This 8-Bit Computer Has Two CPUs (And One Acts As A GPU!)

[@nickolasgaspar9660]

Back in the 1979 Jay Miner implemented that exact solution (Atari 8bit line) through the ANTIC chip which it came with its own set of instructions. In essence it was a second custom cpu designed to produce graphics and feed them to the video chip (CTIA/GTIA). This shows how much ahead Miner's ideas were!

[@Cybergorf]

He later did it again with the "Copper" for the Amiga - reducing the instruction set to only 3 types: move a value to a register, wait for a position on screen and a loop instruction. these little programs, called "copper lists" run independent of the CPU and typically repeat itself for every screen (50/60 times per second)

[@markwiygul6356]

And that's how the Atari 130XE could address up to 128KB of memory: ANTIC addressed 64K and the 6502C addressed 64K. Programming could only use the ANTIC's memory by programing memory banks, swapping them to the 6502C when it was needed. Atari Writer Plus could use the extra memory, so could XE-Term, a modem program, and DO 2.5 could use it as a RAM-Disk. Otherwise, since the 130XE was the final advance forward for the 8bit line, games couldn't utilize it, nor most of other software which was written for compatibility for the entire line, without any bonuses for the 130XE

[@old_liquid]

Its also alike Nintendo and Sega did with their graphics addons for SNES/MD cartridges. Nintendo used ARM cpu for graphics and Sega used Samsung CPU.

[@bobjones-ey5gl]

@@Cybergorf What was "Copper" ? was it part of Agnus ? along with the blitter ?

[@bangerbangerbro]

@@old_liquid Just because it was designed by a guy in Cambridge doesn't make it ARM, however it was almost certainly inspired by it, more basic though, evolved from his Konix DSP probably, as that is where Argonaut recruited him from (Ben Cheese).

[@GianmarioScotti]

My favourite kind of vintage computer video - about a lesser known home computer!!

[@CurtisBoyle]

The FM-7 had a really excellent version of the OS-9 operating system as well (multi-user, multi-tasking, based loosely on Unix), like the Dragon 64 and Coco 1/2/3's had. And it used the dual CPU as well, which would be worth taking a look at if you can get a copy onto it.

[@NoelsRetroLab]

Yes, absolutely! I definitely want to check that out as soon as I get some good way to load programs and the OS into the machine. Should be really interesting and I've been meaning to check it out ever since I learn about it on the Dragon.

[@wcarlin]

If it doesn't have one already, the FM-7 needs a native DriveWire client so it can boot OS-9 via serial. One of the tiny clients, like you load from cassette on the CoCo.

[@mikafoxx2717]

OS-9 is honestly sad, such a good operating system on such a dead-end system. At least they had much better basic than Commodore ever did, and they made it a lot more capable in the coco3. The uncommon processor made it hard to ever port to anything else so it kind of died with the Coco's, which never got the popularity of the original z80 trs-80's, nor the IBM clone Tandy 1000.

[@CurtisBoyle]

@@mikafoxx2717 - OS-9 was later available for the Motorola 680x0 series (nicknamed OSK), and then OS-9000 was multi-platform (Intel x86, Motorola 680x0, DEC Alpha, PowerPC, etc.). But it was never marketed as a home platform (with the except of Tandy and the Coco line), which is too bad. It was meant more for the industrial market. Amiga almost used it as their OS (it had most of the things that they wanted - pre-emptive multi-tasking, small memory footprint, etc. but they had no GUI ready, so Microware lost out). While I would agree with you that the Tandy 1000 line eventually outsold the Coco, I would disagree that it outsold the Z80 based machines - even Tandy themselves for years in the 1980's stated that the Coco was it's best seller. They had sold over 500,000 Coco 1's alone by the summer of 1983, and between all 3 models it sold at least 4 million units.

[@RobSchofield]

I've got three of these machines and they are superb!

[@deang5622]

3 handed mutant?

[@null7936]

WoW a 6 cores 8 bit machine ... Maybe Fermilab, Darpa, NASA who knows ? ( ;

[@horusfalcon]

Wow! That's quite a beast of a machine. It's always a pleasure to watch your videos, and I hope you have lots of fun doing the next one. The FM-7 seems like such an interesting architecture. Small surprise that this beastie didn't fit in your suitcase. (You always have such cool toys!)

[@sandmanxo]

It's rare to see an 80s computer i haven't heard of at this point. Nice video and looking forward to more next year when you can get your hands on it again.

[@alisontelford9339]

Hi Noel! Very cool video -- what an advanced device and a neat history! If the government had supported it, sure seems like it could have worked out, a la UK microcomputing industry (Tho of course we have the example of the NABU here in Canada which failed even with support). Anyway, I don't know if it's been mentioned -- the youtuber Usagi Electric said in a recent video that his day job is technical translation of japanese to english, and he's a vintage computer guy, so perhaps he might be of help if there are remaining translation challenges? No idea, just thought I'd mention it! Cheers and thank you for the wonderful content!

[@thebyteattic]

This is why this is the best retro-computing channel on KZbin, and why the Japanese made many of the best microcomputers of the 80s.

[@NoelsRetroLab]

Wow, thank you so much! That means tons coming from you! 😃

[@WacKEDmaN]

"all the best stuff is made in Japan!"

[@another3997]

That's a bold claim, considering the relative lack of impact Japanese micros had on the home markets outside of Japan. Especially this particular micro, which seems to have died a quick death just after birth. And judging by the lack of information out there, apparently nobody even turned up for it's funeral. Interesting machine? Yes, definitely. A great machine? With only 8 colours and a generic AY sound chip? Not so sure.

[@manicdataminer]

Well, there is one... no, two modern 8-bit computers with two... no, three CPUs... I keep forgetting their names... At the tip of my tongue... :-)

[@thebyteattic]

@@manicdataminer 🤣

[@AshtonCoolman]

The games for the FM-7 series were out of this world compared to the stuff the United States and Europe had. The Japanese really were far ahead in gaming back then.

[@NoelsRetroLab]

Any particular games you would recommend I check out?

[@TheMalMeninga]

@@NoelsRetroLab Not OP but I played Galaga and Silpheed on an FM-7 many years ago, and they were quite good for the time.

[@kitsuneneko2567]

They still are in some ways. Especially if you like things that jiggle.

[@another3997]

Care to share any info, sources, reviews so we can look for ourselves? There's not a lot of info out there. But if the computer was as good a games machine as you say, why did it fail in Spain, and why is it seemingly never mentioned alongside other popular Japanese micros? In 1979 Atari launched their 8 bit computers, far ahead of anything the Japanese (or anyone else) had at the time, at least in the home computer market. 3 years is a long time in which to catch up.

[@belstar1128]

@@kitsuneneko2567 I think American and west european games where on par with japanese games from 1995 to 2015 .but now western games are going down hill again

[@julianbrown1331]

Love this computer - its what inspired me to try and create my own design with a dedicated graphics CPU

[@julianbrown1331]

Just a small detail to point out - the BBC micro could do the same trick using the TUBE interface. If you want to see a side-by-side comparison, try running Elite in single processor mode and the same game in TUBE mode with a second 6502. The performance differences are amazing

[@MrDave6309]

The 6809 is a awesome cpu.. and has proved itself awesome for graphics in this system and some arcade boards ( splatterhouse ) you have a great looking machine there !

[@douro20]

The Namco System 1 has THREE 6809s. One of the sub CPUs ran the tile engine and another ran the sound. One of the DMAs on the sound 6809 is actually tied in to the FM DAC to allow for digitized sound playback; it's the first time I've seen anyone use the FM synth DAC in this manner.

[@akakakakakak3084]

6809 is the most powerful 8 bits microprocessor.

[@M0UAW_IO83]

Damn, ancient history but a schoolmate and I designed a dual Z80 system back when the Spectrum was still being sold, our idea was one CPU with ROM and 64K RAM for code, the other CPU for graphics and both sharing an expandable block of RAM that paged in and out on demand so the main CPU could upload new graphics data or code for the second CPU. Lost touch with him after I left school but, if you're out there Ben M, say hi...

[@guaposneeze]

At one point, Apple had a weird microkernel OS that was meant to be licensed to NUBUS card vendors making things like video cards out of CPU's that could run a semi-normal OS under the hood. Basically this exact idea, just scaled up to the late 80's.

[@Cybergorf]

@@guaposneeze en.wikipedia.org/wiki/A/ROSE Apple put a 68000 on these NUBUS dev boards running A/ROSE - a full preemptive multitasking realtime kernel, that was way more advanced than the OS of the Macintosh II these cards plugged into.

[@remaincalm2]

But would it have solved the colour clash issue? ;-)

[@tonyfishlock7240]

Nice one Noel it's always interesting to see new and different retro equipment and you do it so well thanks

[@zxprojects]

mcleod_ideafix here! Me alegro de volver a verte. Precisamente, el Secoinsa FM-7 lo conocemos un poco en mi departamento. Debo tener 2 o 3 de ellos en mi despacho. Cuando vuelva a la universidad en septiembre, miraré de los que tengo, uno que viene con su caja y todo, a ver si incluye alguna documentación. Por otra parte, preguntaré a mis jefes, los más antiguos del lugar, por si ellos supieran de documentación técnica que pudiera estar guardada desde hace 30 años o más en alguno de los armarios del seminario del departamento. Saludos!

[@TanjoGalbi]

You said in the Commodore 64 DIN socket pin 3 carries sound so you could use that cable for the video out, but all 3 of the DIN connectors you showed also have sound on pin 3, you could do the same for any of the cables for those machines.

[@309electronics5]

I have seen some consoles and computers use 2 cpus, one as the main and the second one for sound or something else but never seen a computer with a cpu that acts as a gpu. Pretty nice!

[@deang5622]

Atari did it with their range of computers: 400, 800, 600XL, 800XL.

[@charlesd4572]

This is awesome - brilliant stuff. The technical detail, editing and production is quality.

[@NoelsRetroLab]

Much appreciated!

[@WacKEDmaN]

what an absolute monster 8bit!.. the architecture is so unique and seems quite powerful..especially for a machine that was going into schools...im interested to see more about this machine... it would be interesting to see if you could add some new routines to the 'gpu' rom, or even optimise the routines ...i guess with all the extra gpu ram, and SRAM blocks.. this had no issue with the 64K tax in the spanish market!

[@DouglasFish]

I'm blown away you don't have more subscribers. I hope you're doing well with this endeavor? Your production quality is pretty darn solid

[@AppliedCryogenics]

I hope you get to tear down an FM-77 AV! (Same dual 6809's, but 4096 colors! Also, a genlock, I think.)

[@superandroidtron]

You actually _can_ write custom programs for the SUB-CPU, there’s a hidden way to do it: the name of the programmer who wrote the SUB ROM “YAMAUCHI” must appear at a certain place in memory to make it work.

[@NoelsRetroLab]

That's great to hear. I'm going to have to track that down for the next video on the FM-7!

[@douro20]

There's also the larger FM-77 which has similar specs but is a full-fledged desktop system with two floppy drives and and the FM-77AV which has a larger pixel buffer allowing it to display up to 4096 colours at 320x200, and FM sound (although a bit limited in that regard).

[@robbyxp1]

The architecture reminds me of the BBC Micro with its IO processor (the BBC micro 6502 itself) handling the video/IO, and a second processor (via the BBC Tube interface) running the application. This was the original idea by Acorn for the Proton (BBC prototype). You can get now lots of co-processors for the BBC with the Raspberry PI.

[@bobjones-ey5gl]

Acorn BBC - Econet GOOD TIMES - although ELITE (Braben & Bell) was almost unplayable (way to fast) if your (primary school) Beeb had a second 6502 processor connected via the Tube interface.

[@Phenomz75]

seriously, they should have done this with the Commodore C128 which also had two CPUs (a 6510 and a Z80). We all know well what the CPU is capable from scene demos (it just doesn't leave any computing power left for anything else).

[@dans.8198]

Unfortunately, on the C128 the 6502 and Z80 shared the same bus and RAM. Therefore one of the CPU must be halted for the other to work without conflicts. The Z80 on the C128 was only meant to run CP/M. However, the C128 also had an extra video controller chip (VDC) with its own separate RAM and interesting functions including blitter, but not powerful enough to be used for games and demos.

[@Cybergorf]

@@dans.8198 You are wrong about the VDC`s demo capabilities: kzbin.info/www/bejne/iGnFdpqOa9Ngh68 For the "parallel" use of both CPUs: The Version of CP/M for the C128 moved many things to the 6502 to free up memory for the z80 side of things. So while really just one professor could operate at a time this gave this special CP/M version about 50KB for programs.

[@MrCenturion13]

3 CPUs, not 2. 64 chip, 128 chip and a Z80 chip.

[@dans.8198]

@@MrCenturion13 The “64” chip (MOS 8502) was the same as the “128” chip.

[@mldalex]

the z80 was exclusively there just for the cp/m mode

[@foogod4237]

6:20 I don't understand why you would jump to that weird conclusion about the layout of the video RAM. (I don't have access to this system, so I can't check, but given the design of the earlier version you have the schematic for, that is certainly not the way I would have done it, or assumed anyone would do it) The original version has three banks, each of which are 8 bits wide (1 bit x 8 chips). So why would you not assume that the new version also has three banks, each 8 bits wide (4 bits x 2 chips)? Then the only change in the architecture is that each bank consists of two chips instead of 8 (which means the rest of the system design can be identical to the original one (same shift registers, etc)).

[@websterleone]

That's neat! I had a similar idea for a hobbyist 8-bit computer I was floating around of using a second CPU as a GPU. I'd really love to see people write software that takes that kind of architecture to its limits.

[@ahmad-murery]

Very interesting computer indeed, I'd love to see more about this machine, MSX too has a separate VDP so that the Z80 can delegate graphics operations to it, from the other-hand this made the communication between the cpu and the VDP is slower (as I read) Thanks Noel!

[@BarnokRetro]

What an awesome machine! I look forward to seeing it going through it's paces.

[@ChrisDreher]

11:12 The 2KB ROM definitely is code (good call). From inspection and educated guesses, the ROM is divided into 4 sections that are bank-switched into address 0xFE00 - 0xFFFF. On the 6809, address 0xFFFE is the reset vector and all 4 banks have the reset vector send the CPU to address 0xFE00. In the 4th bank, there is just 1 instruction: BRAnch instruction that loops forever on itself. The other 3 banks start with a BRAnch instruction that skips over the next 9 bytes which appears to be 3 JMP instructions (likely a jump table). I'm not a 6809 expert but still interesting early results.

[@ChrisDreher]

Just took a peek at the 32KB ROM 8449 and its last 512 bytes has the same structure as the 4 banks mentioned above. All 5 banks are different. Interesting...

[@ChrisDreher]

After using a diff tool on the 4 banks of data (ignoring the one empty-loop-forever bank), they have a common structure. They mainly differ in the first ~64 bytes and the last ~64 bytes. The middle portion of each are identical other than a few bytes due to some jump offsets being slight different (this middle portion is what the 1st of the 3 JMP instructions mentioned above jumps to). The 1st 64 bytes is where the main differences are. The last 64 bytes are typically empty except for one bank where the 1st 64 bytes JMPs over the middle section to continue executing in the last 64 bytes.

[@johnczaia9124]

Thank you, Noel! Never heard about this interesting box. Always looking forward to your excellent videos :)

[@kobe3576]

That computer is a rebranded version of the Fujitsu FM-7. There is also an FM-8 computer that, despite its name, it was released *before* the FM-7. The FM-8 was similar, but it also supports Bubble Memory. I would recommend you to get one of these as it is very interesting.

[@frioglobal]

The FM-8 was really outstanding. Two CPUs, built like a tank and supporting solid-state non-volatile memory cartridges, really something special!

[@adilsongoliveira]

This machine is a beauty! Loved the architecture. Thanks for bringing this up to us!

[@raven4k998]

this machine is like having a dual cpu amd 5950x system but one of those 5950x cpus does graphics only it's crazy

[@Psy500]

There is emulator for the FM-7 but the ones I know of are in Japanese so take a bit of effort to get setup yet that does mean disk images does exist like on the Internet Archive. As for games well you showed a bit Ys, the FM-7 version of Ys II does have a nice intro but it was meant for the later FM-77AV

[@andrewdunbar828]

Are any of them open source? We can help translate.

[@Psy500]

​@@andrewdunbar828 Mame can kinda do FM-7, XM7 is better yet Japanese and was discontinued.

[@andrewdunbar828]

@@Psy500 I was gonna check if MAME could do it. Unfortunately MAME is so confusing to use for computer emulations I have to relearn all the counterintuitive steps every time I haven't used it for a while.

[@jonathan_herr]

@6:38 isn't there a couple other DRAM chips just below the 6x64 DRAM area?

[@espfusion]

You actually can upload custom code to the sub CPU. I'm not sure if the commands fir it were ever officially documented but they're there. The two CPU setup is interesting because it means normal code doesn't have to worry about managing the sizeable 48KB of VRAM or having to share a bus with it while the sub CPU can get full direct access to it. On competing machines like NEC PC-8801 or Sharp X1 you had to bank switch the VRAM into the main CPU address space one 16KB bitplane at a time. I think this setup was mostly there to provide high resolution graphics while letingt programmers focus all the CPU time on processing without worrying about the graphics subsystem getting in the way. I don't think graphics draw performance was that much of a consideration.

[@Curt_Sampson]

The subsystem command to execute arbitrary code was informally known as the "Yamauchi command"; if you do a search for "FM-7 YAMAUCHI" you'll come across some good information (in Japanese) on using it. (I'd give you a link, but KZbin tends to suppress comments with links, I find.)

[@frioglobal]

I was going to write the same comment. It is indeed possible to upload custom code to the sub CPU. Each of the main 8-bit Japanese systems had a different approach to obtaining high resolution display, it is very interesting to understand and appreciate them, and their evolution.

[@elijahvincent985]

I managed to upload the full tape for Mappy for the FM-7 a few years ago from a recording of a bad-conditioned tape. I cleaned it up and removed any excess noise that could affect the loading process. Enjoy! kzbin.info/www/bejne/eHfcpWyKg7JsmdU

[@minombredepila1580]

Secoinsa !!!. More than 30 years without hearing the name... Wowwww

[@electricadventures]

Quite an interesting machine, that I hadn't seen much on, thank you.

[@Zeal8bit]

Another great video and what a beautiful motherboard this is! Routing it must have been an interesting puzzle to solve, even more if that was done by hand. Do you happen to know how many layers that PCB has?

[@NoelsRetroLab]

Thanks! I don't know how many layers it has, but it has to be more than 2 for sure.

[@ChadDoebelin]

Thanks for sharing this! I've never heard of the Secoinsa FM-7.

[@charlesjmouse]

What an interesting architecture. I have to admit while I was aware of the FM-7I payed it no attention because I assumed it was just another Japanese 8-bit and as I don't know the language why bother? Thanks for educating me.

[@juanantonio4955]

Great job, Noel. Thanks fron Spain!!

[@jimadams7765]

You've hit retro gold. Very interesting device.

[@loxai0]

Nice vid! I remember another spanish market adaptation, done by Investronica, resulting on the Spectrum + (later also sold outside of Spain). The few advantages on the classic speccy were overshadowed by some incompatibility issues.

[@tony359]

need a larger suitcase! Looking forward to more about this, pretty interesting indeed!

[@misterkite]

@12:12 looks like it needs to output +12V, -12V, and +5V, just like most MSX machines. I successfully swapped out the busted power supply of a National MSX computer with a Mean Well RT-50B which provides all 3 voltages for just $23 on amazon. Plus it was small enough (minus the metal case) to easily fit inside the computer.

[@fcastellanos]

Hi Noel. I guess you can use an automatic japanese to english translator. there a lot that just can be used with an scanner or the mobile phone's cam.

[@MostlyPennyCat]

OK. Now I have an idea. BBC Micro/Master. An expansion cart containing a second 6502-like processor with the command buffer, a program buffer also an output command buffer and matching video RAM. Sadly only the CPU has access to the _real_ video RAM so it won't be quite as fast need you still have to halt the main CPU and copy completed screen data across to main RAM, but augmented with programmable "shaders" you could do some really advanced stuff. Or maybe even just add your own video out port from the cartridge. Or, just like the old 3dfx accelerators, you combine the analogue output of the 2D and 3D cards with a pass through cable.

[@yosi1989]

NEC PC-8801(A/B/mk2AR) is similar, but, 2nd CPU used for Disk control.

[@AS-ly3jp]

This was a great episode!!! (as usual) Thank you!

[@ropersonline]

13:56: The 60Hz-ness and the 12V pin on the DIN connector (13:06) lead me to suspect this probably originally shipped with a custom CRT that was powered from the computer. Kind of like Amstrad used to do it.

[@retroatx]

When I saw the title I assumed it was the Fujitsu FM-7. I was surprised as I had never heard of the Secoinsa. It was actually a group of FM-7 enthusiasts that had replaced the main 6809 with a 6309 that discovered the hidden mode of the 6309 CPU IIRC - I may be remembering wrong. It has been decades since I used one of those but I seem to recall the second CPU was clocked by inverting the clock signals that went to the main CPU so there could never be any contention and neither had to wait to access the bus. Running OS-9 Level 2 on that thing was a dream - to me it was a better Unix than Unix.

[@wirebrushofenlightenment1545]

Hola! - A new channel to me, thanks to the algorithm. I will be watching more, definitely. I used to work for Fujitsu in the UK. Avionics though, rather than computers.

[@NoelsRetroLab]

Awesome. Welcome aboard!

[@wirebrushofenlightenment1545]

@@NoelsRetroLab Aboard for the voyage with pleasure, Captain! I can't wait to see where we will be going!

[@JoelReesonmars]

I was going to offer to translate a few lines at a time, but it looks like you've got that covered.

[@BilisNegra]

I see you are here in Spain for the summer and your lab is still all set up and complete, instead of having had some of the equipment shipped to the US, That's really quite surprising.

[@NoelsRetroLab]

A lot of the equipment needs 220V to function. And besides, I wanted to have the basics to be able to work there a couple months out of the year.

[@redleader7988]

Very cool. The "gpu" seems to follow the 6502 and ANTIC scheme on the Atari 8-bit line somewhat, where you could upload independent programs to the ANTIC (limited to a small set of graphics related functions) called "display lists" stored to on-chip S-RAM to perform some graphics functions independently of the 6502.

[@user-qf6yt3id3w]

Display lists weren't in on chip SRAM. Antic fetched them from regular memory using DMA. In fact Antic does all the DMA in the system.

[@redleader7988]

@@user-qf6yt3id3w Thanks for the correction. The onboard SRAM is used for other things: "A buffer within ANTIC is used to store graphic data for a single scan line. Its purpose is to buffer data for use on repeated scan lines, reducing DMA overhead. For bitmap modes, it allows ANTIC to only read graphics data for a mode line once, during the first scan line. For character modes, it holds the character name data which is then repeatedly used to fetch each scan line of character data from the character set."

[@user-qf6yt3id3w]

@@redleader7988 Well I never. You're right. And if you look at the die shot the 48 bytes of SRAM take up a decent percentage of the die. So Jay Miner and co must have thought it was really necessary.

[@andrewdunbar828]

Yep that was a classic Jay Miner design. He also designed the Copper graphics coprocessor for the Amiga which had similar capabilities.

[@Antonio_Gallego]

What an amazing find! Thanks Noel!

[@Artoooooor]

Wowowowowow! I just love this architecture :o

[@hernancoronel]

As a 90s collector of the spanish computer encyclopedia “Mi Computer” (yes it was spelled like that), I knew about the Secoinsa Computers even though I live in Buenos Aires, Argentina and still do. Thanks for the video Noel!

[@filker0]

Interesting. This reminds me of how the Digital Equipment Corporation VT240 graphics terminal works. That had a T-11 (PDP-11 chip with an 8-bit bus), an 8085 (I think) and an NEC graphics chip.

[@andrewdunbar828]

How much did they sell for back in the day? I bet they were pricey. I've heard the 6809 was a great CPU but expensive compared to its more popular contemporaries, then there's all that TTL and it seems to be very well engineered. I would bet in the price range of the Apple II and BBC, which of course makes perfect sense.

[@edgeeffect]

2 6809s!!! That gives it the power of the original Fairlight CMI.

[@HelloKittyFanMan]

Nice video signal conversion setup!

[@edoaut]

Thanks for your mention about my projects

[@anatolykosychenko8038]

Hi Ya & best wishes. Thanks for work. Be Happy. Sevastopol/Crimea.

[@misterkite]

The FM-7 is awesome, but weirdly named. It's a stripped down version of the FM-8. You might be more familiar with the 90s sequel: FM Towns, which even had a console version: FM Towns Marty.

[@NoelsRetroLab]

Yeah, definitely weirdly named. I did look up the whole FM family tree on Wikipedia and I had heard of the names very vaguely, but I never got to experience the FM Towns or the Marty (other than it was a music setting in Monkey Island).

[@NathanChisholm041]

@@NoelsRetroLab Yamaha made a FM-7 synth made famous for 80s music.

[@uneune6017]

FM-7's main system can be switched to various CPUs, The expansion slot on the right is for that. MAIN(6809.,6309, Z80, 8088, etc…) + SUB(6809) This slot is also used to connect FM-X (MSX), Control MSX display and sound from FM-7.

[@spacedock873]

Seen a couple of FM-7's at Dragon Meets as the CCH in Cambridge (they were allowed as they are 6809 based). As such it made me smile to see an advert for the Dragon 200 in the magazine you were leafing through! 😊

[@photovincent]

Having been a kid in the ‘80s I still love computers from that time, fascinating video! Interesting architecture as well, having a generic programmable GPU, but like you mentioned briefly towards the end: timing events to graphics may be a pain. I grew up with the C64 and its custom graphics chip, and one of its most interesting features I found to be the hardware interrupts generated on sprite collisions! Any idea this setup had a way to communicate from the GPU back to the CPU 6809?

[@boristheengineer5160]

I've read that Acorn originally planned the BBC micro to have two 6502 CPUs, before opting to split out the second CPU and make it an expansion unit. As I understand it the 6502 "second processor" had very little ROM, and what it has is copied into RAM immediately after reset then the ROM is switched out. It acts as a proxy for the OS on the main board so system calls on the second processor trigger actions on the main board. Without studying the FM-7 block diagram in detail I can't tell but I wonder if it actually works to a similar plan with the CPU having a flat 64k RAM map and just enough ROM to "boot" it, and the other "GPU" does EVERYTHING, graphics, sound, I/O, firmware etc. Also this looks like it could run OS9 (Level 1 anyway) so I wonder if it ever did? Come to think of it what storage does it have?

[@BuddhaPhi]

Awesome video! It's such an unusual system. You should have no issues taking a computer like that through airport security - especially since it doesn't have batteries. I've packed many slightly weird things like electronic drum machines in carry-on bags and have never had any trouble. As long as an item isn't explicitly on a prohibited list you should be good.

[@tigheklory]

I would love to see you get a Coleco Adam computer, there's some really cool things about it like the FujiNet and the MIDI adapter.

[@station240]

Actually the system diagram does explain what that extra 2k rom is. CRT 10k ROM MB8364 MB8516(CG) So that block is actually two ROMs, 64kbit/8kbyte and 16kbit/2kbyte. The second ROM being labeled as CG confirms what I suspected, it's a Character Generator ROM, or set of bitmaps for letters and numbers to be displayed on screen.

[@NoelsRetroLab]

Oh that makes a lot of sense! Thanks.

[@GrugBug-f7j]

Solid! Top KEK! Peace be with you.

[@Curt_Sampson]

It's worth keeping in mind that one of the reasons for the sub-CPU architecture was that the video display system was done much more simply than on many other computers: it's a straight 3-plane 640×200 graphics display, with no hardware support for text, scrolling, or anything like that. (I'm not aware of any other 8-bit computer system that did this as a standard display system, though there were probably S-100 or similar video expansion cards that worked the same way.) This creates two issues. The first is that your frame buffers take up 48K of address space, and the second is that you need to display letters dot-by-dot, as e.g. HRCG did on the Apple II. Given the address space and CPU requirements to do this, dedicating a separate CPU to the task makes a lot of sense. And doing things this way makes even more sense when you remember that the computer was designed to display Japanese kanji. The kanji card for the system had 128 KB of ROM and held several thousand 16×16 kanji glyphs that the sub-CPU could read and render in the same way it rendered normal text. This also offered several fun features. Since a program was rendering text, you could render it at arbitrary sizes and orientations; FM-BASIC supported this. If you're interested in the details of any part of the system, I have made the schematics and a few notes available on GitLab in the `retroabandon/fm7re` repository.

[@benbreeck3363]

The Fujitsu FM-8/7/77 series were the Radio Shack/Tandy CoCo machines they should have built if they had been genuinely serious about taking on Atari, Commodore, Osborne, Timex-Sinclair, the Coleco Adam, and Apple in the home computer space, well, minus the kanji ROM. In the same vein, the FM TOWNS was the high-feature big btrother the Tandy 1000 and 2000 desperately needed to stay competitive in the PC Clone market, especially if they could have commisioned "Second Party" developers to port games, graphics productivity, and language compilers to support the included graphics chipset.

[@bangerbangerbro]

The first thing that came to mind when I saw the title was the Soviet Elektronika UKNC, which has two of those Soviet NC-architecture CPUs that are not PDP-11 clones on chips but are PDP-11 compatible (binary I think but might be only assembly compatible for all I know). That also has 3 bit planar graphics, but is paletted out of a 6 bit palette. It is also much later, 1987. I have no idea if it has any ROM for the peripheral processing unit as it is called, but I would like to think so given its allocated role of graphics processor despite being just a normal CPU. However it probably doesn't. It was very interesting to learn about this machine anyway, thanks!

[@aqualung2000]

Damn. If it weren't for the language barrier on the technical documentation I think I would *love* to get my hands on one of those and start programming for it. What a fascinating computer! Tell me, do you know what clock speed the 6809s run at?

[@NoelsRetroLab]

I knew I left something out! I meant to talk about the DIP switches in the back. Apart from deciding how the computer boots, you can select the CPU speed between 2.0 Mhz and (I think) 1.6 MHz. I'm not exactly sure why that was needed. Maybe FM-8 compatibility?

[@lis6502]

@@NoelsRetroLab rather region selection. I am pretty certain that messing with dip switches would affect pixel clock and refresh rate.

[@RayBellis]

If you added a second processor to a BBC Micro via the Tube interface then the Beeb itself was effectively relegated to an IO processor.

[@zh84]

We had a BBC in our school with a "Z80 second processor" so that it could run CP/M.

[@richardgodivala4680]

Back in the day, I had a 65C102 co pro (similar instruction set, but higher clock speed), and a 80186, which was an Intel chip, and could run DOS and GEM. I never had the Z80, or 32016 (expensive research kit) - and there were others. Now, you can get a Raspberry Pi Co Processor, which can emulate any of the original ones, and can run in native mode. (You can also use the HDMI output on the Pi, but that kind of misses the point...) I have a BBC Master with a Pi CoPro, a Pi 1Mhz Bus Disk emulator, and a Pi Econet Hub/Bridge ... kind of like in Star Trek, where the ancient Voyager craft has all the new kit added to it to make it keep running, but in the core, it's still the original hardware and programming....

[@AudieS71]

Have you checked out Amiga's with processor accelerators?

[@norwegiangadgetman]

Have you seen the Epson HX-20? It also have 2 CPUs, but running at 0.6MHz. But because one is handling the IO, it feels a lot faster.

[@RixtronixLAB]

Nice video, thanks for sharing, well done :)

[@retroand]

Actually arcade boards use this technique extensively. I have many with a 68k+Z80 tandem, others with multiple Z80 and even one with three 6502... It would be quite interesting to find the schematics of those boards in order to learn how they do coordinate.

[@marksmith9566]

Please bring it to one of the VCF shows!

[@NoelsRetroLab]

That's a great idea! I couldn't fit it in my suitcase, but next summer I'll bring it back with me and try to take it to some of the VCFs.

[@marksmith9566]

@@NoelsRetroLab Sounds like time to look for a steamer trunk ;-)

[@der.Schtefan]

Bit blitting is the Single most powerful feature on 80s home computers. The REU on the C64 was often just used to move data around many times faster than the cpu could.

[@Ty-Jack]

This video seems cool 😎

[@Ty-Jack]

This video seems cool Iike you (it a complement If you thought otherwise)

[@Brettski777]

Hi Noel, would the 'chain'command be tape loader related ? All the best . Brettski

[@luisrodrigues7162]

In Portugal whe had the project minerva on public schools in 90's to introduce computers on hi schools. It was a mainly 8088 and 286 processors based computers like PHILIPS and UNiSYS respetible.

[@mytube9182]

Fujitsu Micro 7 was my first computer science teacher. At age 14 I've learned BASIC programming with this 8-bit PC at it's what've changed my whole life for good. I never get any interest in a thing of electronics before that. At my school the math division had bought a big set of Fujitsu Micro 7/8 and initiated application for computer science and BASIC programming course. I thought it sounds cool and just wanted to impress friends so I applied. In the class I just followed the tutor blindly without any exact idea what were these for. One day I picked the BASIC tutorial sheets for killing time reading and WHAM I don't know how but I got the damn whole idea how the programming of that time worked. Then my life from that moment on is programming.

[@greenaum]

You got the circuit diagram wrong at 16:34. If you want to lower the voltage you need a voltage divider, that's 2 resistors per channel, one of them going to ground. Resistors work on current, not voltage. If you look up "voltage divider" you should get it, it's not really complicated.

[@Clavichordist]

Visual Technology did that with their V-1050 CP/M computer in the early 1980s. The Z-80 was used for the regular processing along with its own RAM and a 6502 was used along with its own 32K or RAM was used for graphics.

[@mikafoxx2717]

The Nintendo 64 (and most SGI's that it was based on) were similar in this aspect, sort of. The graphics was another MIPS cpu, but basically run more like a microcontroller with microcode that controlled the (mostly) fixed graphics pipeline. The main CPU did the geometry, and the second did all the rendering with passed parameters. The second did have things like 4k of graphics cache on die, but it was indeed a second processor with code and cache built in. Remember that SGI created OpenGL.. basically a way of abstracting draw instructions that the microcode in the graphics CPU would interpret into it's own instructions. Very few games actually licenced the ability to change the microcode so it was usually a fixed pipeline in essence, but modifying it would be like adding features to OpenGL, like adding a posterization layer or such.

[@recordtronic]

I wonder if that ROM on the lower left has character glyph bitmaps?

[@jaapverhoeven422]

Interesting: this architecture with two 6809's for cpu and video is very similar to the Qix Arcade board design! Communication between the two happen there via 2kB (or 4k?) dual-port ram, and signaling via the 6809 fast IRQ lines. The Qix video subsystem has a lot of dedicated video rom which (logically I assume) contain next tot the display routines a lot of graphics and sprite data of the game itself. There are differences, but it is also very similar.

[@andrewdunbar828]

Well if there's no Qix port to the machine we'll have to encourage Noel to write one! (-:

[@jaapverhoeven422]

@@andrewdunbar828 Haha! Interesting, but this computer has no 6845 crt controller AFAICS. And the qix arcade uses that 6845 to put out a fixed 256x256x8bit resolution. Not to mention the 30-40kb of game roms on board sprinkled in the CPU/data, video and sound sections....

[@andrewdunbar828]

@@jaapverhoeven422 Yep they're far from identical but they have at least one interesting thing in common.

[@rashidisw]

I do curious if the IBM PGC where somehow inspired by this architecture style . The thing has its own 8088 which function as GPU afterall.

[@speedrider3145]

I do wish modern computers are like this.

[@TheDiveO]

maybe just a 2k boot ROM for the GPU, and it gets booted from the first CPU and fed code from the "main" ROMs?

[@colinkraus7139]

Cool videos. Thank you.

[@DAVIDGREGORYKERR]

The 6809 was also used in CMI's Fairlight V.

[@danielmasek8171]

Looking at the 2 kB ROM you called "GPU rom", it looks much like a boot rom, with four 512 Byte pages, mapping to 0xFE00 -- you can see the reset vector at the end of each page pointing to that address. The last page just does in infinite loop, but the other 3 are each different.