EEVblog #1308 - 1970's Intel MCS-85 8085 Design Kit

[Kleinstein]

As far as I see it from behind the Z80 had 3 strong points: it could run 8080 code and was cheaper than 8080 or 8085.  It also had an easier interface, especially when using dynamic RAM.  The direct interface to DRAM was an advantage over the 6502 and could easily justify a slightly high price.

The price to compare is not just the CPU but also the other chips needed around the system (IO, clock, RAM). In those days the IO chips were also quite expensive.

I don't remember the prices from the very early days, but later the Z80 and 6502 were about the same price and the IO chips possibly more expensive. So the CPU price was often only a small fraction of the system. For some odd reason the 8085 stayed quite expensive.

[DrG]

/-/ The 6502 is only revered because it was the first device many people used, and nostalgia is strong. /-/

But you have to add in the REASON it was the first that so many used...the reason for the impact that it made....I guess that accessibility can breed reverence if you would not otherwise be able to get involved.

6502 price - US$25 https://www.eevblog.com/forum/vintage-computing/what-was-you-first-computer/msg2897904/#msg2897904 AFAIK, nothing close was cheaper.

For example, Z-80 (arguably MUCH more powerful and appearing after the 6502) US$200 (single quantities) https://www.americanradiohistory.com/Archive-Byte/70s/Byte-1976-08.pdf [p.38]

The prices of one off parts relate to their volume price about as well as a politician's words relate to their actions. People used the 6502 because it was in the box they used. It was in the box because it was cheap, and the original sources - basically MOS Technology and Rockwell - were able to crank them out in adequate numbers (which, or course, relates to them being cheap). The Z80 wasn't all that expensive if you were serious about buying it, and Zilog did well at ramping production (later Zilog wrecked the Z8000 by being very slow at ramping production). The Z80 would never have got to second place in consumer systems, behind the 6502, if it hadn't been pretty cheap.

"The prices of one off parts relate to their volume price about as well as a politician's words relate to their actions." Sounds to me like colorful deflection.

What exactly were the volume prices for the two chips and when? That information could possibly bolster your point, but it is lacking from your response. The citation I gave said that Z-80 multiple quantity prices (although I don't even think large quantities were available at the time of the article) got it down to US$80 - more than three times the original single unit price of the 6502.

BEFORE it was in the box that they used (save the KIM-1 developer board), it was a chip which, I believe, EEs investigated BECAUSE it was cheaper than anything else they could get their hands on. This was before the the Apple and the TRS-80 and so on.

What I read you saying is "The Z80 wasn't all that expensive if you were serious about buying it..."  What does that mean, "wasn't all that expensive"? Is that some kind of projective test? Is it OK to say that "it wasn't all that cheap" and the statements somehow cancel each other out?

What I distinctly do not see you writing is how much the quantity price for the Z-80 was and how that cost compared to the quantity price of the 6502....as well as the 6800 and Intel chips of the time... and of course when the comparison is being made. In a sense, you are talking like the politician that you mock - making the statements without backing it up with any facts. If you had presented some convincing facts, maybe I would change my mind.

The 6502 was the cheapest of it's kind and it was accessible. That is a major reason that so many talented people learned and developed with the 6502. Its use in the Pet and the Apple came later. To ignore the price differential and do so without data is simply not credible in my view - not if one wants to offer reasonable interpretations of how the history unfolded the way that it did.

But, yes, you are entitled to opinion and I don't have any difficulty with you continuing to respond without convincing evidence.

[Kleinstein]

AFAIK the 6800 was quite expensive in the early days. A reason why Motorola did not like the 6500.

In the same byte Magazine there was a price of some $40 for the 8080A (single - though probably not as much discount for quantities). But you also had extra chips for clock and system control (offered at some $12.50 each). I addition the 8080 still needed 3 supplies - this also adds costs and complicates the board.

The Z80 being very new back than, means the price could be just for the very early units, likely going down when more chips are available. Even at $80 the Z80 could compete with a 8080 system when using DRAM that gets away with a single 5 V supply. At least in the beginning the Z80 only needed to be comparable / cheaper than a 8080 system. It looks like the Z80 was also faster than the 8080 - especially with fast memory.

Some of the early memories (DRAM and EPROM) also needed an additional voltage. So the extra supply for the 8080 was not that bad and odd. Even the initial C64 series used a video chip with an additional higher supply (AFAIR some 9 V).

[DrG]

AFAIK the 6800 was quite expensive in the early days. A reason why Motorola did not like the 6500.

In the same byte Magazine there was a price of some $40 for the 8080A (single - though probably not as much discount for quantities). But you also had extra chips for clock and system control (offered at some $12.50 each). I addition the 8080 still needed 3 supplies - this also adds costs and complicates the board.

The Z80 being very new back than, means the price could be just for the very early units, likely going down when more chips are available. Even at $80 the Z80 could compete with a 8080 system when using DRAM that gets away with a single 5 V supply. At least in the beginning the Z80 only needed to be comparable / cheaper than a 8080 system. It looks like the Z80 was also faster than the 8080 - especially with fast memory.

Some of the early memories (DRAM and EPROM) also needed an additional voltage. So the extra supply for the 8080 was not that bad and odd. Even the initial C64 series used a video chip with an additional higher supply (AFAIR some 9 V).

I agree with you on all points and consideration (well except that I think you meant that Motorola did not like the 6502, not the 6500).

I think that you are right that the introductory price for the Z-80 as quoted was likely for very new units. On the otherhand, the 6502 introductory price of $25 could be viewed in the same way.

I'm really not all that knowledgeable about the retro history, but because I was there (I mean alive not as a player), it interests me.

The impression that I get is that the 6502 actually started a price war. In a few years, I think that the fiscal landscape had changed a great deal - but that is also a few years of development. https://en.wikipedia.org/wiki/MOS_Technology_6502

This thread https://retrocomputing.stackexchange.com/questions/2760/how-much-did-the-6502-and-z80-cost has some interesting "answers". I would love to see a hard-core database of prices and quantities and dates.

Nevertheless, I maintain that integral to the impact and popularity of the 6502 was soundly anchored in it's low initial price. If you build it, they will come...and if you sell it cheaply, they will develop. 

[floobydust]

1802, six rad-hardened used in the Galileo spacecraft and it's the only MCU MPU with a SEX op-code 

I don't remember it all but the 1802 concepts were difficult - any register can be the Program Counter, you can't load immediate- everything uses a register as a pointer. Having the return at the beginning of a subroutine? It is more efficient with multiple operations- but when I was a kid, it was too tough to figure out.
Code copy pasta from https://www.atarimagazines.com/computeii/issue3/page52.php

Code: [Select]

6502
AD 3412 LDA ADDR1 ; READ MEMORY
 
1802
F8 12 LDI A.1 (ADDR1) ; SET UP
B8 PHI R8 ; R8 TO POINT TO
F8 34 LDI A.0 (ADDR1) ; THE DESIRED
A8 PLO R8 ; MEMORY ADDRESS
08 LDN R8 ; READ MEMORY

;***************************************

; 1802 subr call
D7 SEP R7 ; GO TO QSUB

; 1802 subr QSUB
D0 QRET: SEP R0 ; RETURN TO MAIN PROGRAM
CD QSUB: LSQ ; IF Q IS SET, SKIP 2 BYTES
7B SEQ ; SET Q
38 NBR ; SKIP A BYTE
76 REQ ; RESET Q
30 00 BR QRET ; GO RETURN

[coppice]

1802, six rad-hardened used in the Galileo spacecraft and it's the only MCU with a SEX op-code 

The 1802 is only an MPU, not an MCU. Its not the only one with a SEX op-code. Several others, like the 6809, have them.

[floobydust]

OK fixed

[EEVblog]

Nevertheless, I maintain that integral to the impact and popularity of the 6502 was soundly anchored in it's low initial price.

It was. They famously sold them out of glass jars at a trade show for $25 a pop
https://spectrum.ieee.org/tech-history/silicon-revolution/chip-hall-of-fame-mos-technology-6502-microprocessor
Not sure if Chuck mentioned that on our podcast:
https://theamphour.com/241-an-interview-with-chuck-peddle-charismatic-chipmaking-coryphaeus/

[retrolefty]

Nevertheless, I maintain that integral to the impact and popularity of the 6502 was soundly anchored in it's low initial price.

It was. They famously sold them out of glass jars at a trade show for $25 a pop
https://spectrum.ieee.org/tech-history/silicon-revolution/chip-hall-of-fame-mos-technology-6502-microprocessor
Not sure if Chuck mentioned that on our podcast:
https://theamphour.com/241-an-interview-with-chuck-peddle-charismatic-chipmaking-coryphaeus/

 I was told by no other then the Woz was that the reason he chose the 6502 was indeed because he picked one up for $25 at a bay area computer show. I asked Steve Jobs the same question be gave me some story about upgrade potential, always the marketeer even at the start. This was when they were driving around the bay area selling their Apple (1) at small early computer groups and clubs.

[DrG]

Here is a very interesting interview with Ralph Ungermann, Federico Faggin, and Masatoshi Shima (3 of the 11 people that were Zilog in 1976) about the creation of the Z-80 and its early days. http://archive.computerhistory.org/resources/text/Oral_History/Zilog_Z80/102658073.05.01.pdf

Turns out that they did, indeed, sell a few of their first chips for US$200...

Ungermann:  We went after the PC business as we know it today.  And we brought our prices down to allow us to do that very quickly.  When we came out with the product, we priced the Z80 at $200 a chip, and it took us just a few days to understand from our customers that they needed a lower price to be able to go after the market.  However, we also had a number of system companies that were thought that they could get the IBM software to run on the Z80.  They could crack a lot of IBM markets, but that of course, didn’t ever happen, because the Z80 was not designed to compete with a mainframe.   

Faggin:  Yeah, one of the first customers that we had was Cromemco  In fact, Roger Mellon came to the office and I remember personally selling him one Z80 for $200, and Ralph didn’t say it, but NEC was actually our first customer, and I think you sold the Z80 to the competition.  <laughs>   


Ungermann:  When we were packaging the product up for shipment, somebody knocked on the door of the office and it was an NEC guy with four $100 bills for two chips.  And then right after that a very successful Japanese computer company came in with four $200 bills.  That was just the start of the copying.

There is an excerpt about their strategy against copying. I don't completely understand it (I don't understand it at all). Anybody that could explain a bit?

Faggin:  Yes, we were concerned about others copying the Z80.  So I was trying to figure what we could do that that would be effective, and that’s when I came across an idea that if we use the depletion load the mask that doesn’t leave any trace, then I could create depletion load devices that look like enhancement mode devices.  And by doing that we could trick the customer into believing that a certain logic was implemented, when it was not.  Then I told Shima, “Shima, this is the idea how to implement traps.  Put traps, you know, figure out how to do the worst possible traps that you can imagine,” and then Shima with his mind, that was steel mind, was able to actually figure out a bunch of traps that he could talk about.

Anyways, it is a long an interesting interview and kudos to Computer History Museum https://computerhistory.org/

[Kleinstein]

The anti copy traps somehow did not work that well - there was a 1:1 copy of the Z80 in eastern Germany.

[David Hess]

The 6502 is similar in speed to the Z80 at about 4 times the clock. AFAIK the 8080/8085 are a bit slower than the Z80. So performance wise  the 8085, 6800, 6502, Z80 were not that far apart.

The raw clock rate is not very informative because it is multiplied by the number of internal phases.  So a 4 MHz Z80 is comparable to a 2 MHz 6502 because the former uses 4 phases and the later uses 2, and then you have to include how many clock phases are used for memory access.  A 4 MHz 6502 would require much faster memory than a 4 MHz Z80.

And the 6502 and Z80 are not very different compared to some other processors of the time.  Some of the 8080 derived microcontrollers used 12 clock phases!

[floobydust]

Comparing CPU's throughput is very complicated. An entire university course on that, math between different architectures was no fun. RISC vs CISC and % of applications doing loops or bulk memory transfers, support for DRAM refresh also affects your benchmark results. It can be an endless argument.
Looking at how many clock cycles per instruction including the mux'd address/data bus, you still end up against how efficient code is with 16 registers verses say three A,X,Y.
I remember the 68000 is a slow dog with DTACK it needed an extra clock cycle- that sadly put it to pasture.

[Fungus]

So a 4 MHz Z80 is comparable to a 2 MHz 6502 because the former uses 4 phases and the later uses 2

IMHO a 4MHz Z80 would run like more like a 1.5MHz 6502, despite the 6502 only having 3 registers, etc.

[DrG]

I learned 6502 Asm before Z80 Asm and this was the early 80s. To me, the Z80 instruction set was richer, a little more complicated, but, again, to me, seemed much more powerful.

A big thrill was to run an asm program to white out the screen on the TRS-80 and compared to the BASIC version, did so in a flash.

I never had any real experience with the 8080, but the word at the time was that the Z80 was a souped up version of the 8080. My myopic hobbyist view at the time was that there was no point in learning 8080 with the Z80 and the TRS-80 M1. That later changed with the 8086/8088 and IBM and so on.

One point really stuck out and that was the Z80 “prime” registers. AF BC DE HL had ‘shadow registers’, AF’ BC’ DE’ and HL’ and you could exchange values between the registers using a single instruction (EX and EXX). Ideally suited for interrupt processing I guess. I only tested them and never really did use them otherwise, but I thought it was an extraordinary concept and I don’t remember hearing that any processors at the time had anything like that. Didn't know it then, but I was learning about banked registers.

I also found out, if I am recalling correctly, that the TRS-80 M1 ROMS never used the prime registers at all. Further, speaking of those ROMS, they were dissected at great length and that was exciting.

A few years later, I encountered the PIC and when looking at the instruction set at first, I just had a massive WTF? …but with some practice…those too started making sense….but they had so little memory 

[DrG]

More old man reminiscing…..Another big difference between the 6502 and the Z80 in my eyes was regarding, what hobbyists at the time called, “real world interfacing”. That was an incredibly exciting concept – you could turn stuff on and off. The 6502 was “stuck” with memory mapped I/O but the Z80 had an IN and OUT signal that you could use with the lower 8 bits of the address bus and you had 255 I/O ports without messing up the 64K address space.

When I entered Grad school I was introduced to these 6 foot racks of electromechanical equipment for instrumentation. We had to hand wire and make all of these lock up relays and pulse formers and pre-determining counters and a whole bunch of crap. People would walk by the lab and think that we were the University phone company from the cacophony of clicking and flashing lights.

The situation was a great motivator to figure out how a few lines of code and a little interfacing could replace all of those parts…and those racks became programs before graduating.

Chasing that capability was why I learned anything at all about TTL logic. The idea of tri-state logic seemed absolutely correct to me. The how of my learning was mostly due to folks like Steve Ciarcia and Don Lancaster and a host of others who wrote their hearts out.

Not being an EE and being completely immersed in a very different profession, I would, later, for relaxation, sequester myself in the back of the library (where the old books and some overstuffed furniture was [shout out to all those greatly tolerant librarians]) with BYTE and KILOBAUD (and also the mathematical recreations column of Scientific American) and all of the, relatively few, books that they had on the subject. It was fun in a very refreshing way.

OK, back to the present. 

[David Hess]

So a 4 MHz Z80 is comparable to a 2 MHz 6502 because the former uses 4 phases and the later uses 2

IMHO a 4MHz Z80 would run like more like a 1.5MHz 6502, despite the 6502 only having 3 registers, etc.

For simplicity I left out that some Z80 instructions are only 3 phases and left out ISA differences.  Based on memory cycle time which is a hard limit, A Z80 with a higher clock rate should be compared to the 6502.

This is also why older CPU microarchitectures do not scale up indefinitely.  At some point, a longer load-to-use latency is needed to work within the memory access time which requires a change in microarchitecture.

[floobydust]

What's this 6502 does pre-fetch? That has extra nerd appeal 

[HKJ]

The Z80 needed 4 clock cycles for the basic instructions, some of the more specials ones need up to 20 clock cycles. Then there was the block instructions that could use 1000's of clock cycles, but could be interrupted, they where extremely fast to move or search memory (for the time).

[Fungus]

The Z80 needed 4 clock cycles for the basic instructions, some of the more specials ones need up to 20 clock cycles.

A NOP on a Z80 was (is?) 4 clock cycles.

Then there was the block instructions that could use 1000's of clock cycles, but could be interrupted, they where extremely fast to move or search memory (for the time).

But it's much faster to do multiple LDI instructions than a single LDIR. 

LDI=16 cycles
LIDR=21 cycles per byte, except the last byte which is 16 cycles.

Moral: Unroll those loops!

[HKJ]

Then there was the block instructions that could use 1000's of clock cycles, but could be interrupted, they where extremely fast to move or search memory (for the time).

But it's much faster to do multiple LDI instructions than a single LDIR. 

LDI=16 cycles
LIDR=21 cycles per byte, except the last byte which is 16 cycles.

Moral: Unroll those loops!

Unroll a loop to scroll a 4kbyte screen memory? no thanks.

Try doing either of that above on a 8080 or 8085. It very long time ago I have played with either processor, but these types of operations was much faster on the Z80. It also had a nice 16bit register plus offset addressing mode (IX+offset), the instruction was slow (As far as I remember it was also about 20 clock cycles), but again faster than doing something similar on a 8080/8085.

[Fungus]

Moral: Unroll those loops!

Unroll a loop to scroll a 4kbyte screen memory? no thanks.

You can unroll it as far as 32 byte chunks (for example), the savings will be almost the same. 

Fun facts:
* On the computers with Z80+TMS9918 video chip (eg., MSX) the same trick can be used with OUTI instead of OUTIR to update the screen.
* Uridium (a C64 game) really did unroll a loop to copy an entire screen of data (40x25 = 1000 bytes)

It also had a nice 16bit register plus offset addressing mode (IX+offset), the instruction was slow (As far as I remember it was also about 20 clock cycles), but again faster than doing something similar on a 8080/8085.

The IX/IY addressing modes added 12 clock cycles vs. the equivalent HL modes (two extra bytes of instruction fetched, plus offset math). Definitely a thing to be avoided in loops.

[Bassman59]

Will a 1970's era Intel 8085 design kit power up after 40 years?
A look at the Intel MCS-85 System Design Kit and some vintage computer and processor history.

This exact kit was the standard for a lab class in computer architecture and program that I took as a junior at Stevens Institute of Technology (Hoboken, NJ) in 1986. Even though by this point "better" processors were widely available, for purposes of learning programming (in assembly, of course) and architecture it was fine.

Except there was the one time in class and my lab partner and I could not get the kit to do anything, and we soon came to the realization: "this thing is broken." Of course this was back in the days before everyone had a 'scope on their bench, and we didn't have any other such tools (not even a meter) in the classroom, so when we told the instructors that the kit was broken, they didn't believe us and told us our program didn't work. After a few heated minutes of back and forth, I finally said, "Let's take this kit down to the basement and let Chris look at it."

This basement lab was where proper benches were set up and you could check out test equipment (Simpson meters! HP 'scopes with tubes! Power supplies!). The person who ran the lab was this hulking big guy called Chris, who had this Van Dyke beard and always worse (American) football jerseys and sat at a desk inside the cage where the gear was kept. He was the sort of person who didn't take shit and had no time for idiots, but if you came to him with honest questions and concerns, he'd go over and above to help you. I knew him from various other classes and such, so when the instructors and I got to the lab with the 8085 kit and I said, "This doesn't work and they don't believe me," he immediately took it to the bench, pulled off the cover, plugged it in, and metered a few things and said, "he's right, the thing doesn't work." He then pulled another kit from a shelf and handed it to me. The instructors were pissed, of course, but since not finishing the day's project meant that your grade got docked, well, I didn't care. With the new kit in hand, I went back to the classroom, and finished the project.

And I've been playing with little micros ever since.

[alank2]

1A consumption!  Wow.  I really enjoyed the video.

I'm working on an emulation trainer based on an AVR mega1284 that will emulate the 8080/8085.  i chose the same 4x4 hex key layout - I just can't think of a better way to do it.