Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site brl-tgr.ARPA
Path: utzoo!watmath!clyde!bonnie!akgua!whuxlm!harpo!decvax!ucbvax!ucdavis!lll-crg!seismo!brl-tgr!tgr!tweten@AMES-NAS.ARPA
From: tweten@AMES-NAS.ARPA (Dave Tweten)
Newsgroups: net.micro
Subject: Re: NEC V20 ---> 8088
Message-ID: <1493@brl-tgr.ARPA>
Date: Sat, 14-Sep-85 01:08:23 EDT
Article-I.D.: brl-tgr.1493
Posted: Sat Sep 14 01:08:23 1985
Date-Received: Mon, 16-Sep-85 00:23:08 EDT
Sender: news@brl-tgr.ARPA
Lines: 65

I recently bought an NEC V20 and installed it in my Z-151, which I am
using to write this message.  When I pried the 8088 out from next to my
8087, I noticed that it too had been a NEC part.  Contrary to earlier
comments in this forum about NEC 8088s not working with 8087s, it had
worked flawlessly with my 8087 for the previous year.

Preliminary experience is that the V20 speeds up some programs
noticably, and has no effect on others.  That is to be expected.  If a
program is 8087 limited or I/O limited, speeding up the 8088 will do no
good.  It has worked at least as well as the 8088 for any program I
have tried.

The only "negative" effect of the V20 is it causes Zenith's disk-based
diagnostics for CPU-board crystal frequency, and for floppy-disk driver
crystal frequency to fail.  I presume the tests compare crystal cycles
against a wait-loop counter.  Since the NEC V20 "waits faster" the
tests fail.  Sorry, no time yet to do benchmarks.

	From: Charles R. LaBrec 

	I haven't really heard many specifics of the NEC V20.  Is it
	really a case of design stealing or just a case of duplicating
	the 8088 instruction set?  Would someone care to enlighten me?

I don't presume to be an engineering law expert, but by no strech of my
imagination can I conceive to the V20 being an 8088 carbon copy, either
legal or illegal.  The following information was gleened from Intel's
"iAPX 88 BOOK" and from the NEC document titled "V20, uPD70108,
HIGH-PERFORMANCE 16-BIT MICROPROCESSOR, PRELIMINARY INFORMATION", dated
May 1985.

  .  The time for a register-to-register ADD is quoted as three clocks
     for the 8088, two clocks for the V20.  NEC's literature claims
     that is due to dual 16-bit on-chip busses for the V20, as opposed
     to a single bus in the 8088.  That supposedly permits two-cycle
     register-register instructions (get both operands, return result),
     where the 8088 uses three (get one operand, get the other, return
     the result).  A quick scan through the respective instruction
     timing charts indicates that the relationship holds for all
     trivial two-register instructions (this obviously doesn't apply to
     multiply and divide).

     Intel's register-register 16-bit operand, 32-bit result multiply
     is quoted at 118-113 clocks.  NEC's is quoted as 41-47.  The
     equivalent divide times are 165-184 cycles for Intel and 38-43 for
     NEC.  Yes, I too noticed that NEC claims to divide faster than they
     multiply, and I can't explain it either.

  .  NEC claims to use a separate address resolution unit on the chip,
     instead of using the arithmetic unit.  Their effective address
     calculation time is two cycles for any mode.  Intel's ranges from
     5 to 12, depending on mode.

  .  The NEC chip has an expanded instruction set.  By my estimation,
     it includes all the 80186 set plus several more.  It has bit-field
     insert and extract (perhaps useful in low level graphics?).  It
     can test and manipulate individual bits in memory.  It has packed
     decimal string add, subtract and compare.  It has a BCD digit
     rotate instruction.  Those are the highlights (as I see them);
     there are several more instructions I haven't mentioned.  There is
     also a complete 8080 emulation mode which interests me not at
     all.

In summary, it appears to me that if the V20 is a "pirate" 8088, then
the Z-80 was a "pirate" 8080.  Is our chauvinism showing?