a different coprocessor socket for Weitec 80387 chips?
The Weitec is NOT a clone of the 80387. There are a couple of versions
of the Weitec for 386s. On is a chip the other is a chipset on a card.
Both require a socket for the Weitec and will not work in the 80387
socket.
The Weitecs are not compatible with the 80387's nor do they claim
be. The Weitec is in a different league from the 387s and cost a few
hundred $'s more. It is a very high performance math coprocessor and
is generally used for VERY intensive number calculations.
A program that detects the 387 and uses it if present will not use
the Weitec unless it is specificly designed to also work with the
Weitec.
The other questions I will leave for others.
Mike
--
====================================================================
Michael Johnston
id8rld06@serss0.fiu.edu or
26793271x@servax.fiu.edujsims@vuse.vanderbilt.edu (J. Robert Sims) (04/19/91)
In article <3295@kluge.fiu.edu> id8rld06@serss0.fiu.edu (Michael N Johnston) writes: >>> 4) Do I need a different coprocessor socket for Weitec 80387 chips? > >The Weitec is NOT a clone of the 80387. There are a couple of versions >of the Weitec for 386s. On is a chip the other is a chipset on a card. >Both require a socket for the Weitec and will not work in the 80387 >socket. Some clones _do_ use the same socket. If you use an 80387, you insert the chip in the middle of the PGA socket, and if you install a Weitek, it fills the whole socket. This means that you can't have both in your machine at the same time, even though they can exist together if there are two sockets. > The Weitecs are not compatible with the 80387's nor do they claim >be. The Weitec is in a different league from the 387s and cost a few >hundred $'s more. It is a very high performance math coprocessor and >is generally used for VERY intensive number calculations. I have seen comparisons (PC Magazine?) that indicated that the Weitek's performance was quite exaggerated. It was only blindingly fast with certain special optimized benchmark programs, and in real applications was only a little bit faster. The only league difference is price. > A program that detects the 387 and uses it if present will not use >the Weitec unless it is specificly designed to also work with the >Weitec. True. However, almost nothing will use the Weitek. The only packages I've heard of that will are a few obscure finite element analysis packages, and those will also use the *87 processors. > The other questions I will leave for others. > >Mike >-- >==================================================================== >Michael Johnston >id8rld06@serss0.fiu.edu or >26793271x@servax.fiu.edu Rob
tinyguy@cs.mcgill.ca (Yeo-Hoon BAE) (04/23/91)
In article <1991Apr18.211744.12564@vuse.vanderbilt.edu> jsims@vuse.vanderbilt.edu (J. Robert Sims) writes: > >I have seen comparisons (PC Magazine?) that indicated that the Weitek's >performance was quite exaggerated. It was only blindingly fast with >certain special optimized benchmark programs, and in real applications >was only a little bit faster. The only league difference is price. > >Rob Isn't some of the Sun Sparcs use Weitek as their Math Chip? It's rated figure is about twice of that of 486 at same clock speed... (I'm talking about SPECmarks for floats) So it _should_ have some significant improvement over 387... -TG
davidsen@sixhub.UUCP (Wm E. Davidsen Jr) (04/24/91)
In article <1991Apr18.211744.12564@vuse.vanderbilt.edu> jsims@vuse.vanderbilt.edu (J. Robert Sims) writes: | True. However, almost nothing will use the Weitek. The only packages I've | heard of that will are a few obscure finite element analysis packages, and | those will also use the *87 processors. There are compilers available for C, F77, and Pascal, in MS-DOS and UNIX. If you mean commercial programs for the 387 don't use a totally different chip, that's true. People who are thinking about a Witek don't find finite element programs obscure. -- bill davidsen - davidsen@sixhub.uucp (uunet!crdgw1!sixhub!davidsen) sysop *IX BBS and Public Access UNIX moderator of comp.binaries.ibm.pc and 80386 mailing list "Stupidity, like virtue, is its own reward" -me
jgay@digi.lonestar.org (john gay) (04/24/91)
From article <1991Apr23.081807.27116@cs.mcgill.ca>, by tinyguy@cs.mcgill.ca (Yeo-Hoon BAE): > In article <1991Apr18.211744.12564@vuse.vanderbilt.edu> jsims@vuse.vanderbilt.edu (J. Robert Sims) writes: >>I have seen comparisons (PC Magazine?) that indicated that the Weitek's >>performance was quite exaggerated. It was only blindingly fast with >>certain special optimized benchmark programs, and in real applications >>was only a little bit faster. The only league difference is price. >>Rob > Isn't some of the Sun Sparcs use Weitek as their Math Chip? It's > rated figure is about twice of that of 486 at same clock speed... > (I'm talking about SPECmarks for floats) > So it _should_ have some significant improvement over 387... I think that most of the speed gained by the Weitek coprocessor is because it is memory mapped. The new Cyrix memory mapped '387 is supposed to be just as fast as the Weitek (or very close) and can be used as either memory mapped or interrupt driven. There is an article about 386 coprocessors in the May '91 Dr. Dobb's. Give it a read. john gay.
pjh@mccc.edu (Pete Holsberg) (04/26/91)
In article <1991Apr24.165319.3600@digi.lonestar.org> jgay@digi.lonestar.org (john gay) writes:
=I think that most of the speed gained by the Weitek coprocessor is because
=it is memory mapped. The new Cyrix memory mapped '387 is supposed to be
=just as fast as the Weitek (or very close) and can be used as either
=memory mapped or interrupt driven. There is an article about 386 coprocessors
=in the May '91 Dr. Dobb's. Give it a read.
And the Cyrix is a good deal less expensive. We just ordered a whole
bunch at $398 and change.
Pete
--
Prof. Peter J. Holsberg Mercer County Community College
Voice: 609-586-4800 Engineering Technology, Computers and Math
UUCP:...!princeton!mccc!pjh 1200 Old Trenton Road, Trenton, NJ 08690
Internet: pjh@mccc.edu Trenton Computer Festival -- 4/20-21/91stu@valinor.UUCP (Stuart L Labovitz) (04/27/91)
jgay@digi.lonestar.org (john gay) writes: > > I think that most of the speed gained by the Weitek coprocessor is because > it is memory mapped. The new Cyrix memory mapped '387 is supposed to be > just as fast as the Weitek (or very close) and can be used as either > memory mapped or interrupt driven. There is an article about 386 coprocessor > in the May '91 Dr. Dobb's. Give it a read. > There is also an excellent review of the Cyrix EMC87, 83D87, and 83S87 (along with comparisons to the Intel 80486, 387DX, and 387SX) in the May 1991 issue of IEEE Software (pg 119-120). This review, entitled "Cyrix Chips Outperform Intel Math CoProcessors," was written by Sorel Reisman (reachable by e-mail at s.reisman@compmail.com) and discusses how the 80386 with a Cyrix co-processor can, generally speaking, be faster than the 80486 (!). Stuart L Labovitz home: stu%valinor.uucp@dayton.saic.com or stu@valinor.uucp or dayvb!valinor!stu work: slabovit@blackbird.afit.af.mil As far as the laws of mathematics refer to reality, they are not certain; and as far as they are certain, they do not refer to reality.