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? Brought to you by Super Global Mega Corp .com