EEVblog #746 - Sharp X68000 Retro Computer Teardown

[Fungus]

I'm surprised how well the style held up. It doesn't look like it's 20+ years old. I was assuming early 2000's.

Yes, nice design. I'd totally have something like that on my desktop today if they made PCs like that.

Carrying handle is perfect for LAN parties (or whatever they call them these days)

[pickle9000]

I've been looking at some of the games for the Sharp, you can certainly see why it was sought after. The games are really arcade quality, both the sounds and graphics are bang on.

It's so surprising to such an amazing product, I had no clue it even existed until today. Many thanks to the person that sent it in.

[hyperneogeo]

So, can it run MetalSlug?

Metal Slug was made wayyyy after the x68000. (Think PSX era).

[Fungus]

I've been looking at some of the games for the Sharp, you can certainly see why it was sought after. The games are really arcade quality, both the sounds and graphics are bang on.

Yep.

I had no clue it even existed until today.

Nor me, although it's not really surprising; Japan had lots of weird home computers in the 80s that nobody in the west ever heard of.

Many thanks to the person that sent it in.

It's been a trip down memory lane.

A bit of a cop-out for the power supply though, what's with the "out of time" business? There was still half an hour of video left...

[Stonent]

True arcade perfect requires a 16 to 20 MHz 68K processor, Zilog Z80 (as a sound controller) and the Yamaha FM Synth + DAC.
That will cover most capcom and Sega up to System 16, possibly System 18.  I think System 24 may have used dual 68K processors.

Though not a computer the Neo Geo was probably was the best 68K hardware system for doing this.

The Sega Genesis (Megadrive outside of the US) had a slower 68K processor but had the Z80 and yamaha chipset.  It suffered by having a lower color bit depth video and did not have a dedicated DAC for non-FM synth audio.

The Super Nintendo's CPU was slower still and had a lower screen resolution than the Genesis, but had more RAM and better audio DAC so it could use wave table synthesis.  It also had a palette of 256 colors out of 32k vs 64 colors out of 512.

The Amiga had a faster 68k processor but no yamaha synth. The dedicated graphics hardware made it easy to offer the video experience needed for arcade graphics.  The lack of the yamaha synth meant that it would have to use sampled audio.

[German_EE]

Am I the only one who thinks that IBM were INSANE when they chose the 8086 for their PC rather than the 68000? I remember seeing single board 68000 systems in the workshop when I was repairing PCs and the Motorola based systems were an order of magnitude faster than the PC-XT. Also, the memory map of the 8088/8086/80286 is a mess 

[Fungus]

The Amiga had a faster 68k processor but no yamaha synth. The dedicated graphics hardware made it easy to offer the video experience needed for arcade graphics.

The Amiga could scroll the screen all over the place using hardware/'copper' but it had no sprites worth speaking of. Sprites is where the X68000 would totally destroy an Amiga.

Yes, you could turn on more bitplanes as an overlay and use the blitter instead but that reduced you to 8 colors on screen and totally blocked the CPU in the visible screen area due to lack of RAM bandwidth (the CPU only ran in the border areas and vertical flyback, just like a ZX81!)

The X68000 has those high-speed RAM chips for a reason.

The lack of the yamaha synth meant that it would have to use sampled audio.

...and those same sound samples would eat up all the RAM in no time, leaving nothing for graphics.

Stuff like this is what made the Amiga really terrible for games (but excellent for demos!).

[miguelvp]

They were competing with CP/M 80 (Z80/8080) based CPUs, so it made a lot of sense to base it on the 8088 (not 8086).

The OS was going to be CP/M 86 but they struck a "deal" with Microsoft, and used off the shelve components, and the rest is history.

[Fungus]

Am I the only one who thinks that IBM were INSANE when they chose the 8086 for their PC rather than the 68000? I remember seeing single board 68000 systems in the workshop when I was repairing PCs and the Motorola based systems were an order of magnitude faster than the PC-XT. Also, the memory map of the 8088/8086/80286 is a mess 

IBM deliberately chose something a bit crap because they didn't want PCs to eat into their 'real' computer business. It was also compatible with CP/M, a big plus in their target market.

In the long term Intel turned out to be a far better choice. By the time of the 386 it was leaving Motorola behind, the 486 left it for dead.

[Paul Moir]

It's lore and I'll try to find some proper evidence but my recollection is that IBM never really expected their PC to take off.  That's part of the reason they didn't go to the expense of developing their own OS or purchasing the exclusive rights to DOS from Microsoft.  Furthermore, the original PC was based off some 8085 running word processor IBM already had, and dropping in an 8088 made for a simple "16 bit" modern PC on the cheap.

The 68000 needed a real 16 bit external bus, which added a lot of expense.  For example it's hard to route 16 bits on a two layer board, and DRAM was typically by 1 bit so you needed 16 RAM chips minimum.  The 68008 didn't have a BUI like the 8088 so they suffered a terrible performance penalty.  Also the 8086 instruction set was translatable from 8080 code so theoretically existing CP/M software was easily run, or at the very least easily converted.  This partly explains the messed up memory system too, and I'll point out that this has been the historical success of the 8086 platform:  that compatibility is more valuable than orthogonality.

[dexters_lab]

i seem to recall some legacy reasoning with the original XT and the 8088, the 68k was a much newer processor design... the Atari ST and Commodore Amiga were both fresh new designs at the time and were made for a very different market space

nice to see the insides of this X68000 machine, some very serious design there, some big dedicated hardware that would make the ST and Amigas custom silicon go week at the knees!

But then $1500 price tag, in Japan... wow, if that was sold to the US or EU markets it would have been $2000+ by the time i was exported and on the shelves. Nobody would have bought it!

it's physical monolithic black box design reminds me a bit of the NeXT, which interestingly was also 68k

[bktemp]

As a comment for mentioning the amount of 74 series logic used in that computer, have a look at this 386 mainboard from IBM:
It is a mix of ASICs (unter the metal cans), PLDs (PALs, MMI 72x7486) and a lot of 74 series logic. They did even make their own memory modules and ICs instead of using the standard SIMMs.

[Fungus]

As a comment for mentioning the amount of 74 series logic used in that computer, have a look at this 386 mainboard from IBM:

Yep, those first boards were monstrous. Even so all floppy disks, serial ports, parallel ports, hard drive controllers, etc. were on separate ISA cards.

I remember being amazed by my first "all in one" ISA card that had hard disk, floppy, serial and parallel on the same expansion card.

Kids, today..

[Fungus]

It's lore and I'll try to find some proper evidence but my recollection is that IBM never really expected their PC to take off.

It's a classic tale: The top bosses were against it, a small team of heroic engineers hacked the first one together in record time. ...the rest is history.
https://www-03.ibm.com/ibm/history/exhibits/pc25/pc25_birth.html

The 68000 needed a real 16 bit external bus, which added a lot of expense.  For example it's hard to route 16 bits on a two layer board, and DRAM was typically by 1 bit so you needed 16 RAM chips minimum.

The 8-bit bus of the 8088 was a big point in its favor.

[RupertGo]

There were two problems with the 68K for the IBM PC - Motorola couldn't deliver parts, and it had disbanded its design team before it realised it needed peripheral chips (I seem to remember the thinking was that the 6800 stuff could do the job, but I wouldn't swear to it.)

IBM wanted something quick and dirty, and Intel had all the bits ready. If you look at the original PC, it's basically Intel reference designs glued together. As someone who had to write a lot of low-level 8086 code back then, I still bear the scars. I also did 68k on the Amiga, which was cosmically nicer, but that was never where the money was. (The 386, however, was enormously better than the 8086/286, and writing native 386 code felt like doing grown-up programming for the first time...)

The Sharp is an amazing machine, it's most of an arcade box with a general purpose computer in there as well. The design effort must have been really something.  But there can't be many post-Lisa 68K designs with 5.25" floppies.

[hikariuk]

It doesn't show well in that video but it was a totally smooth 60Hz parallax scroll over most of the screen. Useless for a game because it used all the RAM and only had enough CPU left for a single sprite, but not a problem for a demo.

Was that using the BLiTTER on the later STs or just running on the M68k?

[hikariuk]

I still have my M68k assembly language programming guide book.

[Fungus]

It doesn't show well in that video but it was a totally smooth 60Hz parallax scroll over most of the screen. Useless for a game because it used all the RAM and only had enough CPU left for a single sprite, but not a problem for a demo.

Was that using the BLiTTER on the later STs or just running on the M68k?

Just on the 68000.

It's been 30-odd years since I wrote that, but... IIRC the trick is to store 16 pre-shifted versions of the graphics in memory so no data shifting is needed during screen update (you pay the price in RAM). It also has 4 colors so you only copy half the usual amount of data (normally 16 colors=4 bitplanes, the demo uses 4 colors=2 bitplanes).

[Sigmoid]

IBM wanted something quick and dirty, and Intel had all the bits ready. If you look at the original PC, it's basically Intel reference designs glued together. As someone who had to write a lot of low-level 8086 code back then, I still bear the scars. I also did 68k on the Amiga, which was cosmically nicer, but that was never where the money was. (The 386, however, was enormously better than the 8086/286, and writing native 386 code felt like doing grown-up programming for the first time...)

Actually, while the PC may have been a sorry piece of hardware when viewed through the eyes of the enthusiast, looking at it from a management and engineering perspective, it's a masterpiece of unprecedented beauty. They built something without building anything.

Nowadays it's corporate fad to brandish bon mots like "maximize the amount of work not done". Well, they certainly did that. They also brought open architecture into a field dominated by vastly incompatible proprietary solutions.

Well done, IBM.

[hikariuk]

Actually, while the PC may have been a sorry piece of hardware when viewed through the eyes of the enthusiast, looking at it from a management and engineering perspective, it's a masterpiece of unprecedented beauty. They built something without building anything.

The 8086's rather bizarre register conventions are also an artefact of some crafty engineering (like BX being the only register you can use for memory access).  It made porting from the 8080 based CP/M machines automatable.

[Kevman]

Man, I've been wanting an X68000 ever since I saw the Game Sack episode on it a few weeks ago. I was shocked to see Dave with one, what a coincidence! Also stupidly jealous.

It seems to have some serious advantages over the Amiga (though you can't beat MODs). Remember, though that its 2 years newer and 2 years is a loonnngg time in the computer world!



Interesting to see the Demoscene come up again here.  Has anyone been to any Demoparties lately? I do the livestreaming for the American ones.

[pickle9000]

So I guess the next question is will Dave try and bring the thing back to life? Could have a very high view count seeing as it's rare. Then perhaps auction it off or use it as a contest prize.

[TiN]

Love those chip names and silk screen on bottom for debug aid

[fsanches]

MAME has an X68000 driver (in the MESS portion of the project), but I am not sure how well does it emulate this system.
MAME developers would definitely benefit from it in order to improve the emulation. I am not sure, Dave, what will you do with this unit after you're finished recording the teardown videos. If it is going to simply stay sitting in a storage room, then perhaps you could consider donating this unit to one of the MAME/MESS developers. Perhaps R.Belmont would be the ideal choice. See this thread where he mentions his work on trying to emulate the X68000 system:

http://forums.bannister.org/ubbthreads.php?ubb=showflat&Number=28935

It is worth mentioning here, as well, that MAME/MESS source code is a very rich source of technical documentation on these kinds of devices, since the project's main stated goal is to document the history of arcade game hardware and old digital equipment in general as well.

The X68000 driver source-code can be seen here:
https://github.com/mamedev/mame/blob/master/src/mess/drivers/x68k.c

It is filled with comments covering the technical architechture of this system and I would call your attention specifically to this portion which ties together the several parts of the system emulation:
https://github.com/mamedev/mame/blob/master/src/mess/drivers/x68k.c#L1637

happy hacking,
Felipe Sanches

PS: I am a code contributor to MAME and MESS but I am not yet a member of the core development team.

[RupertGo]

The PC may have been a thing of beauty from some perspectives, but never from mine. We had to build hardware and write software for it, and got bitten on the bum by many of the shortcuts in the specification. The ISA bus, for example, which assumes that worst-case (within spec) timing will never happen, and has a whole bunch of other stuff just undefined. Including, if I remember correctly, the mechanical spec for the mounting bracket for expansion cards. And as for the undershoot...  early ASICs did not like a bus-worth of that dumped to substrate by the clamp diodes. And that's before you start to battle with what's actually implemented by the cloners. 

And the BX register wasn't the only one that could access memory. SI and DI did that too. What was it - Accumulator, Base, Count and Data for AX, BX, CX and DX? No, those weren't problems; a Z80 guy like myself liked the extra room and non-orthogonal just gave one's code a lovely baroque feel.

The segment registers, though, one did not like. One did not like them really rather a lot, although it was possible, once you'd been through the pain of learning to think that way, to commit some extraordinarily evil tricks with data structures that those airy-fairy C programmers found quite beyond the pale.  And no snooty MMU to tell you off.

We've come a long way in thirty-odd years.