You remember last year. lol just busting your balls man

Yes.

I miss it a bit. I am glad that he has moved on to creating content he enjoys again, but I wish somebody would pick up the mantle.

@@MyScorpion42 Problem is more and more games not allowing editing of the game files, single player games mostly excluded, but even then they continue to make it more difficult, that and most of the methods are kinda same-same anyway since a lot of games are built on existing engines and all you need to do is find the right files. There's only so many different ways of explaining what the same options do and once people know what they do they can just edit it themselves.

Wait I remember those videos. Are they deleted and gone for good?

I am jealous. I was very sad to hear of his passing back then. I am glad you got to meet him.

My first 6502 was an Ohio Scientific single board computer kit. The graphics weren't great, but good enough that I could write fun games in BASIC. My next computers were Commodores and they had much better graphics systems. Rhe graphic sprites and sound system on the C64 were great.

I had the VIM-1 which was a improved KIM-1 they said if you like KIM you will love VIM I had to program it with hexadecimal numbers (just like the KIM)

Admittedly, there were only 150+ opcodes to know.

you have a long life, wow! nice

@@disgruntledtoons you must be fun at parties lol

Thanks for sharing.

So basically: You didn’t take as long to make the chip as us, therefore you stole from us. How does one win a lawsuit on that?

@@engineerbot That was page one in the lawsuit. There were many more pages I never read. MOS technology ran out of money. As I remember it the first lawyer bill was $400k. They could not pay a second one like that so they had to settle. Someone else will have to post the terms of the settlement. The possession of damaging material by one of the former Motorola employees, even tho not used or even seen by others, would have made further litigation extremely expensive. Motorola could pay it. MOS Technology could not. I was not privy to these discussions.

@@harryb1251 that's interesting.

@@harryb1251 That's the basic model of operations for large businesses. Make it impossible for your smaller competitor to actually fight it in court even if you're in the wrong because you have an entire office floor for your lawyers on retainer and the small business just doesn't have the liquidity to fight you.

I was working from his characterization on that same biography so I am happy it came through

​@@LowSpecGamer Tramiel (BTW drop the i, it sounds like Tra-MEL per the man himself) was likewise well known for being fiery AF, that negotiation would have been legendary. He was a notorious micromanager and the friction between him and the chairman of the board Irving Gould was not great even though his performance had made Commodore immensely profitable. It's honestly weird looking back at the 70s and 80s in computing and realising just how much larger than life all these people were in the industry, so many hard nosed characters dictating the way the industry would evolve.

Im 100 like

Jack was the kind of person most people would dislike, I bet. There were many jerks in that generation of tech CEOs (counting Jobs), but he took the biscuit.

@@ButterfatFarms - *_I don't think he cared what he was selling in the end. It eventually turned out to be consumers electronics, calculators, watches, computers and a few other odd niches._* When I was in my teens -- the mid-1960s -- my father gave me a small "portable" B&W TV. The brand was "Commodore"! After a year or two, the tuner developed a malfunction. This was quite common with the "electro-mechanical" tuners of the day. There were even _stores_ that sold nothing other than tuner parts (for the TV repair trade) -- I remember visiting one in the Bronx, near Yankee Stadium, to buy parts for my mother's Zenith TV (also B&W, as were most televisions of that, to get parts for the _Commodore_ TV, I had to visit the sole supplier, _Commodore,_ which was located at that time deep in the heart of Brooklyn. So, I got onto the subway and began my _long_ travel. When I arrived, I found a place that was like a big warehouse/garage type operation. Definitely _not_ an ounce of "retail" to be found. I bought my parts and was on my way. A few years later, the "pocket calculator" was born. Nifty, but expensive -- something like $400 in 1965 dollars (or whatever year it was in the mid-60s, I can[t remember). But then an inexpensive brand was born, and my father got one for me. Yup, "Commodore" (I have to wonder if the fact that they were located in Brooklyn, and he worked at the Brooklyn Navy Yard, had anything to do with his purchases). Ah, memories... PS: I had the 2nd or 3rd TRS-80 sold in my region. Went in for warranty service nearly 20 times, five main boards, and still ran like a one-legged blind man on a pogo stick. PPS: The one shown in the photo in the video is the Model 3, which came out a few years after the TRS-80 (retro-named the "Model 1").

You were one up on me. I couldn't stop my pacman eating the frickin maze !!!! ;-)

Should have prototyped the logic in something like BASIC and then converted it. I couldn't even afford an assembler so I wrote my own BASIC program that converted z80 Assembler instructions (ZX81) into machine instructions that were POKEd directly into memory. I did intend to go a step further and create a BASIC > assembler > machine instructions compiler but I discovered drink and girls and that was the end of that! :)

For me again it was the 6502, writing demo's on the C64 in the late 80's. Not even using an assembler, just punching hex into memory using a monitor.... lol I managed to learn every hex code for every instruction and could just sit their typing away, hex after hex after hex for hours on end. Was 14 or 15 at the time.

Ahh those were the days. I was also writing 6502 assembly in the 80's and x86 by the end of the 80's.

​@@JoannaHammond. Me too. Punching hex into a keypad - I couldn't afford the full keyboard for my 6502 computer until much later. Codes such as A9 (LDA) are burned into my brain.

Too bad the Z80 half was useless by the time the Commodore 128 came out.

And they wouldn't have bothered except they found the Z80 cartridge for the 64 wouldn't work on the 128, and that incompatibility was considered a bug that they had to fix

I personally don’t like this direction,what’s that all that sarcasm funny generic talk and all those drawings,it annoys me,as if he tries to make funny out of those incedible stories to force you to take it with a grain of salt,am sorry but if i want to have some fun i will watch a comedy but if i want to watch a history about the past whether it’s politics,economy or about video games,i don’t wanna any genericness in it because i consider that BS.

Motorola has been around since 1928 originally started as Galvin Manufacturing Corporation. In 1930 they sold their first Motorola branded product a Car Radio named Motorola a combination of the terms Motor and 'ola' from the then at the time popular Victrola radio name. And well the rest is history.

@@cybercat1531 after receiving a $500 investment to start Motorola the founders allowed the investors to drive around the block one more time in the car in which the new Motorola radio had been mounted for demonstration! The car caught fire and burned to the ground but the investors had already coughed up the check!

The name is a clue, it's a characteristically old-timey American name lol.

You think that's crazy? Nintendo was founded in 1889

@@xSaintxSmithx You think that's crazy? Nokia was founded in 1865

And unmentioned here, MOS second sourced the Z80 for a time. My first computer was a MOS corp Z80 (hand built by the way).

@@scottfranco1962 I doubt that. MOS were probably not able to put the 8500 (fast) transistors of the Z80 design on a single chip for a reasonable price. Maybe you are confusing MOS with Mostek. Mostek and Synertek were andvanced firms that built the first Z80 chips before startup Zilog had its first plant ready. They continued as second sources for years. The European second source for Z80 was SGS-Thomson (now ST). After a few years (late 1970s) there were also clones from NEC, Sharp, Toshiba, Rohm, GoldStar/LG, Hitachi, and others.

@@herrbonk3635 Correct, it was Mostek.

Don't forget Acorn. Without them we're probably still marveling over flip-phones.

@@shaunnichols1743 Never heard of them before. looked it up. RISC and ARM are something i also never knew before. Thanks

What does it say? I can't see it.

What does it say? My screen is bad so I can't read it.

It says "If you get clocks for less than $5, buy the mos6501 and give your purchasing agent a bonus."

It would be much longer until computers would hit the consumer market. Computers would of continued to be big business machines. Its nuts to think about. It was the perfect invention at the perfect point in history.

It wouldn't have stopped Woz. He'd have built something around the 8080 instead

Don't forget that the Nintendo Entertainment System and Super Nintendo used variants of the 6502.

and no Atari 400, 800, BB Micro, PET, Vic-20, Atari 7800, Acorn, .... (Even the Lynx and TG-16 use variants of the 6502).

I still got lots to improve. It takes time to build a new brand. You can help me by sharing it!

@@LowSpecGamer what could you possibly improve other then your pronunciation (no offense intended, but that's literally the only thing I can think of that could be better, I'm genuinely asking what needs to improve.) These videos are some of my favorite content on the platform, your style is just so great. To be honest, you found and nailed your own unique form of presentation.

@@MrGamelover23 If you actually mean his accent when criticising his pronouciation I'd not have it changed in any way shape or form. He is perfectly comprehensible and on top of that very distinguishable in tone and accent! It adds character! I'd rather have profound substance in content and production value. Since years I'm still here and excited over every video he releases. Other creator who are obsessed over perfect pronunciation and tone (aka like to here themselves talk) but seriously lacking in substance of their content (like e.g. coreteks) quickly lost my interest.

@@glockmanish hey don't get me wrong, this guy's videos are amazing, but I still need subtitles for some parts. This guy makes some of the best content for platform's ever had. I was just asking what could possibly be improved since he said he has a lot to improve?

@@LowSpecGamer hey, I apologize if my earlier comment was insulting or offensive or hurtful in any way shape or form. It was not my intention to be rude or hurtful and I hope that you weren't hurt.

It is awesome that 6502 accepts its 8bit nature and has a ton of instructions which work with 8bit immediates. No weird register pairs. Instead 16 bit values are addressed as memory, just like strings. Clean. I like that little Endian works with the way immediates are read and the way pointers in the zero page are read. Yeah, Intel had carry look ahead, but 8 but ADC saves a few transistors. I think that MOS invested a lot of transistors into cycle efficiency. Unlike RCA 1802 or Intel, a lot of stuff could be done. We have like 4 busses and a huge PLA. Only later the MOS fab produced better and better chips, and soon their clock frequency wasn’t the limit anymore. Rather in the C64 the complex system bus limited speed.

He was a colorful character. I respect his shrewdness if anything

I think he was just not a “visionary” but a good “businessman” that really kept an eye on what actually made his company money. Many companies over the years have gone bust because they lose sight of what their core business is and Jack needed definitive economic proof that computers were a worthy venture before committing.

@@LowSpecGamer and 3 different companies wanting one isn't proof?

"Business is war" Jack Tramiel

@Alfred Wedmore 😄

@@LowSpecGamer In the end his style of doing business also was the downfall of Jack! Nobody who ever had to deal with him businesswise wanted to do it a second time, just look at the Amiga guys how desperate they were not getting into Tramiels clutches and were saved last minute by Commodore from that fate. Also just look how Epyx was screwed over by Tramiel on the Lynx and basically went bankrupt because of it! In the end Atari under Tramiel stood alone and no one wanted to deal with them anymore.

yes and no -- the whole point of these cores was they were cheap enough to be used anywhere. Even while IBM PC's were being powered bu the 8088, keyboards were already being powered by the simple-but-not-THAT-simple 8085. There were actually smaller alternatives to the 8-bit boys but nobody remembers them now. Basically calculator chips with minor modifications. Slow mask-programmed 4-bit behemoths found in a lot of instrumentation until they were replaced by application specific chips.

Commodore truly does not get enough credit. It is a shame they did not survive through today like Apple, Microsoft or Intel

The Amiga was not a Commodore computer. It broke with all the fine traditions: no instantly ready BASIC in ROM, no PETSCII, no backwards compatibility. Not surprising because it was not made by Commodore but was bought in. They should have made an own 16-bit computer backwards compatible with the C64 instead - that could have been an IBM PC (clone) killer.

@@NuntiusLegis It wouldn't have mattered.. The PC didn't win because it was the best, or the most fully-featured. It won because it was the *default.* It didn't matter how good Commodore made the Amiga or a hypothetical C256 or the like, because only they were making it. Just like Apple was the only ones making Macintoshes. *Everyone else was making IBM Clones.* So programmers made more IBM-Compatible software. And in the end, all the fancy hardware in the world can't change the fact that *software* drives hardware sales.

​@@watchm4kerWith the C64, Commodore had achieved the world market leadership for micro/home/personal computers, outselling IBM PC clones, Macs, and everything else by far. The C64 was the default. IBM PC clones won the race in the end, because they were continuously developed further in a backwards-compatible way. Had Commodore done the same in time, our current PCs would be descendants of the C64.

@@NuntiusLegis No. They wouldn't. Because companies couldn't make C64 compatibles without running afoul of Commodore. Nobody else could make a C64. Nobody else could make an Apple II. Nobody else could make an Atari 400. Or an Adam. Or an Amstrad. Or an Acorn. But they *could* make a computer that was just as good, if not better, than IBM, and let someone at home use the same software they used at work. Commodore, Apple, Atari, and Acorn were not competing against another company. They were competing against an entire industry, and one they could not compete *within.* (See what happened later with Commodore) If you want a counterargument, though? *The MSX.* That was a computer system designed specifically for licensed manufacture by multiple firms. Sony, Toshiba, Panasonic, and more could make MSX computers, all compatible, all interoperable. Had it been pushed harder in the US, and had Zilog been able to keep pace with Intel's hardware, it might have become the standard for home computers, worldwide. The only company that could stand up to the IBM PC was Apple, and that took a truly Herculean effort to achieve, carving out a niche as the graphics computer of choice for anyone working in art, design, or publishing. Even then, they tried to grow the market by attempting to license out hardware manufacture, but it was too late for anyone to really care. Amiga found a smaller niche as the first low-cost video graphics machine, and it held on amazingly well. Of course the company that really got the last laugh over them *all*... is Acorn.

If anything I feel like it’s not only not left us but is only getting much more common. Now that processors are this powerful in general even the low spec ones blow stuff from the not so distance past out of the water

Isn’t there this indexed zero page indirect addressing mode, which no other CPU has? It seems to be used for Demo effects only. It is like if on a 68k you could use the 3 lsb of a data register to select one off the 8 address registers to use in a reg-mem instruction. This is insane. I guess 6502 cannot move addresses between address registers like a 68k. I Wonder how you teach a C compiler to use IDs in Y instead of pointers? And use zeroPage+X for lists. Global optimization? Seems like C is the wrong language for 6502. Similarly, why is it so difficult to put pointers to the stack frame on the call stack? Ah, no addressing mode for this. 6502 cannot even peek SP+signed immediate8 .

objective oriented programming is a fucking lie... hopefully in the next few years we see a trend in indie game developers of coming back to Functional Programming (all functions inside a single file of code, code easier to read, etc)

@@FeelingShred That's a bit random and absolutely nothing to do with my original comment.

@@mandolinic haha Look, another C++ advocate LOL 🤣🤣🤣

@@FeelingShred C++? Yeuk.

@@FeelingShred it is good to have a data structure of some sort... but OOP mostly means "37 different warring designer's data structures in a mangled glob that barely functions". Just give people the data, in a simple flat structure, and they can do anything in any language

We had a lugable Kaypro 2. I loved playing text adventures on it.

@@hicknopunk like I said above, you can find ladders in a .jar file. Did you ever play a game that was a galactic economy Sim? Now THAT was a fun game. Wish I could remember the name of it and find a copy in a jar as well. Good old kaypro. I remember one time my dad took it on a business trip and a coworker of his that had the same machine set it up to allow me and my dad to instant message each other, in like 1985. Take that instant messenger.

@@jonathonschott the games I remember most were infocom games like Zork, Planetfall, Hitchhikers Guide, etc. I also had arcade games made up of all text characters like centipede, pac man, some racing game. I was a kid using it, so I mostly remember playing text adventures at night before my dad had to take the computer back to work on monday

I wish processors were still in that margin, with 1 dollar per core. 8 core processor? 10 dollars

Not sure, if there was no 6502 everbody and his kitchen sink would have used the z80 instead!

Sounds about right

@@LowSpecGamer Yes, Atari was still reeling from the crash of '84, and had already missed the boat on what would become the NES. They were DONE, and their parent company wanted them gone.

I am going from username alone but are your Harry Bawcom? I am super glad to hear you enjoyed the video. Hopefully you noticed the piece our artist did at 7:28 with the original Motorola team that should have you on it (and we would love to send you a print of it, if interested). Also, let me know if you are ever available for interviews. I would to know more about these events or what you did afterwards! I am available at alex@lowspecgamer.tv

Registers are called global variables by some. Similar to the zero page. So with a bad enough coding style you use them up quickly. C and Pascal have structure and a compiler can check spans without any calls and then check for non-overlapping lifetimes of variables ( first and last access ). This can be expanded to parameters and return values. Even in 1993 AtariJaguar came with scratchPad memory which forces you to load small, but mostly complete programs into 4k of SRAM ( like on PET ). And you load data into your 64 registers, which are not that global anymore. Sad, to use a compiler only for those snippets. You have to link them all using a custom tool chain. Ah, and no subroutines allowed. I mean prohibitive slow. All macros and loops.

@@ArneChristianRosenfeldt tons of registers make programs also faster, you do not have to performan constant on off stack operations for storing away data. There is nothing worse than having only a small number of registers, well yes there is something worse, having a small number of registers, and half of them dedicated to just one single task to skimp on internal wiring logic!

@@werpu12 6502 has the decimal logic wired to the Akkumulator, but otherwise it has a register file just like z80 . ARM, MIPS, JRISC, and SNES have special registers for multiply and accumulate. So basically, just as 6502 they have an accumulator. I always thought that maybe they keep carries at every fourth position like Z80 ALU? Also MAC already reads two source registers per cycle. You want two read ports on the file? I may want to add one bidirectional port to also allow two register barrel shift. And division is just slow, so indeed it uses hidden registers. No need to expose those in the ISA, if you have a scoreboard as JRISC does. MIPS does not. MIPS blocks if memory access takes longer than a cycle (plus addressing mode, 5 stage pipeline for memory vs 3 for reg-reg, so actually two cycles? MUL also fits in three cycles ( no addressing mode )). MIPS delegates multi cycle stuff to a coprocessor. Yeah, int, float, and vectors are not that different. For vector add, just block the carry. Same for exponent and mantissa. Modern, deep pipelines can deal with instructions of different execution time. So 64bit pointers, double floats, and MMX all in the same registers! So you don’t like the flag register? RISCV is for you! I think that the forth port to the file could be used to set or get a carry from to any register for ADC. Power has 6 flag registers. I just want them aligned to the file. Stack pointer is a special purpose register as is instruction pointer, but these are loved.

Today's PCs are just a lot of processor cores, RAM, ROM, and a keyboard. Even GPUs became just more processor cores. Wozniak was right all along: hardware's job is to run software.

@@ischmidt Try to us your smartphone to control a 3d printer, try to use your smartphone to write a real program, and you'll see the difference between a real system with expandable IO and the disposable tinkertoys without it. IO is so easy to overlook but so important.

​@@tsm688Smartphones can do both of those things, and they can do them comfortably when you plug in a USB-C hub to add a keyboard, mouse, and display. The entire point of the Arduino and Raspberry Pi ecosystems is precisely that a lot of I/O oriented stuff doesn't need a PC.

@@ischmidt on a platform where programming languages are literally banned? The best and nearly only way to program for phone, is to not use phone to do it...

I was really, really hoping someone would catch that. I am glad it was you.

Yes! There are ton of small technical stories, like the process they had to go through to create the masks, that I just could not get into for lack of time

I don’t get this. I thought that RISC has a lot of registers in order not to access memory for data. Large registers and barrel shift to pack data. RISC replaces the microcode ROM with cache RAM for code. RISC tries to avoid wait states, but always had allowed for them in case of a cache miss. MIPS was an experiment to see how we can reduce wait states as low as possible. Just consider DRAM and the address multiplexing. It is kinda double data rate. When MIPS was designed, a load or code fetch was still single cycle. When it came to market in 1987, fast page memory introduced variable latency.

​@@ArneChristianRosenfeldt Large registers can only help eliminating load/store operations by few percents not completely, also "RISC replaces the microcode ROM with cache RAM for code." wouldn't that exacerbate the problem, since microcode ROM is faster(at that time) and directly wired to ALUs and can do multiple operations in parallel?

@@niks660097 I don’t get the few percent thing. I may be biased. I try to write assembler for AtariJaguar. Load Store does not block there and two instructions are loaded from RAM per cycle. So as long as I have reg-reg in between, von Neumann goes brrrrr. And a lot of code snippets I saw and my own ideas have a lot of reg-reg instructions. JRISC is cheap and only has one instruction format ( 6Operarstion:5src:5dst ). So there is a ROM which translates the Operation to the control lines. It only has 64 entries because there are only a few instructions and each is single cycle. So and that was a lie. JRISC has multicycle instructions also. RISCV has different instruction formats which slow down the decoding a bit. Still fits in a single cycle for a low enough clock rate. I don’t understand what 6502 does with the instruction opcode before the PLA. I thought it is so smart to emphasise decoding speed. So the opcode is latched and then using large transistors with lots of fan out blasted over the 128 transistors of the PLA for maximum parallel processing. For extra speed this is even a balanced signal almost like ECL. But why the same apply the same urgency for the following cycles of the instruction?

That was my first computer, too. Bought as a kit in 1979. I am now building a replica of it using the 65c02 and modern memory chips (32kb in one chip is pure luxury, compared with the 2 chips needed for 1kb we had originally).

Ooo, me too but 4k (to start with)

@@christopherlawley1842 I put 16k RAM on mine, basically by soldering an extra 8k on top of the first 8k of ram, worked fine; even with overclocking it x2 :)

I know, I had to simplify to not add yet another tangent to the video. Since it is a variation of 6502 it still works within the narrative of the video

I can only find resources saying it picked up the one in the name some short time after the initial launch. Looks like in some official context it didn’t always have the one as most people would expect

I knew parts of the ARM story but I had NO IDEA they directly took ideas from the 6502. That is brilliant. I am totally going to add that to my eventual ARM video. Thank you for telling me.

I think I've seen a video with Sophie Wilson say that she was inspired by the 6502 when designing the ARM.

I GOT to make that video eventually

6502 sets flag on load. ARM does too? But is totally the wrong thing for RISC and load store. Even flags are not so great of an idea. Now JRISC got all those variants ADD, ADC, ADDtransparent to denote if flags are read or written. Power even has 8 flag registers. Or is this okay? After all CPUs have float regs and vector regs. Vec8 could write all 8 flag registers.

@@ArneChristianRosenfeldt All RISC becomes CISC once you discover enough needs

6502 needs fast memory and originally could not accept another bus master on the 16 bit address bus and only had limited current output. So you better pair it with fast = slow RAM and add bridges for all bus pins. I wonder how the bridge can be fast enough to tell the SRAM that write enable is not meant for it? Address comes a bit before enable, so there is that. I still Wonder if there should be a shadow copy of SRAM in the DRAM extension to correct misguided writes? Kinda write through cache. Write only happen at half the clock rate, or do they? Push program counter is faster, but doesn’t leave SRAM. DRAM reads need to prolong phase 1 of the CPU.

Ricoh in Japan also based its processors used in the NES and Super NES on the 6502.

About the only maxed-out 128 application was some high flying business BBS software. Trying to sell CPM, in 1985!? They really did not read their market for that one.

1080 60 is low end. Mid level is at least double that. Even the 2060, a 3 year old budget card, runs games at 1080 100+

@@TheLongDon well top end pheonix should be around a 2060 apparently. I don’t think a 2060 could play anything you throw at it at 1080 60 tho.

I hyphenated low-mid not meaning low through mid but in the sense that it would replace lower mid range cards like the 2060 which is not a low end card imo. An mx450 or 1660 would be a low end card. Mid range cards would be 60&70 series imo. In retrospect, i may have been a bit too hype saying low-mid. It will only replace lower end cards.

@@TheLongDon What parallel universe are you from that 1080 60 is low end? I want to go to that universe without the chip shortage.

@@TheLongDon depends on game, E sports yes, AAA it's well within mid-level

Pretty "low spec" with just a 8-bit stack pointer though. And memory couldn't keep up. Because the 6502 did not use the available memory speed very efficiently; only a fraction of a cycle was allowed for memory access, due to the transistor level design. A 6502 at 1MHz needed a 300ns memory, which is less than 1/3 of a clock cycle. In contrast, the Z80 allowed two full clock cycles for memory to respond. So using the same speed memory, the Z80 could be clocked around six times as fast as the 6502. This meant that ordinary 8-bit operations as well as normal jumps, calls, returns, etc. could be 2-3 times as quick on the Z80, using the exact same speed (and therefore cost) RAM and ROM. But the big difference was in 16-bit and 32-bit arithmetics (such as for floating point). The Z80 could do a 16-bit addition in 11 cycles, while the 6502 needed at least 20 (much longer!) cycles for the same thing. Being able to be clocked faster, less cycles for 16-bit add/sub, and several internal 16-bit registers meant that the Z80 could be about 12 times as fast as the 6502 in practical "number crunching" tasks. Again using the same speed, (cost) memories.

@Jimbo Bimbo But Elite was first written for the BBC Micro - a 6502 machine

@@herrbonk3635 The practical speed difference is nowhere near what you indicate. z80 11 cycles for a 16 bit add is only for adding bc, de, or the stack pointer to hl (or adding hl to itself). Doing the same adds to ix or iy is 15 cycles. And you have to get the data into the right registers somehow to start with. With as few registers as both the z80 and 6502 have, that usually means the data comes from and goes to RAM. The 20 cycles for the 6502 is to add an arbitrary pair of zero page locations (there are 256 of them) to another arbitrary pair of zero page locations, and store the result in possibly yet a 3rd arbitrary pair of zero page locations. So your 20 cycles gets you the same functionality as a 16 bit machine with 128 registers and 3-address instructions. To do something as flexible on z80 requires e.g. "ld hl,($nnnn); ld bc ($nnnn); add hl,bc; ld ($nnnn),hl" which is 16+20+11+16 = 63 cycles and 11 bytes of code, vs 6502 20 cycles and 13 bytes of code. Sure, if you can keep all your data in registers on the z80 it can fly, and it does have more registers than the 6502, but it doesn't have *enough*, usually. 6502's 256 bytes of Zero Page is enough "registers" for anything you can imagine -- the same as a modern 64 bit machine with 32 registers e.g. RISC-V or ARMv8. I found that typically over a wide range of programs a z80 needs three times more clock cycles than a 6502. z80s generally ran at 4 MHz, while 6502s were either 1 MHz (a little slower than a z80) or 2 MHz (significantly faster than a z80).

@@BruceHoult No, it does not mean that it comes from memory, that was one of the main points iirc. Firstly, no one uses IX/IY for this purpose, other than perhaps a naive beginner in a school assignment. Again, even with zero page addressing only, the typical 1 MHz 6502/10 still needs 20 cycles = 20us for a 16 bit addition or subtraction. During those 20 clocks, the typical 4 MHz Z80 had 80 cycles to spend. Floating point routines for the Z80 could store up to three 32-bit mantissas in HL'HL, DE'DE and BC'BC. With exponents in A and A' and/or B,C. This was very useful for repeated shifts and adds, such as in mul and div. So these could therefore be performed with almost no memory data traffic at all. So very unlike the 6502 in this regard. The Z80 EXX-instruction was needed ONCE per 32-bit operation (via carry) and took the same time as *ONE* clock cycle on the 6502! The Z80 sure needed a different crystal than the 6502, but, again, it could cram more power out of the same memory ICs than the 6502 could, especially in 16/32-bit calculations. Memory speed was the main (economical) limiting factor in the 1980s. Even an old "4 MHz" rated Z80 chip could actually be run at 6, 8 or 10 MHz, when coupled to the faster memory ICs that became common in the latter part of the 1980s. The clock speed difference didn't even need to be large. The ABC 800 (based on a 3 MHz Z80) had a significantly faster BASIC than the pretty optimized BASIC for the BBC Micro (based on a 2 MHz 6502). The latter was in itself known to be quicker than say M$ BASIC for 6502. en.wikipedia.org/wiki/Rugg/Feldman_benchmarks

@@herrbonk3635 Thanks for the link! Unfortunately it seems you didn't read it, or look at the table of results: "They conclude that the 6502 is the highest performing of the CPUs, agreeing with comments Gates had made in his letter." Just as you apparently didn't read or didn't understand my comment. I will repeat and rephrase: there are micro-benchmarks (individual loops or functions with few variables) where the z80 is faster because it can keep everything in registers, but this doesn't translate to being faster on entire programs, and certainly not by the margins you claim.