The ultimate total number of ZX Spectrum computers produced

[Lethargeek]

I've missed something. Some of early Russian home computers had means or add-ons for hardware+software emulation of ZX Spectrum, namely:
- Орион https://zx-pk.ru/threads/26616-emulyato ... atnye.html
- Вектор http://sensi.org/scalar/ware/605/
- Арго http://sannata.org/konkurs/2011/kt1119.shtml

if we count these with proper Spectrums, then the Amiga is a PC clone

[p13z]

it always puzzled me why the scene was/is mostly european even for the american home computers (both 8 and 16 bits ones)

...many valid answers...

but not for the Apple-II it seems, afaik it was much less popular in Europe, and so there wasn't a scene to speak of

...and a very valid response...

So, another thing that occurred to me is this:

Before the WWW, software/audio/video piracy was not really thought of, or treated as, a real crime anywhere - other than the USA. In America there was a real threat from civil proceedings and organisations such as the FBI. The first time I ever heard of anyone actually being in trouble for piracy of any kind came in the mid nineties - when the FBI and FAST went for BBS operators based in Europe, from America.
The "scene" largely originated from people showing off their skills for hacking/modding/cracking and redistributing games, just for the notoriety and/or a measly income from BBS phonelines (which probably never even covered the cost of running them). I guess this could have been a scary enough prospect if based in the USA to prevent any real scene from blossoming?
Thoughts?

[Alone Coder]

One more clone - Samsung SPC-650 made for Korean market. No sales figures are known.

New info about ex-USSR: people I've asked confirm than ZX Spectrum was more widespread than all the Soviet home and school computers taken together, the most notable was the abovementioned 600K+ ZX Spectrum video author (previously I was only comparing the figures to BK/DVK/UKNC line, that is 1 million, and ignored the 8080 based computers that were many).

New info about Spain (well, old, from ERBE circa 1986):
"Importabamos directamente de Inglaterra. Tendriamos un margen de un 20% o 30%,
porque comprabamos producto final acabado. Pasados dos anos, teniamos el conocimiento de
que en el mercado espanol habia una base instalada cercana al millon de Spectrums." (https://grupomainjobs.com/wp-content/up ... HEOSE2.pdf p.22)
Comparing this to September 1985 figure of 180 000 Spectrums imported by Investronica (controlling 70% of the market)(https://ddd.uab.cat/pub/tesis/2018/hdl_ ... mc1de1.pdf p.255), we see that Investronica must have produced around 1 million clones in post-tax 1985-1987 (if 1987 was not a successful year, or else there must be more). Later the Amstrad's clones were imported, and this explains the big Spanish game making activity as compared to ex-USSR. MicroHobby was sold 100 000 copies per week in 1986 (https://grupomainjobs.com/wp-content/up ... HEOSE2.pdf p.119) that is well above ZX-Review's 30 000 copies per month in 1994. However, MicroHobby covered other 8-bit platforms, too.

[Alone Coder]

The "scene" largely originated from people showing off their skills for hacking/modding/cracking and redistributing games, just for the notoriety and/or a measly income from BBS phonelines (which probably never even covered the cost of running them). I guess this could have been a scary enough prospect if based in the USA to prevent any real scene from blossoming?
Thoughts?

I think, the scene activity blossoms if a child owns one single computer platform for his whole childhood, to learn it top to bottom and love it. USA were too rich for this

[Ivanzx]

One more clone - Samsung SPC-650 made for Korean market. No sales figures are known.

New info about ex-USSR: people I've asked confirm than ZX Spectrum was more widespread than all the Soviet home and school computers taken together, the most notable was the abovementioned 600K+ ZX Spectrum video author (previously I was only comparing the figures to BK/DVK/UKNC line, that is 1 million, and ignored the 8080 based computers that were many).

New info about Spain (well, old, from ERBE circa 1986):
"Importabamos directamente de Inglaterra. Tendriamos un margen de un 20% o 30%,
porque comprabamos producto final acabado. Pasados dos anos, teniamos el conocimiento de
que en el mercado espanol habia una base instalada cercana al millon de Spectrums." (https://grupomainjobs.com/wp-content/up ... HEOSE2.pdf p.22)
Comparing this to September 1985 figure of 180 000 Spectrums imported by Investronica (controlling 70% of the market)(https://ddd.uab.cat/pub/tesis/2018/hdl_ ... mc1de1.pdf p.255), we see that Investronica must have produced around 1 million clones in post-tax 1985-1987 (if 1987 was not a successful year, or else there must be more). Later the Amstrad's clones were imported, and this explains the big Spanish game making activity as compared to ex-USSR. MicroHobby was sold 100 000 copies per week in 1986 (https://grupomainjobs.com/wp-content/up ... HEOSE2.pdf p.119) that is well above ZX-Review's 30 000 copies per month in 1994. However, MicroHobby covered other 8-bit platforms, too.

Interesting information here!

Just a little correctio, Microhobby did not cover other computers, it was a only-Spectrum magazine

[p13z]

I think, the scene activity blossoms if a child owns one single computer platform for his whole childhood, to learn it top to bottom and love it. USA were too rich for this

That is probably another pretty valid reason, yeah.

[1024MAK]

but not for the Apple-II it seems, afaik it was much less popular in Europe, and so there wasn't a scene to speak of

Apple computers were available in the U.K., but were significantly more expensive (*) compared to the other computers that were available (so seen as poor value for money). And the number of shops/retailers that stocked them was rather limited. In comparison, Sinclair, Commodore and many other computers were available from lots of different shops/retailers. Including big name stores in U.K. ‘high street’ (main) shopping areas.

In the education market in the U.K., Acorn with the computer developed to the BBC requirements, called the BBC Micro / BBC Model A / BBC Model B that has a very good version of BASIC (later called BBC BASIC) became the most common computer in U.K. schools. Although some schools also had other computers, the government encouraged schools to buy only approved computers (meaning models from U.K. companies).

Hence for education purposes at home, a lot of families bought a BBC Micro.

So in the U.K. there was next to no education market for Apple, and due to the cost, next to no games machine / home computer market either.

I don’t know what the situation was in the rest of Europe.

In terms of now, there is a German Sinclair forum. And many German owners of Amstrad machines. Similarly, there are French owners of Amstrad machines with a smaller number of Sinclair machines. Lots of Spanish owners of Spectrum machines. But I don’t know the whole picture. As I don’t follow the Commodore scene (8 bit or Amiga) and I know even less about the Apple scene (assuming there is one).

(*) note that generally speaking, most products made in the U.S.A. and imported to the U.K. were more expensive.

Mark

[1024MAK]

I think, the scene activity blossoms if a child owns one single computer platform for his whole childhood, to learn it top to bottom and love it. USA were too rich for this

I’m not sure about this. Just look at the variety of computers available in the U.K. in 1982 to 1984

• Acorn Atom
• Acorn BBC A
• Acorn BBC B
• Acorn Electron
• Advance 86a
• Amstrad CPC 464
• Atari 400
• Atari 800
• Camputers Lynx
• Commodore Vic 20
• Commodore 64
• Colour Genie
• Dragon 32
• Genie I
• Genie II
• Jupiter Ace
• Mattel Aquarius
• Memotech 500
• Memotech 512
• Multitech MPF II
• Nascom 2
• Newbrain
• Oric 1
• Sinclair ZX81
• Sinclair ZX Spectrum 16K
• Sinclair ZX Spectrum 48K
• Sinclair QL
• Sord M5
• Tandy TRS-80 MC-10
• Tandy TRS80 model 1
• Tandy Colour Computer
• Texas Ti99/4A

All these were less than £400. And yes, the sales figures for some of these was not very high. I may also have left some out!

In comparison the cheapest Apple was the Apple IIe at £980!

So for the owners of the less popular machines, you will often find them switching to more popular models in search of games (or maybe other commercial software).
And as reliability was not great for various computers, people also often bought a different make if their machine died. And as various manufacturers withdrew from the market, the number of available machines dropped. So your choice became more limited.

In addition, only the popular machines had dedicated commercial magazines published. And the only source of information for most people was the supplied manual, information from user groups, or the commercial magazines. So although there were general computing magazines (that were not make or manufacturer specific), the best source of information was definitely the various magazines dedicated to the popular machines.

Mark

[PeterJ]

You missed the Mattel Aquarius [mention]1024MAK[/mention]

But in all seriousness I agree with you it's all about four things:

Software availability
Price
Magazine support
School usage

It's like competitors to Android and iOS. Many have tried and failed but once you have a small number of huge successful systems it's very hard to break into that market.

I did enjoy reading Your Computer though which being multi-format they reviewed all these oddities.

[1024MAK]

You missed the Mattel Aquarius @1024MAK

I think my DRAM refresh circuits may be failing

Added the Mattel Aquarius and the QL. Almost certainly there are some others still missing…

Mark

[Lee Bee]

Funny to think only a million Amstrad Speccies ever made. Those machines seemed to be everywhere in the late 80s and such a huge part of my life. Most of the boys I knew had a Speccy, CPC or C64.

Even if you include all the clones and variants, Speccies are still only about a fifth of the 62 million NESes produced!

[Alone Coder]

Even if you include all the clones and variants, Speccies are still only about a fifth of the 62 million NESes produced!

It's not possible to write your own NES game on a NES ZX Spectrum is a workstation compared to it

[Alone Coder]

This 1989 Santaka ad http://naunaunau.narod.ru/articles/0512 ... 989-12.pdf p.65
mentions several other ZX Spectrum clones:
- УНИ ПЦ (Hungary)
- Роботрон 29001 (GDR)
- Меритум-1 (Poland)

I have no information about these.

[Alcoholics Anonymous]

...But I don't know of any Spectrum users in Canada.

Well now you do I was a TS2068 owner back in the day. For Canadian numbers, you can accurately divide down the US numbers by about 10 for almost anything. There was similar ownership level, the same brands, the same magazines, and so on.

The C64 was definitely not aimed at business use as previously stated in this thread. Business computers were CP/M machines and then PCs. The C64 was a games machine from the very beginning. I think the video game crash in North America in the early 80s was actually a move to home computers from the Atari and Intellivision consoles. I think the shift happened because (1) the C64 and other computers were more capable than the Atari 2600 and Intellivision in terms of games and (2) pop culture became focused on computers through films like Tron.

For a very brief time, Timex held around 50% of the US computer market. There was a strong establishment of ZX81/TS1000 users but this was just before the computer market transformed. Before about 1982/1983 or so, the computer market was about hobbyists and experimenters and the ZX81/TS1000 opened up computer ownership to a much larger part of that group because it was super cheap compared to the TRS80, Apple I/II, Atari 800, Commodore PET, etc. But as the price of the TS1000 fell, it started getting picked up by the general population curious about computers and the number of units sold rose dramatically. This was helped by the fact Timex was selling computers alongside its watches through retailers all over the US and Canada. The VIC-20 is oft claimed to be the first computer to sell 1 million units *within the United States* but the TS1000 was only a month later. If you consider the ZX81 and TS1000 the same computer then I think it's quite likely that title belongs to the ZX81/TS1000. The TS1000 was not the right machine for the regular people that started buying and it began to garner a bad reputation among those buyers because they couldn't "do anything". I'm sure this reputational worry figured into Timex's calculation about whether to stay in the computer business.

Anyway, a market was discovered and other manufacturers followed suit cutting prices. The people buying needed something to do with their computers and that was going to be playing games. The same transformation happened in the UK as you can see from the computer mags where they went from technical to aimed at games and kids. Timex decided to exit the computer market just as the exponential growth in sales was starting so the TS2068 was pretty much stillborn as it was discontinued after only being on sale for 6 months. The community continued to support it in magazines and with sw/hw for many years after.

As far as schools go, the schools I went to had TRS80s, Apple IIs and PCs. I never saw Commodore computers in the school system (education is controlled provincially here and I lived in 3/10 provinces during that time) and I don't think that was the case in the US either although I don't know that for sure. We were learning basic on Apples and Pascal on PCs. The Apples probably did very well in education because they were easily networked and CP/M cards were available. I did see C64s for teaching Comal in a Danish school.

I do agree with the premise of this thread though -- I do think the number of zx spectrum units sold worldwide is deeply underestimated as its clones were available in a lot of countries and sometimes as the dominant machine in those countries, with none of them being counted toward that popular 5 million figure. In contrast, the C64 was only made by Commodore and it alone sold worldwide so could count every single unit made. However I do think it likely that the C64 still remains champ, even if the 20 - 30 million figures are considered unlikely, simply because it became so big in the United States where a large fraction of computers worldwide were sold.

[Alone Coder]

I've checked the sources of 12.5 million figure for C-64. There are discrepancies.
Namely, the last used report (end of 1990: 11 million, dead link) doesn't match the previous report (1988 January To date, over 7 million Commodore 64 and 128 computers have been sold. [806.12] https://web.archive.org/web/20160317010 ... 641988.htm). This looks like an extrapolation of 7 million figure, as if the author thought it was about C-64 only, but it's 7 million Commodore 64 and 128 computers TOGETHER.

Serial numbers slightly save the situation. However, the boards 326298, /A, /B, /C (launched at the same time) might be different configurations of the same board with common numbering (I know this method at my job). The same goes to 250407 with letters (launched at the same time) and 250425 with letters (launched at the same time). This correction reduces the number of C-64 to 9.5 millions.

[Alone Coder]

Another reason to doubt the official numbers of C-64 is the total number of the versions. Even if we count these configurations as different versions, there are only 16 versions of C-64.
Compare it to 14+1 versions of ZX Spectrum from Sinclair (5 million units) and 5 versions of ZX Spectrum from Amstrad (1 million units). https://spectrumforeveryone.com/technic ... um-models/
Also compare it to 9 versions of Amstrad CPC (3 million units).
Or compare it to 164(!!!) versions of MSX (9 million units): https://www.msx.org/forum/msx-talk/hard ... bers-topic

The only computer that looks like C-64 in this ratio is Amstrad PCW - with official (and bloated) circulation of 8 million units (6 versions).

[Alone Coder]

One Yugoslavian (namely Slovenian) ZX Spectrum clone is known, named Ines (from Spica). For unknown reasons, is requires an extra card for ZX Spectrum compatibility.
Any sales figures?

[Pegaz]

One Yugoslavian (namely Slovenian) ZX Spectrum clone is known, named Ines (from Spica). For unknown reasons, is requires an extra card for ZX Spectrum compatibility.
Any sales figures?

Its not a clone, Ines is just a custom case with a professional keyboard, for Spectrum 48k.
It was produced in Slovenia (Ex-Yu), also theres a business package (Word Processor, Database, etc.), under the same name.
btw, Spectrum clones are not manufactured in ex-Yugoslavia, but Slovenian company Iskra, imported and sold Spectrum 16K, under the official Sinclair license during 1984.
Due to many reasons (state taxes, mostly), it was very expensive, more than 48k model, which could be easily obtained in neighboring countries like Austria, Germany...
Sales Figures are small, insignificant and have no global impact at all.

[Alone Coder]

Very interesting. Why do such a case, even without the national symbols on the keyboard?

[Alone Coder]

100 000 for Argentine

Almost right. People sent me some info about Czerweny:
"According to an article in lanacion, they made about 96000 units..." https://www.youtube.com/watch?v=hBMPcIs ... RzmGSVdqSD

[Pegaz]

Very interesting. Why do such a case, even without the national symbols on the keyboard?

Who knows... As I heard, "Iskra" didn't want to really manufacture Spectrum 16K, just made a contract with Sinclair, to get parts in Kit form, assemble and sale.
Like I said, the whole job was a failure, because people could buy a 48K model cheaper than this officially "produced" in Yugoslavia.
Similarly happened to Oric Atmos, which was imported in parts, assembled and sold in Yugoslavia, as Oric Nova, by Slovenian company "Avtotehna".

[Alone Coder]

The first source from Argentine:

Durante 1984 y 1985, Czerweny Electrónica llegó a producir unos 4000 Spectrum
por mes, en una línea de producción que también se encargaba de producir las
plaquetas y partes que requería IBM Argentina.
Sin embargo, un cortocircuito en la fábrica de Paraná provocó un incendio que
destruyó por completo las instalaciones en junio de 1986. Czerweny Electrónica
intentó retomar la producción, pero el producto estuvo fuera de mercado por un
tiempo prolongado.

La historia de Czerweny CZ Spectrum, la computadora Sinclair con sello argentino - LA NACION
https://www.lanacion.com.ar/tecnologia/ ... id1886082/

[Alone Coder]

- 600 000 Timex Color Computers were sold in USA by August 1983 (ZX Computing Aug-Sep 1983 p18)

This was Timex 1000. Timex Sinclair 2068 (the model for the USA and Canada) was produced in 80000 units https://www.retrothing.com/2009/05/time ... puter.html

Another source for Poland:
"TMX Portugal Computer sold only to Poland 200.000 FDDЗO00 units and about 700.000 Timex Computers."
https://zxpress.ru/article.php?id=17935

Another source for Argentina:
The first 2068 computer in Argentina was the TS with NTSC system, but some years later the TC 2068 comes with PAL system (here the official TV Standard is PAL) and the ZX Spectrum Emulator Cartdrige (both, the TC and the Cart was made in Portugal... I remember the label in the bottom of the unit...)
http://timex.comboios.info/history.html
(TS2068 must be within the above-mentioned 80000 figure, but TC must be counted separately)

The first source for Brazil:
https://books.google.ru/books?id=sm9rkw ... rt&f=false p.98
This sums up to 630409 8-bit computers from Microdigital (i.e. TK90X and TK95X) in 1984..1990.

There are 4 board versions of C128, C128CR, C128D: https://www.amiga-stuff.com/hardware/cbmboards.html . By proportion with 16 versions of C64 (9.5 millions based on serial numbers), it reduces the total number of all versions of C128 to circa 2.4 millions, not 4 millions.

With all the new data, the first message must be corrected (-600000 Timex, -1369591 Microdigital, -1 million Investronica, +116000 Czerweny (two years and 5 months of 4000 per month), +500000 ex-USSR, +90000 GDR, +10000 Hungary, +20000 Korea, +10000 Meritum?, +5000 Ines? (if serial numbers and official stats count for computers, not spare parts), +25000 European amateurs? (for a guess: a ZX-81 clone Galaksija was built by 10000 amateurs in Yugoslavia https://www.youtube.com/watch?v=hBMPcIs ... UVh4AaABAg )). This reduces the sum to 12.1 million ZX Spectrums. (Including Amstrad's 1 million, I don't see a reason for a bigger number.)

Amstrad models serial numbers:
The highest number I've seen for +2 was U 305209.
Spanish +2: S-033577 https://en.todocoleccion.net/video-game ... ~x55790745
+2A: 2 278902 https://imgur.com/gallery/HZ9Zm
+2A with different numbering: S030060 https://retroordenadoresorty.blogspot.c ... sue-2.html
Arabic +2A: G 021758 https://www.nightfallcrew.com/11/08/201 ... c-version/
Spanish +2B: GC 03966 031B https://woodvineblog.wordpress.com/zx-s ... b-spanish/
+3: U080886 https://retrorepairsandrefurbs.com/2021 ... storation/

[Alone Coder]

Note: Microdigital numbers include non-Spectrum machines! Paulo suggests that Microdigital may have produced about 300000-350000 TK90s and TK95s.
He also gives new counts for Timexes in Argentina:

It's very difficult to estimate the number of Speccy clones in Argentina, but we know that Drean C64s were number 1 at sales, with about 300000 units (and maybe some imported C64s), MSX number 3 with about 60000 MSX Talents sold and perhaps some more imported, let's say 40000 (probably much less) = 100000 and number 2 Speccy clones: CZs = 116000 (max) + TSs = 50000 (that's my very optimistic guesstimate, seen that the total number worldwide was 80000) + TC 2068 perhaps another 50000 (I think I'm being very generous here, since I consider there were much less TCs than TSs) = 2160000 (and I've seen some Argies commenting on YouTube saying MSX was number 2)

The number of Spectrums in Czechoslovakia needs to be reconsidered. Previously it was counted as 100000 units, but there are 505 games from this country (compare fo 2035 from ex-USSR and 3422 from Spain), that looks more like 500000 units. This might be too high for a 15.6 million country (for example, there are just 810000 for 38.3 million Poland: Timex + Elwro + Meritum), so let's take 330000 by proportion.

This reduces the total number of ZX Spectrums to 11.9 millions.

It also appeared that the 12.5 million C64 count was criticised by Pete Rittwage, the keeper of C64 Preservation Project (which serial numbers have been used in the research).
The main reasons:
- the 12.5 million figure used old data (there are now much more serial numbers in the database)
- numbers from different factories were taken together.
Sadly, C64 Preservation Project is now down, but the forum posts are still there:

So I updated my numbers based on 250469 Rev 3 and Rev 4

250469 (ignored, they are likely in numbers below)
250469 A 1996933 and HKC386449 and HKM035641 = 1996933 = 2419023
250469 B 2326852 and HKC251969 and HKM437319 = 2326852 = 3016140
250469 3 1058246 and HKC421629 and UAL020353 = 1058246 = 1500228
250469 4 2012879 and HKC1126178 = 2012879 = 3139057
250469 TOTAL 7394910 or 10074448
KU-14194HB CS14017 and F39658 = 53675

Counts
326298 286/437079 0.06%
250407 819/6508919 0.01%
250425 504/3736946 0.01%
250466 147/1143851 0.01%

250469 577/7394910 0.0078%
or
250469 577/10074448 0.0057%

KU 110/53675 0.2%

The questions are-

1) Did some people just not enter the HKC and HKM and they are the same number series? That is not likely based on that it is never a high number on those series. If I don't count them, there are still 7.4 million boards. If I do, it's over 10 million boards.

2) Did the numbers start over again for each revision? If not, there are substantially less. Too few.

Look at the small percentage of entered machines compared to what was made. We have an impossibly small sample.

The number of KU board entered is a huge percentage of what might exist. That means 1) a MUCH larger percentage of collectors with this board entered the serial, or 2) Our number of boards in existance is much to low. Same with 326298 to a much smaller amount, but I am willing to say more collectors with early machines have entered their information.

It's more likely a mix of these problems.

Also, one single person entered board revision "C" which nobody else did. I assume it was a mistake but why would they do that?

https://www.forum64.de/index.php?thread ... ost1548964
"Did some people just not enter the HKC"? For board 250407 certainly yes, because the highest HKC1211455 matches the highest number 1238489.
For board 250425 we know 4 numbers from HKC, the highest is HKC922576 (this gives the "German tank" count of 1153219 HKC units of board 250425), that is also close to the highest reported number 1323460.
For board 250466 we also know 4 numbers from HKC, the highest is HKC366795 ("German tank" count of 458493), that is also close to the highest reported number 530806.
Only for board 250469 there are 7-digit numbers that don't match the numbers from HKC.

I have changed the c64preservation site to output the board serial in the list- it was hidden for a while. You can sort only by KU board and see how the numbers that are there are spread out. I tried not to use data where there were only a few, and I purposely subtracted instances where, for example, all serial numbers started with 1.

For the "F" series KU, it was obvious the first digit is likely always 1, so I removed 100000 from my calculations. The CS ones are pretty spread out, so they were likely out there. All in Europe from 1982/1983.

https://www.forum64.de/index.php?thread ... ost1548740
This excess 100000 may be also true for unit F138546 with 250407 board.

326298 A 337532
326298 B n / a
326298 C 99547
326298 TOTAL 437079

250407 1238489 (ignored, probably in below numbers)
250407 A 1247737 and F138546 and CS112159 = 1498442
250407 B 1203526 and A62352 and CS62307 and DE55519 and HKC1211455 and S66259 = 2661418
250407 C 1218502 and HKC962606 and UA167951 = 2349059
250407 TOTAL 6508919

250425 1323460 and HKC922576 and JA054265 and UA143461 (ignored, probably in below numbers)
250425 A 1069758 and HKC701441 and UA130409 = 1901608
250425 B 971770 and HKC266130 = 1277900
250425 C 063708 HKC493730 = 557438
250425 TOTAL 3736946

250466 530806 and HKC366795 (ignored, probably in below numbers)
250466 A 438001 and HKC243358 = 681359
250466 B 407179 and HKC55313 = 462492
250466 C HKC176609 = 176609
250466 TOTAL 1143851

250469 (ignored, probably in below numbers)
250469 A 1996933 and HKC386449 and HKM035641 = 2419023
250469 B 2326852 and HKC251969 and HKM437319 = 3016140
250469 c 382140 = 382140
250469 TOTAL 5817303

KU-14194HB CS14017 and F39658 = 53675

TOTAL 17,697,773

https://www.forum64.de/index.php?thread ... ost1548473
Note: 250469 Rev 3 and Rev 4 are excluded in his final count (they might be misreadings of B and A).

The boards 326298 to 250466 were produced in 1982-1987. And the sum shows 11 826 795 of them. So, this calculation openly contradicts the official figure "1988 January To date, over 7 million Commodore 64 and 128 computers have been sold. [806.12]" https://web.archive.org/web/20160317010 ... 641988.htm (Compute's Gazette, May 1988, Volume 6, Number 5, Issue 59). Why this mistake?
a) some serial runs were counted several times;
b) some serial runs don't start from 0 or 1 or have another numbering scheme.
As we can see, almost all boards have 3 modifications named A, B, C. But the author sums them up instead of counting as a single serial run.

For an optimistic count or C64, let's count HKC units of boards 250425 and 250466 separately from non-HKC data, and use unit number F138546 as it is.


(Note: "For the "F" series KU, it was obvious the first digit is likely always 1, so I removed 100000 from my calculations." https://www.forum64.de/index.php?thread ... ost1548740 )

9390512 C64's with 16 boards gives 2347628 C128's with 4 boards by proportion, with the total number of 11 738 140.
Any pessimistic count will even reduce this number.

[Polrai]

This is from msx.org:
Let's estimate the number of some of the Z80 computers sold:
1) MSX - less than 10 millions;
2) Amstrad CPC and PCW - more than 11 millions;
3) Tandy TRS-80 - about 10 millions;
4) ZX Spectrum with clones - less than 10 millions

I think that's quite reasonable...

Investrónica produced the Inves+ only from December 1986 to December 1987, meanwhile Amstrad España continued to import ZX Spectrums from GB, I don't know how many they made, but they were far behind the ZXs, that's why I'm guessing 250000 (they were plagued with problems, it was too little too late).

By the way, it's Pablo, not Paulo (yes, that's me)