An interesting article about the history of the '486' cpu

[Andrew]

A mixture of pictures and text Brian

[Brian Gaff (Sofa]

I think sadly at the moment, Intel are falling behind on processor development and its not a new problem. AMD seem to be running rings around them on their own processors, and the new chips with different architecture are now very powerful, the arm etc. Brian

[Jethro_uk]

Whatever happened to Sinclairs idea of intelligent fault-tolerant wafers that simply blew the links to duff dies and worked as entire assemblies of chips ?

[Jeff Layman]

I think The Borg picked up on that. ;-)

[Steve Walker]

Probably yields improved and the amount crammed into a tiny section of the wafer made it unnecessary - his original intention was for the 512KB RAM pack for the QL to be wafer scale, but we have moved so far past that that it is perhaps unnecessary.

[Max Demian]

Doesn't half have some rabbit the author. Charles Darwin, complete with picture?

I recall that early IBM PC ATs had a clock rate of 8MHz, and games of the day were designed to run as fast as possible, so, when faster machines were available, they became impossible to play. Some machines had a button on the front to switch between 8MHz and the maximum, which could be up to 66MHz. The button was usually labelled "Turbo" and you pressed it for full speed.

[Steve Walker]

That started with the IBM PC clones. 4.77MHz Standard, 8MHz Turbo. AT clones were 6MHz and 8MHz at the start , but got faster later.

Our first PC used an odd processor, the NEC V20 - which was pin and code compatible with the 8088, but had extra instruction matching the 80186 and internal differences meant that it could sometimes do more per clock cycle than the 8088.

It also could be switched in code to emulate an 8080 (so could run the existing CPM-80 and applications, not that we ever did) and could run mixed 8080 and 8086 code.

[Bob Eager]

I had several PCs, and I switched all of them (V20 for 8088, V30 for

8086).

I also fitted turbo switches!

Yes, I liked the idea but never used it in the end.

[Peter Johnson]

The first PC I assembled myself was a 486. I remember when the cpu was first released there were reports of it overheating which implied that Intel had cocked up big time. Intel's response 'It's OK, put a fan on the cpu to cool it' was initially greeted with some scepticism I recall.

[Jethro_uk]

I vaguely remember the original spec was that the 486 was the 386+387 maths co processor as a standard offering.

That lasted about as long as a marketing lunch until Intel realised that a 486 with a failed MPU was a 486SX ....

This instigated the CPU-as-car vibe that dominated the late 80s and 90s with ever confusing codes and designations. All to ensure that level 3 employees were getting something level 2 weren't.

[Theo]

Chips have long been fault tolerant - parts get configured based on what works and what doesn't, and one design is sold as multiple SKUs at different price points based on functionality and measured performance.

It's easier to manipulate chips rather than wafers though, and they're more compact. Flash memory chips like micro SD cards are a stack of maybe 32 dice wired together. That's equivalent to a wafer, but you get to throw away the terrible ones.

Even with all this, keeping everything in a wafer means you're still susceptible to major faults eg if somewhere on the wafer is a power supply short you can't just set a configuration bit to disable it.

These days silicon is used like a PCB, and chips are made from a substate of silicon with wiring and multiple dies (from different fabs/processes) bonded on top.

The main place we still use wafer scale chips is solar cells, where we aren't too fussy about minor defects and area counts most of all.

Theo

[Martin Brown]

The original IBM AT was 6MHz clock which meant that it was OK to export to behind the Iron Curtain whereas later ones and Compaq & Dell clones had a 8MHz clock speed and were not COCOM approved for export to Russia.

I recall one particular week when Computer Weekly had two news reports. IBM salesman of the year gets award for sale of 2000 PC ATs to Moscow State University juxtaposed by some West German business man jailed for

3 years for exporting 200 Beebons to East Berlin. Apparently the graphics on it were just slightly too good and they stamped on him.

I still have my 486(TM) bronze key fob in daily use which has a defunct chip from the early prototype failures epoxied to the front. The feature size was just about perfect for decorative holographic effects.

[Vir Campestris]

I vaguely remember only the '486 had a cache. But ICBA to check.

Andy

[Adrian Caspersz]

My first PC, was the 80186 co-processor module inside a BBC Master computer running DOS+ and GEM, and MFM hard disks (from a skip at work).

Got bored waiting for it to boot, and the graphics were at best CGA with limited colours (or none?). So went for an 80286 12MHz AT a large Northern Telecom case from an auction, and then followed every release of Intel CPU and Microsoft OS on the NT3.5/NT4/Win2000 side of things, until Windows XP.

Bought many video cards, disk controllers, sound cards, motherboards. I eventually worked in PC parts RMA for a trade distributor, found skills in programming and application support - and now am happy to take a corporate engineered box of the shelf, rather than build my own.

Most of my friends back then invested their work earnings in gaining property. I sadly blew mine on what is now computer junk, and most of that perished in a garage flood - so nil retro gains today on eBay.

Tech idiot consumers like me will no doubt be telling the same story to others in time to come ...

[Theo]

There was a story that the US State Department prevented export of BBC Micros to somewhere because they contained a US-built 6502 processor and could <something something munitions>. Exporting the Apple II (same 6502 CPU) - no problem at all.

Theo

[Steve Walker]

A totally different area, but back in the '90s I worked for a large, industrial compressor manufacturer. We sent 8 off, skidded, gas-pipeline compressors, driven by 120 litre, V16 gas engines; along with all the ancillaries, control panels, two control rooms (2 sites of 4 compressors each); all the instrumentation, etc.; to Libya.

All properly approved and authorised by the Department of Trade or whatever it was.

This was a (supposedly) 3 year project, so stuff was all over our factory floor for years.

Half-way through the project, the top guy left our offices to become CEO of European Operations and was replaced by an American. He lasted 3 days, before asking what all the green painted stuff on the shop floor was, then jumping on the company jet back to the States when he heard it was the "Libyan job", under fear of him being locked up, as he wasn't allowed to have any dealings with Libya at all.

Despite being based in Manchester, our project team and the shop floor guys working on it, all officially reported to the Aberdeen service centre for the next couple of years, to get around the problem!

[Jethro_uk]

Weren't larger computers also never bought, but leased ?

I vaguely recall a story about a University trying to sell a PDP-11 only to discover that DEC could (and did) prevent the sale.

[bert]

ICL sold a 4-70 mainframe to Russia but then someone in customs heard it had a "clock" which hadn't been declared. We had Russian engineers over for training complete with minders. Some of them did a bunk to Canada. There was a story they wanted a CDC and wanted to pay for it with Xmas cards.

[Andrew]

My first computing job was at Redifon Computers in Crawley. They made key-to-disk data-entry systems based on a DCC copy of a data general 16 bit mini. Because of licensing restrictions they were limited to the UK and eastern europe. They had to write tape-emulators to suite a whole range of iron-bloc computer systems, which were typically heavily modified locally built copies of western computers. The Redifon software had to handle this in addition to multiple eastern european languages for programming screens and manuals. I don't remember any talk of restrictions other than the data general operating system license which limited them to those countries, though the modified CPU boards etc all made from scratch at Crawley probably wouldn't run pure the pure DG Nova operating system because they altered the address MSB to access a full 64K bytes, while on a DG nova this would give multi-level indirect addressing (I think, brain is rusty).

[Andrew]

Systime (Of Leeds fame, with an indoor fountain) fell foul of US export retrictions and had a death fight with DEC themselves because their machine were based on DEC VAX. There were rumours of PDP-11's being exported as juke boxes.