AI Slop Video claims first 40007 GAs were not made by Ferranti but by SGS?!

As you may know, YT is being flooded with AI-generated retro computing video slop. I just watched another one claiming that the first 464s shipped with an SGS ULA made in Milan, not by Ferranti, and that the tape recorder was so unreliable that you could not walk around the machine and reliably load from tape (to not disturb it) :sweat_smile:

Who is generating and watching all that slop, and does it REALLY pay off? These videos are also easily recognized because the AI voice narration can’t even speak “four six four”, let alone “six six four” correctly.

Now, the ULA statement is just nuts - I always thought that the “F” versions (see pic attached) are clearly Ferranti. Not sure about “PS” though. Anybody?

1 Like

Lord Sugar is on the record saying the 464 wouldn’t have made it’s launch date without ditching Ferranti (who couldn’t get their act together) to buy chips from other suppliers instead (eventually landing with SGS.) Ferranti did eventually supply some chips, but they weren’t in the first batch.

https://amstrad.com/product/cpc464/

1 Like

Interesting. Well, this on has a -4, which probably indicates Sept. to Dec. 1984. Anybody has any earlier Ferrantis? So maybe the AI Slop is actually right…

Fwiw I think the tape deck thing is mostly the exaggeration of how painful loading was and how often we all suffered loads errors. The reality wasn’t that bad, but all the big scenes have had stories for years about how your mum walking into the room would immediately cause the game, that had spent the last three days loading, to fail instantly. It’s just repeating generic historic hyperbole because it doesn’t really know better.

I’ve been taking a look at our decap pictures of the different models of the Gate Array and by pure chance I discovered that the 40007 and 40008 use the same style of cells and they match exactly the cells described in the 1983 Ferranti Quick Reference Guide.

The 40010 36 AA and 40010 37 AA do have the names LSI Logic and SGS on the die.

I’ve superimposed one of the cells of the 40007 on top of the relevant page of the quick reference guide and you can see they’re a perfect match!

Could that indicate that the first 40007 and 40008 GAs were made by Ferranti and the 40010 exclusively by SGS?

3 Likes

Are you sure that what you call AI slop is not a video where you have the horrible Youtube auto-translate activated ?

It really gives a feeling of AI slop even with well known youtuber.

Maybe you’re right but check this before, it can be deactivated (even if it’s not easy).

Regarding the original question, it looks like Ferranti was struggling to produce their GAs and that despite their attempts, they couldn’t make it run cooler, so Amstrad went to SGS and they got mass production and better efficiency.

That’s my 2 cents… :smiley:

Yeah, there a about a dozen of these channels by now. They all have AI-narration, AI-generated thumbnails, … and some of these channels are churning out a new video of 30 mins. every 2 days…

1 Like

well, all 40007 (the Ferranti “F” one and the other “PS” one in my picture) had the large aluminum heat sinks… so the improvements wrt. to heat only came later? 40010. Those came without heat sink.

The 40007 and 40008 are both Ferranti gate arrays. The one with the ‘PS’ marking is from Plessey Semiconductors which bought up Ferranti Semiconductor in 1988 as Ferranti started to collapse after the unwise merger with International Signal and Control. Within another year GEC had taken over Plessey and merged their semiconductor businesses to form GEC Plessey Semiconductor. Other Ferranti-originated gate arrays and ICs can be found with GPS markings on the package, although I don’t know if the 40008 was one of them.

The date codes on those parts (e.g. 8823) are most likely year and week of assembly.

Ferranti stuck with bipolar technology for their gate arrays, which had its attractions in their military and industrial markets. However, this was always a more power hungry technology than the MOS process used by pretty much everyone else (including SGS and LSI) for competing gate arrays. As we know, MOS (and later CMOS) was the winner in the end.

3 Likes

Thats a old bigger thread about this in the old forum.

The 40007 and 40008 its a 20ra043 5000R series ula from ferranti

The first gen of 40010 made in Italy by sgs was a bigger chip that the second generación from lsi with less logic Gates.

And yes, the 40007 works,they mount It in the first board versions, the z200 was made to Accept the 40010

Then years latter, until the cost-down revisión was made(that is a sea of Gates).

Amstrad made a special pcb only for the 6128 with last ferranti chips until the cost down was produced.

Generally they send the ferranti chips to the 464 and the lsi chips to the 6128.

Also the first 40010 appears in the CPC 664 that don’t support double gate array.

40007/40008 20ra043 ferranti 5000 series ula.(View spectrum uka book).

40010 1stgen->hsg3170->LL3170>37AA Italy->21X82=1722 cells

40010 2stgen->Hsg3130>LL3130>36AA>18x74=1332 cells

1 Like

So, that seems to indicate than that Lord Sugar misremembered the state of affairs, or wasn’t that involved in the process after all?

My personal opinion was ferranti/amstrad are not sure if they can made work the 40007 until the launch of the 464 so sugar search the plan b with SGS(that also source It with a clone of z80).

Then ferranti can made It work but very Heat(40007) and maybe in low cuantity.

But alan sugar as the gate array its a component essential for cpc that can’t be buy in another source. Maybe prefer maintaing the two sources, as if one fail, he can source from the another and continue development the 40008 to reduce the Heat.

And they do It years latter until the cost-down was finished. Maybe because the lsi product go to end of live and ferranti go to the same situacion. or something else apart from reducing cost.

The 40007 and the 40008 are both of the 16RA family, not the 20RA one. You can see the marking in the decap pictures.

Regarding the final part of the ULA marking, 023, I couldn’t find any information.

Data on the 16RA family is in the 1983 Ferranti Quick Reference Guide mentioned above:

2 Likes

I think the 023 is just the design number.

I have taken a number of photos of Ferranti ULAs (e.g. here’s a small 2C array.)and usually that suffix matches the marking on the package. Sometimes manufacturers are branding parts as their own on the package, as in this case, but the die always carries Ferranti’s design number.

Should be 20Ra in the service manual, i have tell It from memory. Probably a error from amstrad.

It may be that they intended to use the 20RA but ended up managing to put it in a the smaller 16RA, which would have been slightly cheaper too.

Or maybe they use It at begining same as the 40010.

The only way to know that It is decapp a 40007 from the first cpc uk with tall keys and first motherboard with the most early date possible in the serial.

It could also be that maybe they had some prototypes made on a 20RA, but I don’t think it’s really important as they’ll be largely the same but the 20RA will have more cells.

Well,from the point of view of know the complete history,always all are important.