CPC Plus - a missed opportunity?

[alexisread]

As a new member of the CPC world, I wondered what people thought about the CPC plus range - Too little too late?

I was entertaining the idea of using the ArnoldV chip, but modified to allow dynamic reading from separate 8bit and 16bit busses.It would allow easy addition of a 68000 cpu to turn the computer into a 16bit machine which may have stood a very good chance against the contemporary 16bitters.


As far as software goes, Amstrad had bought Sinclair and so had access to QDOS (preemptive multitasking, networking, shared DLL architecture, modular drivers, superbasic shell/scripting and compiler, included office software on microdrive), which by 1990 I believe had a GUI, and could run on the QL so inside 128k RAM.

Assuming that the GUI does run in 128k, that would allow a cheap machine, not much more expensive than the 6128+

If we allow 2*64K 8bit RAM banks for the z80 as per the 6128, but mapped also into the first 128k of the 68000 memory (alternate access between the 68k/z80 on each bank cpu interval), and have 128k 16bit ram on the 16bit bus, and the ArnoldV can access lowram on 2cycles (alternating the 64k ram banks), and highram on 1 cycle, we can have several flexible profiles for use:

note: ArnoldV switchable between 8 and 16bit busses by changing modes in software, peripherals and sound on z80, rom+expansion maps across both low and high ram. I believe that it's possible to do 640x400 (and 768x576 overscan) with the CRTC on a monitor/TV

1) Full CPC6128 compatibility - CPC and CPM profiles can use the 68k as a coprocessor as per the TRS-80 Model 16 in z80 mode
2) Games profile can use z80 copro for sound, I/O eg.gamepads, dynamic raster and sprite fx, HUD blitting/overlays (200K blitting/sec by the z80)
3) QDOS / xenix profile can use z80 for sound/gfx/IO to allow full fastram operation at 8mhz, and also use z80 as an mmu/VM support for the 68k see http://cpushack.com/cpu/cpu3.html as per the TRS-80 Model 16 in 68k mode in 640x400x32coloursx50hz, with (z80 ie. dedicated timing) midi networking from analogue port


Thanks to owning the QL and software, a bundle could include:spreadsheet, word processor (able to read locomotive docs), database, microemacs, c compiler, basic compiler, terminal, fax, email, ghostscript, archiver, vectordrawing

Prospective users could come from:CPC users, QL users, PCW upgraders (thanks to reading locomotive docs), SMS2 users (and could heterogeneously network with them over midi), TRS-80 Model 16 users (note also the possibility to port TRS80-II, xenix, and unix programs to QDOS as per microemacs), atariSTE devs (who would be familiar with the DMA and YM sound, scroll, resolution), amiga devs (familiar with 32colours+sprites), megadrive devs (familiar with the split z80/68k arch)


Note that most unix machines cost an arm and a leg in 1990, mainly because an MMU was required, which usually required an expensive CPU (68020+MMU or 68030, see the Atari TT or Amiga A3000).

It seems unlikely that Amstrad would change from 3inch discs, but the CPCplus aux port would be more useful as a general RS232 port for mice/modems, and the model would include an RF output in addition to the CPCplus ports and cassette port.

Overall, the idea of a cheap 16bitter that can computationally outperform the Amiga/ST (dedicated fastram bus with gfx on the 8bit bus), go toe to toe for games, had CPM and unix compatibility thanks to the z80 (CPM, or as an MMU), backwards compatibility with the CPC, and pitched also as a PCW office computer upgrade (to a real GUI), would have a fighting chance at becoming a classic - a missed opportunity perhaps?

[scruss]

... Amstrad had bought Sinclair and so had access to QDOS (preemptive multitasking, networking, shared DLL architecture,  ...)

At the time of the 1986 purchase, the QL had very little commercial value. Amstrad certainly never did anything with it, and the refinements (like not crashing if you looked at it funny, and replacements for the knot-in-a-box microdrives) came after the sale and wouldn't have been Amstrad's to own.

[Gryzor]

Correct me if I'm wrong, but QL's "GUI" pretty much sucked. IIRC there came others, much later on, that were much better, but the QL was dead by then.

[Bryce]

You seem to be missing the point of the CPC+. It wasn't meant to be a revolutionary new machine. It was a way of buying time and staying relevant while Alan decided what his next move would be. Putting major resources into anything Z80 at the time would have been extremely wasteful, as these people were needed for machines that would stand a chance against the next generation, not the one that was coming to an end.

Bryce.

[dragon]

Well they can be boost the cpc switching with a more powerful z80 at 8mhz for example.  That is not as expensive as 16bit cpu, retain compatibility and can seriously upgrade the filling of the cpc quality in screen.


But alan sugar wants stay in the super cheap cost zone. With the most cheap cpu and most cheap design possibly with one asic.


So at finish the extremely cheap  mode kill the plus, that not have much cpu force to move alot or bigger sprites, hardware or not hardware. When made at same time other cpu intensive calculations for example the aracnides of the prehistorik 2 that slow down because the calculation of the circunference.

[alexisread]

Thanks everyone for the considered responses.

Yes as far as I'm aware the original QDOS GUI sucked - it was two panes and a shell window at the bottom. However by 1990 Tony Tebby had written SMS which was ported to the Atari ST in 1992/93 which I believe has a usable GUI (certainly SMS2 did).

Also by 1990, it was evident that the z80 didn't have a 16bit successor, so architecturally there was little point in continuing to compete with 16bit machines using a high-clocked z80 (though as a games machine that route has merit vis the Konix multisystem).

As the majority of (non-msdos) computers were using 68K cpus, and given the launch price of the megadrive was so cheap, it would have been feasable to add a 68000 and say 128k ram to the system and come in at a cheap price with CPC backwards compatibility, no major development required, as the ArnoldV ASIC would be suitable (tweaking the bus).


Obviously a new architecture requires software, and you're all probably correct in that Amstrad would have a licence to use QDOS but not SMS. As Tony was still developing it, it would be a known quantity to buy in the OS to run on a 68000 machine.

Again, the idea was for a cheap CPC backwards-compatible 16bit machine. That it could run unix and act as a migration path from the by then out-of-date PCW range would also help uptake (see the sales figures for the PCW16 - DOA as it had it's own non-upgradable OS and secondhand PCs were getting cheap).

This sort of fits with the Amstrad MO - a cheap but good machine (though this time it's a 16bit machine) that could be developed in a known timeframe for a known cost.

[TotO]

I think the missed Amstrad opportunity on the 16-bit market was to be not chosen by Sega to distribute the Mega Drive in Europe.
For me, the Amstrad Plus and GX4000 systems are a second choice and we have missed an Amstrad branded console and computer around the Sega hardware... But, the Mega PC as consolation price.

[Gryzor]

If I'm not mistaken, for every pound of increased cost there would be three pounds worth of increase in retail price, so even what you suggest would mean missing price points.

Huh, never heard of SMS, should check it out!

And, funny you should mention Konix which was dead in the water!

[Bryce]

Well they can be boost the cpc switching with a more powerful z80 at 8mhz for example.  That is not as expensive as 16bit cpu, retain compatibility and can seriously upgrade the filling of the cpc quality in screen.

But alan sugar wants stay in the super cheap cost zone. With the most cheap cpu and most cheap design possibly with one asic.

So at finish the extremely cheap  mode kill the plus, that not have much cpu force to move alot or bigger sprites, hardware or not hardware. When made at same time other cpu intensive calculations for example the aracnides of the prehistorik 2 that slow down because the calculation of the circunference.

I think you are underestimating the costs involved here. Moving from 4 to 8MHz would be massively expensive, even more so if the machine is to stay compatible with the older titles. The machine would have probably doubled in price just for these changes.

Some examples:

1 - Amstrad most likely had a deal to buy the Z80 at very low bulk costs. They would now have redundant stock and have to pay for a much more expensive chip.
2 - The RAM would not be fast enough for 8MHz, so new RAM (same supply/redundancy problem as with the Z80) would be required, plus new circuitry which now has to be completely re-evaluated and tested.
3 - The computer would need a metal shield on top and bottom of the PCB.
4 - The firmware would need considerable changes for all time critical functions.
5 - A new Gate Array would be required.
6 - Additional circuitry would be required to get the AY, FDC and CRTC synchronised with the CPU.

And that's just to go from 4 to 8MHz with a Z80! Add a 68K to the mixture and the price would have been beyond anything on the market.


Bryce.

[dragon]

Why do you tell 8 to 16mhz¿?.


I'n the sgs catalog of the 1990 sgs sell 2,5mhz,4mhz,6mhz,and 8mhz.


Somebody have made overlooking here. Really cpc archetecture and the ram can't handle the 6mhz version?.

http://www.bitsavers.org/components/sgs/_dataBooks/1990_SGS_Z80_Microprocessor_Family_Handbook.pdf


They also can have made another approach. Take license and build the z80 in the asic Instead the hdma sound cpu.

[Bryce]

Sorry, that was meant to be 4MHz to 8MHz. I was still on my first coffee and in the process of repairing an Amiga at the same time.
Yes, SGS was selling an 8MHz part, but it would have been many times more expensive than the 4MHz version, with less discount possibilities because it would have been more popular.

Bryce.
Will edit above... and grab another coffee.

[tjohnson]

My thought on this is that Amstrad clearly got the market wrong and it cost them dear .  I bought into the revised models unfortunate few did and the rest is history.  Despite the success of the megadrive it still didn't end well for Sega, one can never bank on continued success.

[Sykobee (Briggsy)]

Yeah, about a year too late and a little short feature-wise of what was really needed to be a worthy cheap UK 'amiga lite'.


Given the Z80 really couldn't be upgraded with the CPC design, the only real upgrade area is the custom ASIC.


What they got right: Smooth hardware scrolling, DMA audio (albeit one channel), 16 16-colour sprites, 4096 colours, raster interrupt (e.g., smooth colour gradients)


Issues:
Sprites need loading/unloading from ASIC - increasing onboard ASIC memory to allow more sprites (even if only 16 could be displayed at once) could have made this function easier to use for animations, and saving main memory.
Colour limitation in MODE 1 - spot sprites helped (switchblade) - but some way to colourise games better was needed (e.g., attributes to colour each character differently, this is how the NES and MS got more colours at this resolution - not a lot of memory needed for this).
Maybe another AY chip for better audio.
Not enough games


But it all comes down to cost. Clearly having a larger ASIC wasn't feasible for this project. They did what they could. In the end it failed, costing everything for what could have been a couple of quid per device. Clearly £99 was a price target they couldn't miss for the console.

[alexisread]

Sykobee, my thoughts exactly.
I wasn't thinking to change the z80 speed at all - in the Megadrive it's clocked at 4mhz for SMS (Sega Master System) compatibility, with the 68k clocked independently. TBH if the 68k was clocked at 8mhz and 16bit ram at 2mhz you don't need to change bus timings at all for the z80 portion.

Better colour and sprite/HUD handling could be done in this system by the z80 acting as a coprocessor to the 68k - the 4mhz z80 can blit around 200K/sec, so we have a choice of either handling the sound, handling sprites/HUD data, or handling raster colour splits with the 68k (and DMA for sound) doing the rest.
btw Mode 1 is limited to 4 colours, but the sprites I thought had their own independent palette of 16 colours as CRAM is implemented in the ASIC? With the z80 handling raster splits shouldn't that allow enough colour for mode 1? I don't know enough about attributes to comment really, certainly the 68k could push enough data (32k) for an extra mode as an alternative.


Obviously a price point of £99 wouldn't be feasable and £150 is possibly too close to the Megadrive launch price. I think that going the computer route would have had merits though as you can look at the sales figures for France for the 6128+ and go from there. If the computer was popular then you could launch a reduced price console when the game lineup was there - that seems to be the make or break of a console, large and relevant game library (and the c64GS/GX4000 weren't relevant despite the large back catalog).
Lastly they did spend £20 million on advertising, so...

[Sykobee (Briggsy)]

To be honest I'm not sure of the final sales figures of the GX4000, 464+ and 6128+. £20m is a lot if they only shifted 100k or 200k in the end!


And that's another thing they could have done - 128KB base across the range - even in 1989 that would have been cheap.


Oh, and stick a couple of (possibly slightly tweaked for the plus) older CPC games on the shipping cartridge as well - not using EPROMS would have made this economically viable, and people like it when systems come with multiple games.


I think the 68k+Z80 would have changed the system too much, made it a Megadrive without the Megadrive benefits. TBH £99 was probably a good target.


The hardware scrolling saved a huge amount of Z80 CPU time over classic games that scrolled in software - allowing larger software sprites at the very least.

[VincentGR]

4-8-16 MHz then it would be another Atari ST.
They should design something completely different but that was the thing back then.
They painted the same machine and sold it for a decade... all of them.

[tjohnson]

Talk of z80, 68k cpu and custom asics all sounds rather fanciful tbh, far more cost and complexity to engineer I imagine.  The Amiga was popular here in the UK at they time but again commodore wasn't able to sustain that success.

[dthrone]

Look, they're technically 'at least as good as the snes' - I have it on good authority!

[siccoyote]

I would have said the plus series just could have done with an 8mhz Z80, but then as has been pointed out the memory would have to be changed. The Graphics had been upgraded so to a layman they look almost as good as the ST / Amiga, the sound is as good as the ST. But it doesn't have the speed or memory to do anything amazing.
Maybe they should have released only the C128+, with a 2nd Z80 built in?
But then it probably would have failed as they weren't commited enough to compete with Atari and Commodore.

[ChaRleyTroniC]

As far as software goes, Amstrad had bought Sinclair and so had access to QDOS (preemptive multitasking, networking, shared DLL architecture, modular drivers, superbasic shell/scripting and compiler, included office software on microdrive)

I'm not 100% sure Amstrad bought the rights to the QL.


But 1990 was too late in any case. Amstrad's real missed opportunity was 1988's Sinclair PC200, which could have been a contender had it not sucked quite so badly.

[Gryzor]

AFAIK Amstrad got all of Sinclair's IP?

[alexisread]

I'm not 100% sure Amstrad bought the rights to the QL.


But 1990 was too late in any case. Amstrad's real missed opportunity was 1988's Sinclair PC200, which could have been a contender had it not sucked quite so badly.

I've not looked at this computer before - you could've stuck the cpc+ arnoldV chip in that (had it been developed earlier) and you're right, it would've been a hit! Ironically it'd look remarkably similar to the Atari STE with the PSG DMA sound, scrolling, GEM desktop, 320x200x16 colour gfx, memory, analogue gameports and disk drive
It was really cheap for the PC compatibility :O http://www.retroisle.com/sinclair/others/pc200.php