The 8-Bit Guy Builds A 16-Bit Computer

One of the better retro historians out there on YouTube is the 8-Bit Guy, and after years of wanting to do something like this, it’s finally happening. The 8-Bit Guy is building his dream computer, heavily inspired by the Commodore 64.

Before we go into what this computer will do and what capabilities it will have, it’s important to note the 8-Bit Guy is actually doing a bit of market and user research before dedicating a year or more to this project. He’s asked other famous retrocomputing YouTubers for their input on what their ‘dream’ retrocomputer should do, and they’ve come up with a basic list of requirements. The Dream Computer will be like working on a 1957 Chevy, in that all the registers are immediately available for peeking and poking. The computer will be completely comprehensible, in so far that one person can completely understand everything, from the individual logic gates inside the CPU to the architecture of the kernel. It’ll run BASIC.

In the age of the Raspberry Pi, one might ask, ‘why not go with a Raspberry Pi?’. To the 8-Bit Guy, the Pi is just a Linux computer. Other retrocomputing projects of a similar scope to this dream computer also fail: The Mega65, a project to resurrect the Commodore 65, will be too expensive. The BASIC Engine fails because it only does composite out, and it runs on an ESP anyway, so you’re shielded from the real hardware. The same problem exists with the Maximite in that the hardware is one layer of abstraction away from the interface. The C256 Foenix is probably the closest to meeting the design goals, but it’s far too expensive, and even without the MIDI ports, SID chips, and other interesting hardware, it would still be above the desired price point.

The ‘requirement’ for this dream computer is to use only modern parts, have VGA or HDMI video out, a real CPU, preferably a 6502, use no FPGA or microcontrollers, and can run Commodore Basic. Also, this computer would cost about $50, with $100 as the absolute, maximum limit (implying a BOM cost of around $15-$25). This is absolutely, completely, astonishingly impossible. I would be deceiving you if I did not mention the impossibility of this project happening with the stated goals. This project will not meet the goal of selling for less than one hundred dollars.

That said, there’s no harm in trying, so The 8-Bit Guy is currently working with a few dev boards, specifically one designed around the 65816 CPU. The 65816 is an interesting chip, in that it is a 6502 until you flip a bit in a register. It has a larger address space than the 6502, and everything from the World of Commodore should be (relatively) easily ported to the 65816. Why was this CPU never used in Commodore hardware? Because a Western Design Center sales guy told a Commodore engineer that Apple was using it in their next computer (the Apple IIgs). The option of Commodore ever using the ‘816 died then and there.

If you’d like to help out on this computer, there is a Facebook group for organizing the build. This Facebook group is a closed group, meaning you need a Facebook account to login. Unfortunate, but we’re looking forward to a year of updates around this dream computer. Building a computer that meets the specs is impossible, but we’re more than eager to see the community try.

81 thoughts on “The 8-Bit Guy Builds A 16-Bit Computer

  1. “The Dream Computer will be like working on a 1957 Chevy, in that all the registers are immediately available for peeking and poking. The computer will be completely comprehensible, in so far that one person can completely understand everything, from the individual logic gates inside the CPU to the architecture of the kernel. It’ll run BASIC.”

    Possible, but quite the hodgepodge.

  2. This project had me excited when I first heard about it but he immediately backpedalled and is using and FPGA for video out in the form of a gameduino. It was almost really cool

      1. As long as he’s breaking one of his rules I would have used a low cost microcontroller. They’re cheaper and easier for the average person to work with, not to mention to source.

        1. He’s now said that he doesn’t want any microcontrollers that are more powerful than the central processor. Which I personally think is shortsighted. Yeah, it’d be weird for support chips to be substantially more powerful than the CPU, but they’re cheap, they’re easy to program, and they’d act like a black box just like the support chips of old. Doing an FPGA has made the whole project more complicated for merely philisophical gain.

          1. It’s not too weird for support chips to be more powerful than the CPU…. *looks at my Ryzen tower and it’s 1080Ti*

            From a retro period… OK, maybe it is odd for a chipset to be more powerful than the CPU… but that depends on what an individual persons’ definition of “more powerful” is.
            i.e. Is being able to draw to screen/memory-manage/peripheral-interface more powerful than mathematics+compare+jump+etc?

          2. @fpgacomputer – he really wants it to be that specific processor. 16-bit version of what the C64 used. This whole project is largely a philisophical choice to begin with, so I can’t criticize it too much for certain choices. I would make different choices, but then, I’m not the one doing the work, either.

            Personally, I think what I want is a single-user version of Linux on the Pi with a lightweight console, access to scripting languages, and a basic graphics mode (just a framebuffer with a 2D drawing API). That would behave a lot like the instant on, complete control systems of old, but without their frustrating limitations.

          3. The TI-99/4 and TI-99/4A, when not running 9900 machine code programs compiled from assembly language, were essentially running the programs on the video display processor. The CPU only had direct access to 256 BYTES of RAM. So for BASIC and Extended BASIC, the 16K in the console was used by the VDP.

            Amiga computers had a gaggle of support chips that directly did things the CPU could not, while the CPU acted as a ‘conductor’ making sure orders were properly dispatched and the whole works operated in concert.

            Meanwhile, in x86 land, they concentrated on making the CPU more and more powerful to handle *everything* while fairly wimpy peripheral chips did as little as possible to shuffle data in and out from ports and storage, and out to the monitor. The cost was concentrated into the CPU while the support chips were inexpensive.

            Then someone hit on the concept of integrating nearly everything except the CPU and RAM into two or three chips, called it (and the company) VLSI or Very Large Scale Integration. (Crazy eh? Like if Dave Thomas had decided to call his franchise restaurant chain Franchise instead of Wendy’s.)

            That led to a very steep drop in prices for 80286 computers using VLSI’s chipsets due to being able to dramatically shrink the motherboard size by not having to make room for a ton of 7400 series TTL chips.

            Did VLSI get patents and license the concept of VLSI, or did other chip companies just copy VLSI and thumbed their nose at them?

          4. @frezik
            i guess you are correct. I was an old C64 guy and would love to see a modern C64 but I have also have no love of the 65xx line of CPUs. The 6809 was much better. What made the C64 great was, price, software base, the VICII and SID chips. What made it terrible was IO and CBM Basic 2.0.
            For this project I don’t get why they do not put a soft 6502 on the FPGA with the video system?
            It all comes down to the reason for the project. If it is to make a Commodore like 8/16 bit computer then this kind of makes sense, I thought that this was to make a spiritual 8bit like machine. For that I would have gone with an ARM Cortex M0 or M4 as the CPU and maybe the same FPGA gpu he is using. The key would be in making a simple single tasking OS and Basic that runs out of flash for the instant-on performance and general simplicity. The user could still learn programming and arm is not any worse than 6502 for machine code. More power to him and I hope for all the best.

      2. Tons of way of generating video:
        – propeller
        – old ZX80 uses CPU to push video data
        – old AVR (8515?) pushing video out with external shift register or AVR pushing video out with its SPI (lower res)
        – See my stm32f030 – VGA monochrome text project
        https://github.com/FPGA-Computer/STM32F030F4-VGA/tree/master/STM32F030F4/Hardware

        If I had to make my 8-bit computer with hands tied behind my back, I would have use one of the modern system on a chip Z80 with built-in DMA to push video to external shift register.

        1. To be frank, a propeller is basically everything he wants if he would have accepted a non-6502 archetechture. It’s a dead simple BOM, too; it just needs the propeller chip, EEPROM, crystal, and the connectors for SD, VGA, and a PS/2 keyboard. IIRC someone has already written a BASIC interpreter that runs directly on the prop. And given that the Propeller 2 is inching nearer to production, there is even an upgrade path.

          1. oh Wow.

            I didn’t realize they have released exactly what I wanted.

            It would be perfects for this if it still has the parallel bus. (I haven’t looked at the specs).

            Only problem is that it’s BGA only. Someone needs to cook some up on 24 pin DIP adapters.

    1. It still is really cool. Just because you don’t appreciate the choice of the gameduino implemented fpga (I’m just guessing that you don’t approve as you commented on it three independent times on this page out of seven comments) doesnt mean everyone does not. I for one am excited to get one!

    2. “This whole project is largely a philisophical choice to begin with”

      I learned machine language programming and microprocessor (MC6800) basics at the register level via the absolutely outstanding Heathkit ET-3400 trainer and course materials. I liked the 6800 architecture and instruction set the most among the 8-bit CPUs of the time, that CPU reaching perfection with the 6809. The 6502 also has a very simple and, therefore, easily understandable architecture. I hope a built-in assembler is an extended goal that he’ll implement beyond just peaks and pokes from BASIC. These days the “ROM memory” (flash) to hold such things is dirt cheap, so why not?:

      https://upload.wikimedia.org/wikipedia/commons/thumb/7/78/6502-architecture.svg/538px-6502-architecture.svg.png

  3. I was really interested in this project at first but he immediately went back on the ‘no FPGA’ thing and is using and FPGA based on a gameduino for the video card.

    1. He isn’t “using” anything, the entire system is still in the design and prototype stage.

      As the gameduino acts as a memory frame buffer to push out video, it’s probably one of the most trivial parts to pop out and replace later, only being topped by simple buttons and switches being easier.

      There is no gameduino in any design, there was only one on a prototype board to test a specific CPU that someone else whipped up in a couple days. Why the assumption this will be in the end product?
      Do you also install bench testing power supplies into all of your final products too? :P

  4. I think the criticism regarding video using the gameduino is a bit harsh… he’s not emulating the Vic.. it’s a stand-alone solution that uses the FPGA – (not being a hard core programmer, I don’t fully understand the pro’s/con’s of each component selection). I believe his original statement about not using the FPGA was in reference to using it to “clone” existing chips. I’ll have to watch his vid again to be sure, but that was my takeaway.

    Also, being a one man show – if he wants to complete this project in a reasonable timeframe, and without unlimited development resources, he has to cut a few corners. I’m sure if someone wants to implement a custom VIC substitute in off the shelf components, he would use that… but that alone would be a huge undertaking.

    1. He’s not using it to clone existing chips, just using the gameduino architecture. He also isn’t working alone, I know of at least two other major contributors to the project.

    2. It’s not a one man show. He’s been working on it behind the scenes a bit with a few others before posting the video announcing the project.

      I think the 8-Bit Guy underestimates the size of his fanbase. He did it with Planet X2, as well: announced it, and wasn’t prepared for how many people wanted it. Now he’s dealing with a flood of Facebook posts asking the same questions and making the same suggestions.

  5. I’ve just found his videos last week. I do like the whole instant-on idea. While I like having workstation-class computers for doing design, I adore small low-power computers, having grown up with Coco’s.

    What’s rarely made today are keyboard-computers, where the computer and keyboard are one. Perhaps a keyboard-box computer would be a good thing, and perhaps that is one thing that the Commodore 64s, Coco’s, Atari’s, and Sinclair, would agree on.

    1. Yes!

      Honestly, if one wants to make a system as if it were still 1988, there’s really no other way than FPGA, bcs you aren’t going to find the VLSI devices anymore, so I don’t understand the prejudice against them, nor the backlash when he inevitably walks back ‘no FPGA’ decision.

      But even if you *could* build from CRTCs and VICs and SIDs and VIAs and CIAs, and the like, and somehow didn’t have trouble packaging the beast, it would still miss about 60% of the total potential coolness factor the moment it gets coupled with a modern PC keyboard. “Oh look, another box connected to a VGA monitor and an ordinary USB keyboard”. Does it even matter if there’s a real 65816 in there, or just a Pi running an emulator?

      I think you nailed it. A great deal of what gave our 8 bit boxes personality was their particular cases and keyboards. This is probably the attraction of putting said Pi+emu into otherwise dead C= breadboxes and etc.

      So what I’d like to see is some kind of computer’s-in-the-keyboard configuration, like the C64 or TRS80 mod 1, (even arguably an Apple II), with actual key switches for authentic feel, and get the joy of being ‘on the metal’ by, yeah, doing the logic inside an FPGA. Or tie one hand behind yer back and use a few CPLD if that gets you closer to the ‘look ma, I grok the VLSI’ feel of using discrete ICs. There’s no magic in FPGAs, it’s not cheating, you can build yourself a 6502 in there, and a 6845 etc, and have access to all the registers too – registers that *you* built, even! You’d end up with a machine that feels authentically retro both physically under your hands and in terms of total accessibility. And ironically, it’d end up being the custom keyboard hardware that would be the biggest cost driver.

    2. You a aware of the Kiwi? Which is a keyboard-computer:
      https://hackaday.com/2013/02/27/the-coolest-homebrew-computer-gets-its-own-case/

      Anyway, the electronic design is one part. The design of a (cheap produciable) enclosure another. If the 8-bit guy has the final schematic the layout process starts.
      Maybe a keyboard-computer enclosure for a standard µATX board could be an independent bridge. It could be used with standard PC hardware and cheap because of the PC mass market. I dunno…

  6. re: the gameduino thing: he mentions in the video this is just for the sake of prototyping. He says he is planning to replace it with a better solution in the future.

      1. I think it’s the best choice to use the open source GameDuino architecture even if not the current hardware setup. I’ll explains in another post from a bigger screen with an actual keyboard.

    1. He hasn’t found a suitable replacement, so is planning to use the gameduino chip with a slightly modified firmware on the final board so far. The firmware will give it parallel-in from the CPU, a (partial) direct bitmap mode, and a Commodore style ‘character graphics’ mode, along with hardware sprites and scrolling and a neat tiny coprocessor thingy that could be very interesting if it ends up user-programmable.

      They’re removing the screenshot and sound functionality from the original gameduino code to free up a little space for more tricks.

      Just think of it as a custom video chip, just like the vast majority of 80s/90s computers and consoles. It’ll be like a souped-up SNES, with BASIC, and 16 gig of storage on an SD card…

      1. This is correct. While there are things about the Gameduino I am not a fan of, it’s a perfect choice really. Using an FPGA is the homebrew equivalent of working at a silicon fab, taping out your own design, just like Commodore did with the SID and VIC-II. There was a whole ideology argument about using FPGAs and they are fine in this design when used in a non-modifiable context. You can’t flash your own design into them whenever you want. Others have these morphable designs and they only have a niche market. My understanding is two FPGAs are being used, one for video generation, and another as a Chipset handling RAM, Chip Select Logic, etc.

  7. I like the project. I’ve suggested to use a bus system that you can easily expand it so it could also be someone elses dream computer :)

    I would prefer a Z80 processor myself as that is what I used to learn programming in assembler on the MSX1 and MSX2.

    I’ll be following his development with great interest in any case!

  8. I think that a lot of people are confusing “his” dream computer, with “their” dream computer.

    The only thing that I can add is that you should be pursuing dreams, not realizing them. Because if you do, then you must find a new dream, which could be more difficult then you think.

    I wish him the best of luck, although I don’t think he’ll need it as I’m pretty sure he’ll pull it off. This isn’t really about the hardware itself, it’s about the fanbase. Because that’s just what a new product needs in order to be accepted and make people writing software for it and a huge fanbase is what he already has because of his youtube channel. In a way this could be the BBC micro computer of the 80’s all over again, where a television station promoted a type of computer.

  9. Interesting idea. But I’m a little confused: why stick so close to the C64? That design wasn’t “perfect”, and if you’re going to build the Dream Computer, why limit yourself to a contemporary design from 1983?

    1. If you think about the goal of being able to understand your computer right out to every last register, the C64 isn’t such a bad choice. That, or maybe the original 68K based Mac. *Maybe* the first Amiga or Atari ST.

      Because most of what came after started getting really complicated. The original PC (or XT) was just about graspable in this way, but when 80286s and 68020s entered the scene, you’re looking at cache systems and virtual memory and sophisticated DMA, and really at that point things start to lose appeal to hardware designers (it starts to become such *work*) and is usually way more complicated than most software oriented types want to deal with. That’s when you start to encounter real OS+device driver issues.

      Before all that, you could pretty much just init your I/O devices and let them be. Which is why I was on the fence about including Amiga and AtariST – the 68K is straight forward enough, but the I/O in these was complex enough to be a challenge to master. But in terms of main CPU, the 6502 or 68K or 65816 don’t require mad scientist levels of devotion to master.

      Yeah I guess if your goal was to create the box you wish you had to *use* back then, you’d aim higher than C64, for sure. But if you’re after the machine you wanted to *develop* on, I can totally see eschewing the heavier hardware. That feeling that something Mysterious was going on, that I’d never figure out, kicked in around 1988.

      (Fwiw, lately it’s proven to be a lot of fun to get down to the metal on certain M4 SoCs)

  10. I hadn’t been to this site in months,,,,

    I see there’s still no real “hacks”

    Site looks more and more like google search result for “common hobbyist electronic projects that have been done a million times before”

    How in the heck does this site stay afloat??? I thought EEVBlog was bad but this is even worse… Blind leading blind leading deaf leading deaf?

    1. +1. It’s depressing.

      There’s a falafel restaurant near where I live that I stopped going to because the quality of the food started going downhill and they compensated by simply pilling more and more salad onto the plate, mentality being that so long as they gave you a mountain of food you couldn’t possibly complain. Feels like a metaphor for HAD. I remember the days where I used to read every single article on this site, and actively participate in discussions in the comments. Nowadays you make a single comment on a semi-interesting article that popped up on your feed and then when you come back to it a day later you have to dredge through all the pages of irrelevant crap that’s been posted in the interim trying to find that one article.

      Ordinarily I’d think “well, the huge amount of noise probably attracts more visitors and that’s why they do it”, but I suspect that’s not the case. I’ve been featured on HAD three times: the first time took down my site within minutes, the second time I was ready, but it still wound up costing me quite a bit in host fees. The third time I saw a momentary blip in traffic that lasted a couple of hours, and that was it.

      I know it’s much easier to trash something than it is to propose a solution, so my suggestion is this: go back to the good old days where only actual hacks and high-quality contents get featured on your RSS feed. Then expand/rework the “If You Missed It” section to include all the other stuff, which we can then casually glance through once we’re here. Either way, something’s gotta give.

      1. The thing is it’s a necessary evil. If sites like HAD (or any company, for that matter) didn’t ultimately wind up being bought out by bigger sponsors with different interests then people wouldn’t invest the time and money needed to start them up and nurture them through the financially tricky early years and we’d never wind up having sites like HAD to begin with.

        Anyways, I’m gonna go check out Hackaday.io for a bit, that’ll cheer me up.

  11. Sadly, I find this guy quite annoying. Sure, his heart is in the right place. He loves his technology (whatever it might be). But when you look at something like this you have to think – WHY? And then question his thinking.

    Let’s just take the Raspberry Pi. It’s NOT a Linux machine! Like anything of it’s kind, it is compute platform. Linux is an operating system that runs on a compute platform and is no doubt the cheapest way of achieving mass market appeal.

    Taking a step back and looking through his project goals, it wouldn’t be a far stretch to achieve most (if not all) of them using a Raspberry Pi – but instead of running Linux, look at programming it at the most basic level – bare metal. Yes, this means investing time into learning how to program bare metal – but guess what – isn’t THAT what he’s talking about? All the fun is learning a new platform, talking to the hardware directly – not through Linux. You want C64 basic then stick it on there!

    In a way, this is exactly what I’ve been through myself. I’ve created a few hardware projects based on 6502/Z80 and the biggest stumbling block has been modern video output. Want HDMI – get ready for pain and prepare yourself for using FPGA – you can’t avoid it. I have all the Gameduino’s and other types, but in the end it’s down to cost. You can’t easily create a retro platform that works on modern hardware and interfaces to modern equipment. Myself, I’ve moved on from the pain and I’m now happily creating retro-style arcade games on a Pi Zero in bare metal. It’s not hard and it’s extremely rewarding.

  12. If someone still makes the chips, he should use the PPU from the SNES, as its the perfect match for 65816.
    But I dont know if its possible to output vga without an converter chip.

    If he drops the 6502 compatibility from his wishlist, he could still use the Raspberry Pi, without linux,
    he would just need to make his “kernal” for the videocore processor.
    vga over dpi is possible, and if someone can get the damn hdmi state machine to work without an NDA also hdmi.

  13. Hey Andy, I see a lot of wisdom in your comment. Today there is soooooo much capable hardware; but the real impediment for the hobbyist/enthusiast is having guidance on getting into working on bare metal. Do you have some blogs on your Pi Zero work? Is it in Assembler or C? What suggestions would you have for others to follow in your footsteps? I am curious and intimidated at the same time.

  14. I put 8 bit per pixel video on the Apple IIe with VRAM and a couple PALs. This is easier on the 65816 because you don’t have page the RAM in as complicated a way. Waaaay back then there was a 64K bit VROM from Hitachi or someone (I’ll check) and I used 8 of then (one would allow a text display), some PALs for timing and to sync external video for capture (It was a frame grabber too) and 74HC parts to latch output and feed an R/2R network and an emitter follower. It was simply ‘OR’ed’ with video sink signals through the follower. this gave 256×256 with 8 bits of gray or a color mapped to it with a daughter board.

    Enough about me, that was just for bona-fides. You can likely get a single chip VRAM now that will do this, and data rate will be about 7MHz, easily allowing memory-mapped video without extra processing hardware. And memory-mapped I/O is the 6502 way.

  15. I fully expect that once the schematics are finalized, the entire system gets eaten by a medium size FPGA that can emulate the entire system before the real system exists, and will run faster and cost less.

    Nevertheless I think it will be interesting to see where this is going. If the project gets enough traction like the RC2014, it could be the start of a nice interesting community.

  16. Just design the thing on an FPGA to begin with. A simple 32 or 64-bit instruction set and perhaps some extra ram and/or rom chips. The whole allure of retro computers is their immediacy, accessibility, and platform consistency. 8 or 16 bits has little to do with it. It was not fun to chain operations for numbers greater than 255 back then and it would not be fun now.

    1. If you plan to sell a few of these “retro” thingies you have to account for cluelessness of your buyers, who want “real” chips.

      If I were looking for a real-chipped neo-retro wtfputer though, I’d specifically look for its video system to be “real”, because that’s what orders the music in those systems. Before designing some of my own though, I also used to think too much of the role of CPU, so it’s understandable that he plans his design from the wrong end.

      1. Agreed, pumping the video out was what gave shape to all those early designs. You could try to minimize your MSI (that’s 74xx, kiddies) count like Woz and end up with oddball memory mapping. You could try to be accommodating as possible of the CPU and end up with unimpressive video specs, seen in any 6809+6847 system. You could damn the video so your Z80 could full speed ahead and accept ‘snow’ in return. Or do like the C64 and insidiously steal the bus every 8th scan line to suck in another 40 bytes for the next character row. One side or the other always took a hit for the other, in some way.

        When you think about it, video (in that day anyway) was really nothing more than a synchronous serial output, running at a rate high enough to nearly saturate common RAM bandwidth, so of course it had to be front and center in the design process. (And a damned funny thing to do, too, considering that the overall rate of change in that signal was normally so *low*, but hey, CRTs, and I digress)

        So, yeah, it’s not Truly retro unless it’s going to output RS170 or NTSC or PAL, and drop signal into an oldey-timey particle accelerator display. No, I’m not trying for snark. If the goal is to get into that particular flavor of knowing every last corner, that decidedly driven by 120..250ns RAM and CRT limitations flavor, you’re right, it needs to reproduce this sort of trade-off.

        Unless we just missed it, and the ‘dream’ part of his ‘dream computer’ was having your CPU free to do CPU stuff, and getting an ideal video output, and not having your cycle counts blown by the video DMA.

        1. The irony of sending the same visual information over and over again while crippling the CPU. The only memory in a CRT was the fading phosphorus. And memory was still too expensive to let the monitor remember even one frame. For those two reasons alone, video output is considered demanding.

  17. The 6502 has been dumped for the 65816, the $50 goal has been dumped for $100, the graphics chip is an FPGA, and this whole project is being coordinated on Facebook. I have no idea whose dream this is anymore.

    1. The $50-$100 has always been the goal. The 65816 was decided well before the video. The reason is the ability to access over 64k w/o MMU. As for the Gameduino FPGA that was the original PoC choice. What has changed is someone provided the 8-bit parallel bus replacement. It can be accessed via pokes like the VIC-II. No other comparable solution so far has been close. That may change. Its early stages and things are very much in organization mode along with seeing where the interest lies.

  18. just make a cart for SNES and be done with it

    >Why was this CPU never used in Commodore hardware? Because a Western Design Center sales guy told a Commodore engineer that Apple was using it in their next computer (the Apple IIgs). The option of Commodore ever using the ‘816 died then and there.

    so thats why Commodore never used Motorola 68K! makes “perfect” sense :)

    No, Commodore didnt use 65816 because they had incompetent greedy management and no one at the steering wheel, they were busy selling scams like floppies with full computer build-in and C128D with 3 (three) CPUs (but computer would only ever use one at a time). Commodore was a company that acquired Amiga, and then proceeded to fire best engineering talent while manufacturing same computer (7MHz, double density floppy) for the next 9 years, and also tried selling $4K in 1992 money workstations with floppy spinning at half speed.

  19. What does this 65816 CPU cost? At octopart I find spools of wire and buckets of paint with that number, but no chips.
    If it’s cheap enough a multi processor solution may be an option.
    Use a 2nd chip to bitbang video and a 3rd for the sound, maybe a 4th for I/O or disk emulation.

    Oops: VGA = 640*480*50Hz = 15M pixels /s

    Seems this project wants to build someting like the C256 foeniks, wich is build on a 200x300mm 4-layer PCB ???
    Only way I can see this made for an affordable price is to put the whole thing in an FPGA.

    Future will tell if this 12-bit guy is either a dreamer or a wizard.

  20. In regards to the graphics, if all you want is simple VGA output (not hard to do) then the suggestions above (with the exclusion of the SNES PPU which isn’t made anymore, would only be useful if combined with the SNES CPU and its special DMA and interrupt and address hardware and doesn’t give you a framebuffer anyway) do make sense.

    But since he wants HDMI, he would need a HDMI support chip anyway (which pushes the cost up) so it makes more sense to have something like and FPGA to do both the video generation AND the HDMI.

  21. I don’t think there is a gate level circuit diagram for the 65816, so starting with this chip already fails the requirement “one person can completely understand everything, from the individual logic gates inside the CPU to the architecture of the kernel.”.

  22. PS: in the Facebook group someone said the core team has already implemented most of the graphics processor inside a FPGA, with 90 sprites per scanline and they are planning to use this. So much for “no FPGA”.

  23. Brian,
    You know something that would have been cool, is before saying the Foenix is “far too expensive” maybe it would be a good idea to actually ask how much it will be? David thinks it is too expensive compared to his dream computer, Lorin thinks it is too expensive hell, the Germans thinks it is too expensive… But none of those people, not one have asked the price, I mean the final price considering the different changes and against the possibility of having volumes. Actually, do you the price that Foenix will be sold?

  24. I get his philosophy, instant-on, and 65**, what he’s used to. But the project’s original constraints weren’t compatible with his own philosophy. The VIC chips *were* effectively what we consider modern-day GPUs. It’s ok they run faster than the CPU, and it doesn’t matter if it’s an FPGA or whatever.

    I would find a low-cost FGPA (e.g., Artix 7) and build a modern day equiv of a VIC (sprites, buffers, modes, scrolling) that has HDMI out (vGA is a poor choice, IMHO). The CPU choice doesn’t matter then. Personally I’d use a STM32 and write the kernel+basic directly on the metal of the SoC. This would give you 1) instant on, 2) on the metal, 3) modern parts, 4) the ability to continue to integrate with modern parts.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.