Should Sinclair have better foreseen the future of RAM?

[1024MAK]

Should Sinclair have better foreseen the future of RAM?

It was possible, at least with the 16K ZX Spectrum model, to expand the RAM beyond the 64K limit of the Z80 microprocessor. This would have been by using bank switching/paging.

For example, it’s possible to have had an expansion providing an extra 64K of RAM, giving a total of 80K bytes (yes, this could also be done internally on issue two through to issue 4S boards, by modifications to the board).

But should Sinclair have made this easier, by providing an extra control line on the expansion port edge-connector, so that a 64K RAM expansion could also be used on 48K machines (giving a total of 112K of RAM)?

An extra control line, or better, two extra control lines, would have allowed a RAM expansion to provide a more flexible bank switching arrangement than the (third party) internal 80K system that was available.

And later on, even bigger expansion RAM packs could have been available, e.g, a 256K one.

Vote NOW!

Mark

[merman]

RAM issues really affected several machines during the early 1980s. Jack Tramiel at Commodore took a gamble on the price going down (and MOS Technologies being able to make RAM chips cheaper) for the C64 - after very dodgy decisions on the amount of RAM in the VIC-20 and Ultimax 64.

You can definitely see the decision to go 16/48 with the Spectrum as part of Sinclair’s thought processes in keeping the machine cheap. The extra hardware/control lines to address extra memory would have complicated the design and possibly another point of failure - with the wobbly old ZX81 RAM pack fresh in his mind, he would have wanted to avoid even more consumer grumbling.

[spider]

I would say yes in a way, but my take on it is:

RAM was very expensive (iirc) at the time.

What should of been foresaw perhaps was with the expanding "home computer" market in general, and more production of machines (not just Sinclair, generally) supply/demand/production costs would of gone down a bit hence over the longer term a RAM chip would slowly be cheaper.

Not a great way of explaining it I know, sorry.

Regarding the (slight lack?) of foresight with expansions yes even though it does complicate the design a bit as mentioned by Merman, this 'extra' would of been less than the RAM cost, unless it involved more control chips (but these were not hugely expensive even then ?)

EDIT... Although much later "in the game" I'm not sure if the MGT Sam way it worked (if I read correctly somewhere I've not tried) in that the switching happened behind the scenes in Basic as least so you could in theory do a POKE 131070 , 201 etc etc. I realise this paragraph is slightly off topic, unless you meant also 'user management' of the expanded memory too.

[bluespikey]

No. He should have stopped being a stingy skinflint with the Spectrum. He was right to move to the QL, but that was hobbled by another load of cheap decisions which didn't belong on a business machine, rather than a consumer machine like the Spectrum. He should have been less cautious.

[AndyC]

48K of RAM was quite a lot at the time and, in all likelihood, designing a banking scheme would've increased the cost of hardware far more than was necessary - especially if you weren't also going to change the ROM to allow addressing it in a pseudo -linear way (as the SAM does).

If you were going to put some sort of banking arrangement in at the time, focusing on the ROM would've made more sense. Could have quite easily implemented something similar to the Amstrad CPC arrangement with Upper/Lower ROMs that can be paged independently. That would've allowed for a much faster/more featured BASIC and given space for 64K of RAM as well.

[Eugene C.]

Maybe he was not thinking of a future over 2 years for the Spectrum.

[Joefish]

It would have been better if you could replace the RAM in 16K chunks from the expansion port rather than just the ROM. It would have allowed for bigger game cartridges and easier RAM expansion. But there's no need for extra RAM control lines, really. The Z80 only has a 16K address space and you can quite easily use the OUT command to control memory paging.

The problem wasn't that there weren't dedicated lines for extra memory paging, but that there wasn't a standard documented way of using an OUT command for memory paging (just as there wasn't - originally - an 'official' joystick address or protocol). An official memory expansion may have led the way, but I guess there just wasn't the demand at the time for what would have been a fairly expensive expansion.

[toot_toot]

But surely the problem was around having a decent (and fast) storage system? I mean, 128K games took up to 15 minutes to load from tape, would we have waited 30 minutes for a 256K game to load??!!

Two things I think would have helped extending the Spectrum’s lifespan, one would have been a better storage option (that wasn’t a Microdrive) and the other would have been to, somehow, get rid of Attribute Clash while retaining backwards compatibility. The Spectrum could have likely competed with the NES, MasterSystem and Gameboy as a cheap gaming system, but by 1990 the Spectrum’s colour clash just made games look old, even when compared to the C64. Insult to injury was games taking 15 minutes to load…

[Matt_B]

Maybe he was not thinking of a future over 2 years for the Spectrum.

Almost certainly. Sinclair had supplanted the ZX80 and ZX81 very quickly, and they were already lining up a ZX83 when the Spectrum was released.

As it turned out, the QL was late and a bit of a lemon, so the Spectrum kept selling.

[Mpk]

He should have brought a GUN to the Baron of Beef.

Also there should only have been 1 model. 48K of RAM, and he should have called it the 64k Spectrum, the fool. No-one knows what RAM and ROM are even now - add them up, bigger number on the box, quids in Rodney. He'd have been President of Eurasia by now.

[Matt_B]

He should have brought a GUN to the Baron of Beef.

Also there should only have been 1 model. 48K of RAM, and he should have called it the 64k Spectrum, the fool. No-one knows what RAM and ROM are even now - add them up, bigger number on the box, quids in Rodney. He'd have been President of Eurasia by now.

It should be the 65K Spectrum, even, because it has 65536 bytes.

[AndyC]

Two things I think would have helped extending the Spectrum’s lifespan, one would have been a better storage option (that wasn’t a Microdrive) and the other would have been to, somehow, get rid of Attribute Clash while retaining backwards compatibility. The Spectrum could have likely competed with the NES, MasterSystem and Gameboy as a cheap gaming system, but by 1990 the Spectrum’s colour clash just made games look old, even when compared to the C64. Insult to injury was games taking 15 minutes to load…

The NES, Master System and Gameboy are entirely (and carefully) designed entirely around their video hardware though. It's not really something you can just hack onto an existing design.

The C64GS, Amstrad GX4000 and even the Amiga CD32 are all pretty good examples of what happens when you try to console-ify an existing computer design - it just ends up compromised in ways something designed from the ground up doesn't have to be.

[1024MAK]

Just to clarify, by extra control line(s), I mean in a similar way to the /ROMCS, but for the RAM. This would have been simple for the upper 32K (just a cheap resistor and making use of a spare contact on the edge connector). But not so simple for the lower RAM area.

Mark

[cj7hawk]

In fairness, almost nothing uses the lower 16K of ROM, so you can easily expand this way - but practical sense at the time was that people only used 48K for commercial products - that included the 128, so things like RAMDISK made sense to use the rest.

The Spectrum had a lot of things other computers of the era didn't necessarily have - such as a full expansion bus. It should be possible to use this to bypass anything inside the case, to a fair extent. You can replace the CPU with an external CPU and then you can decide yourself what you want to do with the internal memory, and the clock signal will still be slowed by the ULA to avoid video memory contention in the lower 16K bank.

Technology was moving pretty quickly - and Sinclair had an eye towards the future - perhaps more than most. I think the Loki might well have challenged even the PC and would have definitely challenged the Atari and Amiga... If only Sinclair had the business sense to make it work. He was very much the Elon Musk of his era, and his business failed before he could make better ideas a reality.

As for RAM, follow the ZX80 and ZX81 progression to the ZX Spectrum - And even then the Speccy came out with 16K for practical reasons, which did make sense at the time.

If you follow the trends of the era, the Spectrum was slightly ahead of it's time, while trying to be a budget system. The QL reflects Sinclair's slowly changing intent... Perhaps the most misunderstood issue of the era was that Business was a bigger market than Games and that Business computers needed more power than gaming computers... How that was changed with the 3DFX VooDoo card for the PC. But keep in mind this was a decade later, so technically, while he was moving in this direction, he was already a decade before his time.

There were a LOT of technologies that emerged in the early 80s which impacted technology development. Sinclair was smart enough to realize picking the best of them and "stealing" their ideas was a good move, but he never got to complete his work.

External RAM however, was not one of them. It was never a good idea, and RAM really did need to be expanded within the machine itself - Ram Pack wobble was a serious issue and bluetak was not a good solution. If all of the memory is on the board, then why go further?

After all, early PC type systems came with as little as 64k of memory at times ! This is with a 8088 processor. It wasn't until Microsoft wrote an OS to handle 640K that manufacturers started adding empty sockets to their boards as a sales gimmick. It worked. Next people had 128K machines, then 256, and finally after that, people started using the full 640K.

I think Sinclair saw the future of RAM and took advantage of it at the right time and price for what he was selling.

[Matt_B]

I'd disagree on RAM pack wobble being an inherent problem with the concept. That was mainly a problem with Sinclair's expansion for the ZX81.

Third party RAM packs for the ZX81 were far more reliable and the ones that were released for the Spectrum were mostly made by the same companies.

Lots of other computers had external RAM expansions without issue too and it was also fairly commonplace for games console cartridges to include additional RAM. I doubt that would have happened if they all suffered from wobble.

[zx64]

He could also add internal 16K RAM to the first 16K of address space and make it accessible one of the following ways:
a. make it possible to turn off ROM and have this new RAM appear there
b. keep the ROM. Both ROM and RAM are in the lower 16K. When you read, you read from ROM. When you write, you write to RAM. This RAM is used as a video memory. So, we would have 16K of video RAM, and 48K of uncontended memory for programs. No color clash and better resolution

[Lethargeek]

Probably the simplest way to get rid of the colour clash (and to make more and better looking game conversions possible) was to add another mode with a bigger and improved ULA - low-res 4 colours 4x8 tiles (with rectangular pixels - common for many other home computers). It would use the same screen memory and the same display/contention timings. Make ink=paper case the mode switch and a selector for 8/16 customizable 4-colour palettes.

[firelord]

I believe the goal was to create very quickly a cheap/affordable machine for everyone.
Better specifications would mean :
1. More design time
2. More expensive
3. Would be starting to compare spectrum with "higher specs" machines. (eg why 64k and not more etc..)

If #3 would happen then #1 and #2 will rise very much.

In general I think Sinclair wanted the main advantage to be the price. Not gaming and not anything else....

[AndyC]

To be honest, if you look at how memory was often used on 128K machines, banking was probably always the wrong choice anyway. It's complicated to manage and all too easy to mess up.

A better approach would probably have been to provide expanded RAM as a form of high speed storage, with some form of DMA controller to allow data to be quickly LOADed or SAVEd from it.i believe the C64 REU works something like this (although there more out of necessity than anything else) and have the ability to very rapidly load blocks of memory into the screen etc can offer significant advantages over a banking scheme.

But, again, planning this up front would probably have cost a lot in terms of extra hardware and been a big gamble on the price of RAM coming down enough to make it worthwhile. I'm not sure that was a given, especially given that pretty much all computer manufacturers at the time we're hopelessly convinced that everyone would be writing their own programs entirely in BASIC and, for that, 48K is already pretty good.

[Lethargeek]

I believe the goal was to create very quickly a cheap/affordable machine for everyone.
Better specifications would mean :
1. More design time
2. More expensive
3. Would be starting to compare spectrum with "higher specs" machines. (eg why 64k and not more etc..)

If #3 would happen then #1 and #2 will rise very much.

In general I think Sinclair wanted the main advantage to be the price. Not gaming and not anything else....

No, depending on the actual specs it may even lead to less time and less/no more expensive. What if Sinclair skipped the 16K option and the designers worked with less constraints? One thing, it was possible to implement a better video mode (taking a bit more memory but more convenient layout thus faster output). Then improve the ROM BASIC maybe. Small improvements shouldn't change the price that much, and it even may have been ended being cheaper than developing and producing both 16K and 48K

[firelord]

No, depending on the actual specs it may even lead to less time and less/no more expensive. What if Sinclair skipped the 16K option and the designers worked with less constraints? One thing, it was possible to implement a better video mode (taking a bit more memory but more convenient layout thus faster output). Then improve the ROM BASIC maybe. Small improvements shouldn't change the price that much, and it even may have been ended being cheaper than developing and producing both 16K and 48K

I agree wil all these and they are correct but, I think the design was done in a hurry.
If they had another year maybe they would be able to make a much better machine but, they might have lost the "cheap/affordable" title from some other company .
Even if it had more faults I think Sinclair would have launched it anyway!
From the movie I understood that this was the general idea.

[AndyC]

I'm also not convinced there was much demand for "a better video mode" at the time. The resolution was a massive step up from the ZX81 and the addition of limited colour was probably seen as good enough. I don't think anyone imagined the Speccy still clinging to life in the early 90's and struggling with early 80's design limitations.

It was a different story by the time the 128K machine came along and there probably should have been additional video modes added at that point - although I suspect that would have suffered from backwards compatibility constraints anyway (if not from the hardware then from developers not wanting to shut out users from their software).

[animaal]

The Sam Coupé was released in 1989, but I think I remember reading that it was scheduled for an earlier release than that, had been delayed, and by the time it was released it was a tough sell when compared to an Atari ST (or an Amiga, although those were more expensive).

If Sinclair had released the Sam Coupe as the successor to the 48K Spectrum, maybe in 1987, would it have been more successful? It seems to have most of the features that people mention in this thread. RAM was very expandable, so an initial 1987 release might have had the same 128KB as the Spectrum 128 and +2 models. I'd have been excited by the better graphics modes and 3.5" floppy drive.

[Lethargeek]

for me the Sam Coupé always looked overengineered and underpowered (even more than the Amstrad CPC)

[Ralf]

No.
For me RAM was never a problem. People managed to put incredible lots of stuff into existing RAM.

And C64 had exactly the same amount of RAM yet it managed to run huge games because it had a disk drive which allowed
data to be uploaded quickly and which became a standard.

Also memory paging was for me always awkward from a programmer point of view and I recall many other people don't like it too.

Zx Spectrum to became better needed extra graphic modes and cheap disk drive or other fast storage. Unfortunately Clive
failed with microdrives, then went bankrupt and then nobody cared to put any serious work into Spectrum development.
128kB models for me were a quick work, offering very little beyond the standard.