CPU family tree - help needed

[kuro68k]

I'm creating a CPU family tree poster and could do with some help adding additional families and verifying data.

PDF file attached.

The idea is to have all the CPUs used in various computers and game consoles. I only go up to 1996 but I could add some more years if there is interest, or maybe just to make it fit a A size sheet of paper.

I don't include slight variations, or second source parts where they are identical to the original. I also don't include microcontroller models unless they were used in a computer.

I'm missing year information for NEC V41 and QED R4700. I'm going to add the Signetix 8X300, because it was used in some popular hard disks. I was thinking of adding some examples of popular machines that used each CPU, but space is an issue. I already need to reformat it to bunch things up a bit.

I know some of the 8/16 and 16/32 bit labels are controversial. I based them on my own judgement, but feel free to make your case!

Any help would be appreciated. After this I was thinking of doing a timeline of computers. Both will end up on my wall.

 

[Plasma]

Looks pretty good. Missing Cyrix on the x86 side.

 

[marcopolo]

Microprocessor chronology

The first microprocessors were manufactured in the 1970s. Designers predominantly used NMOS logic and they experimented with various word lengths. Early on, 4-bit processors were common (e.g. Intel 4004). Later in the decade, 8-bit processors superseded the 4-bit chips. 16-bit processors emerged...

computer.fandom.com

 

[kuro68k]

Looks pretty good. Missing Cyrix on the x86 side.

Thanks. You are right, I forgot Cyrix. That could get messy with their "486 in a 386 socket" type CPUs, which are related to two generations of Intel parts, but also completely original designs. I'll have to think about this. The whole Intel tree is a bit messy because of the many compatible/licenced parts. At lest NEC stopped at the 8086.

I might just make the NEC line into a bubble or something, rather than lines connecting them. Like the 8086 and 8088, my understanding is that they had mostly the same cores internally, so aren't really two separate lineages.

Microprocessor chronology

The first microprocessors were manufactured in the 1970s. Designers predominantly used NMOS logic and they experimented with various word lengths. Early on, 4-bit processors were common (e.g. Intel 4004). Later in the decade, 8-bit processors superseded the 4-bit chips. 16-bit processors emerged...

computer.fandom.com

Thanks, I'll have to go through those. Are there any that stand out? The main issue is determining if each part was used in a notable machine. For example, the NEC μCOM 4. Wikipedia has no info on what it was used in, and it's likely to only have been some obscure trainer boards and industrial products. Even so, I'd include it if the ISA was part of a lineage that went on to be used via descendants of those parts.

My Japanese is not great and it can be difficult to find ancient magazine articles and the like for these systems. The search engine on archive.org is not great.

 

[marcopolo]

Are there any that stand out?

i860
AMD 29K: The K5 was based upon an internal highly parallel Am29000 RISC processor architecture with an x86 decoding front-end (en.wikipedia.org/wiki/AMD_K5)

 

[jqavins]

You've got the Z800 and Z280 in there twice. Is there a reason, or is it a simple mistake?

You don't have RCA's COSMAC processors, the 1801 and 1802. The 1802 has some notability for its use in the Elf design, the VIP (OK, maybe that one's not so notable) and the Studio 2 game console.

Also not present is DEC's LSI-11, a two chip version of the PDP-11/23, which was used in their PRO 325 and 350.

 

[kuro68k]

i860
AMD 29K: The K5 was based upon an internal highly parallel Am29000 RISC processor architecture with an x86 decoding front-end (en.wikipedia.org/wiki/AMD_K5)

Thanks. Do you know when the Am29040 and Am29050 were released? I'm having trouble finding dates for them.

Ditto for the i860. I have 1989 as the year that the XR was released, but no date for the XP.

 

[kuro68k]

Updated diagram attached.

I am having trouble with PowerPC chips. The history is quite confused because there were IBM and Motorola lines. Wikipedia gives different dates to cpu-collection.de. Anyone have any good resources for those?

Also looking for dates for the 6508, 6509, and 6510T.

I am thinking about ways I can group by manufacturer or ISA. Maybe some shading. It would be nice to indicate where things were in production for a while, because for example the i960 only ever had one CPU (at various speed grades) but was a current project for much longer than it's lone entry suggests.

 

[jqavins]

You've got the Z800 and Z280 in there twice. Is there a reason, or is it a simple mistake?

You don't have RCA's COSMAC processors, the 1801 and 1802. The 1802 has some notability for its use in the Elf design, the VIP (OK, maybe that one's not so notable) and the Studio 2 game console.

Also not present is DEC's LSI-11, a two chip version of the PDP-11/23, which was used in their PRO 325 and 350.

Also also not present is the TI TMS9900 family.

 

[kuro68k]

Updated with lots of new additions, thanks everyone.

Still having trouble finding dates for some parts, but I'm working on it. TMS9900 still to do. Might have to make it even wider to fit that. Need to think about how it can be printed. Maybe 3x A size sheets or something.

 

[Eudimorphodon]

AMD 29K: The K5 was based upon an internal highly parallel Am29000 RISC processor architecture with an x86 decoding front-end

To expand on this a bit more/note how this muddies up the diagram: The K5 was AMD's first CPU that used a strategy of breaking x86 instructions into "micro-ops" which were executed by a RISC-like central core, in this case based on the AM29000. Depending on what you're intending to portray in a CPU "family tree" this would argue that you should actually have two lines of descent converging on the K5 instead of having it directly descended from the AM486.

Even more confusingly the K6, which uses an overall similar design philosophy, isn't actually directly descended from the K5. AMD acquired the company NexGen, which briefly sold a micro-op based design called the "Nx586" (introduced in 1994) in 1996 and based the K6 on their next-generation Nx686(*). Despite its name it's not really a descendant of the K5 and probably doesn't really merit a solid line to it.

(* I'm guessing this chart is only meant to have CPUs that actually made it to market, so if you're going to add NexGen to the chart it's open to question whether the Nx686 would deserve its own box or if you should just draw a line directly from the Nx586 to the K6. NexGen did some demos of the Nx686 in late 1995 before the acquisition, and I vaguely remember they actually got as far as listing it as "coming soon" in ads for Nx586 motherboards, but I don't think it ever actually made it out into the wild. It's not entirely clear how much changed between those prototypes and the shipping K6, IE, whether they would count as the "same" CPU. Just to be safe I'd probably do a dotted line to the K5 on the assumption that the AMD team contributed *something* to the final product.)

The Athlon is actually its own interesting kettle of fish, because some of the engineers who worked on it were recruited straight from DEC's Alpha team and the CPU actually uses the same EV6 bus protocol as the Alpha 21264. Whether that would merit a direct (or dotted) line of descent between Alpha (which seems to be absent from the chart) and Athlon/K7 is debatable but, hey, it's a thing. (The CPU core itself is mostly a fatter version of the K6 architecture, so that merits a fairly direct line.)

Another x86 line to chuck on there, if you *really* want to be complete, would be Transmeta, but in the end they didn't really amount to much.

I also don't see SPARC on there, but maybe that's a can of worms you don't want to open.

 

[Eudimorphodon]

... oh. Another CPU that probably merits at least a brief mention if you're covering PowerPC is the Motorola m88k. The 88k was *not* a PowerPC CPU, it was Motorola's independent attempt at a general-purpose RISC CPU, but when Motorola joined the AIM alliance their major contribution to the PowerPC 601 helping the IBM team graft a single-chip version of the POWER architecture core to a bus taken from the m88110, which Apple had used for their first RISC prototypes and already had support chipsets suitable for use in cheap computers available. Variants of this bus were used right through the end of the PPC750 and PPC7400 families.

... And FWIW, you should probably have lines connecting both the 7400 and 750 families to the PPC603. (Actually, make that a line going from the 603 to the 750, and then a branch to the 7400) Both the 603 and the 604 are (basically) direct descendants of the 601 (IE, they should probably be alongside each other instead of in a stack), with the 603 the "lightweight" line and the 604 the "heavy duty" option. Spiritually you could argue the 7400 is more like the 604, but technically the 7400 is a 750 with an Altivec sidecar bolted onto it rather than an actual descendant of the 604.

 

[jqavins]

Since you haven't responded to this:

You've got the Z800 and Z280 in there twice. Is there a reason, or is it a simple mistake?

You don't have RCA's COSMAC processors, the 1801 and 1802. The 1802 has some notability for its use in the Elf design, the VIP (OK, maybe that one's not so notable) and the Studio 2 game console.

Also not present is DEC's LSI-11, a two chip version of the PDP-11/23, which was used in their PRO 325 and 350.

I wonder whether you just haven't got to it, or don't intend to. No, I'm not complaining. The one part of it I really do wonder about is the duplicated Zilog processors.

Is there a reason you want that? If so, please educate me.

 

[kuro68k]

You've got the Z800 and Z280 in there twice. Is there a reason, or is it a simple mistake?

You don't have RCA's COSMAC processors, the 1801 and 1802. The 1802 has some notability for its use in the Elf design, the VIP (OK, maybe that one's not so notable) and the Studio 2 game console.

Also not present is DEC's LSI-11, a two chip version of the PDP-11/23, which was used in their PRO 325 and 350.

Sorted the Z800/280. I have the 1801 and 1802 now.

Do you have a date for the introduction of the LSI-11?

 

[kuro68k]

To expand on this a bit more/note how this muddies up the diagram: The K5 was AMD's first CPU that used a strategy of breaking x86 instructions into "micro-ops" which were executed by a RISC-like central core, in this case based on the AM29000. Depending on what you're intending to portray in a CPU "family tree" this would argue that you should actually have two lines of descent converging on the K5 instead of having it directly descended from the AM486.

Even more confusingly the K6, which uses an overall similar design philosophy, isn't actually directly descended from the K5. AMD acquired the company NexGen, which briefly sold a micro-op based design called the "Nx586" (introduced in 1994) in 1996 and based the K6 on their next-generation Nx686(*). Despite its name it's not really a descendant of the K5 and probably doesn't really merit a solid line to it.

(* I'm guessing this chart is only meant to have CPUs that actually made it to market, so if you're going to add NexGen to the chart it's open to question whether the Nx686 would deserve its own box or if you should just draw a line directly from the Nx586 to the K6. NexGen did some demos of the Nx686 in late 1995 before the acquisition, and I vaguely remember they actually got as far as listing it as "coming soon" in ads for Nx586 motherboards, but I don't think it ever actually made it out into the wild. It's not entirely clear how much changed between those prototypes and the shipping K6, IE, whether they would count as the "same" CPU. Just to be safe I'd probably do a dotted line to the K5 on the assumption that the AMD team contributed *something* to the final product.)

The Athlon is actually its own interesting kettle of fish, because some of the engineers who worked on it were recruited straight from DEC's Alpha team and the CPU actually uses the same EV6 bus protocol as the Alpha 21264. Whether that would merit a direct (or dotted) line of descent between Alpha (which seems to be absent from the chart) and Athlon/K7 is debatable but, hey, it's a thing. (The CPU core itself is mostly a fatter version of the K6 architecture, so that merits a fairly direct line.)

Another x86 line to chuck on there, if you *really* want to be complete, would be Transmeta, but in the end they didn't really amount to much.

I also don't see SPARC on there, but maybe that's a can of worms you don't want to open.

Okay, I have sorted out the K5 and K6. Fascinating history there, at a time when new players could enter the market with highly competitive products built from scratch. I have not included the Nx686 because like you I can't find evidence it came to market. I was thinking of adding chips that didn't make it into production, maybe as grey or dotted outline boxes or something. There are quite a lot of them though - it was very easy to make paper plans that ultimately came to nothing. The main issue though is that there is no way to date them. Some had expected release dates, but many were abandoned before they got as far as thinking of marketing them.

Transmeta is on my list of CPUs to add. I've been debating Alpha and SPARC. I think their family trees are pretty linear at least. I know they were used in some workstation machines, but I don't know much about them. Were they ever used for "home" computers? I have to cut off somewhere or I'll end up with all the IBM S/360 stuff and so on too.

The same applies to Itanium. I have PA-RISC on there so should probably have Itanium too.

... oh. Another CPU that probably merits at least a brief mention if you're covering PowerPC is the Motorola m88k. The 88k was *not* a PowerPC CPU, it was Motorola's independent attempt at a general-purpose RISC CPU, but when Motorola joined the AIM alliance their major contribution to the PowerPC 601 helping the IBM team graft a single-chip version of the POWER architecture core to a bus taken from the m88110, which Apple had used for their first RISC prototypes and already had support chipsets suitable for use in cheap computers available. Variants of this bus were used right through the end of the PPC750 and PPC7400 families.

... And FWIW, you should probably have lines connecting both the 7400 and 750 families to the PPC603. (Actually, make that a line going from the 603 to the 750, and then a branch to the 7400) Both the 603 and the 604 are (basically) direct descendants of the 601 (IE, they should probably be alongside each other instead of in a stack), with the 603 the "lightweight" line and the 604 the "heavy duty" option. Spiritually you could argue the 7400 is more like the 604, but technically the 7400 is a 750 with an Altivec sidecar bolted onto it rather than an actual descendant of the 604.

I have the 88k in there. I'll try to link up the PPC stuff. I need to organize that section better. My goal thus far has mostly been to just get everything added to the diagram with the right dates and connections, and organize it later.

 

[Plasmo]

I don’t think Z800 is ever a real product. It was renamed and produced as Z280. Z80 has several microcontroller version such as Z84C15.
6800 had derivatives such as 6805 and 6808, then 68hc11 and 68hc16
68000 had a 8-bit bus version, 68008. 68000 also spawned the huge CPU32 line, 683xx. P90CE201 is derived from 68000.
32016 also has an 8-bit bus version, 32008
powerPC has a MPC5xx line for automotive applications.
what about transputers?

 

[1944GPW]

How about the Signetics N3000, it was a bit slice design that comprised the N3001 central microprogram chip that managed 2-bit N3002 processing elements. There was also the later(?) N8X3001 processor.
And the special variant of the 68000 used in the PC XT/370 that was mask-programmed to run the IBM System 370 instruction set. I think that might warrant a node branched off the 68000 for that one.
The chart needs a title, and perhaps a key of company logos.

 

[kuro68k]

I don’t think Z800 is ever a real product. It was renamed and produced as Z280. Z80 has several microcontroller version such as Z84C15.
6800 had derivatives such as 6805 and 6808, then 68hc11 and 68hc16
68000 had a 8-bit bus version, 68008. 68000 also spawned the huge CPU32 line, 683xx. P90CE201 is derived from 68000.
32016 also has an 8-bit bus version, 32008
powerPC has a MPC5xx line for automotive applications.
what about transputers?

Thanks. I think you are right about Z800, but I'm not sure what to do with it. Could remove it entirely.

6805 and 6809 are there, I can add the 6808. Problem is I can't find dates for any of them. They might have been only used in one machine, i.e. made to spec for a customer. Do you have any data on where they were used?

I added the 68008. How could I forget that? Used in the Sinclair QL.

I am not adding microcontroller versions, unless they were used in micro computers. Similar to the no mini/mainframe computers rule. I might make exceptions if there is some compelling argument, like they were used in a popular home computer but not as the main CPU. For example I was thinking of adding the 6500/1, which was used in Amiga keyboards.

I also try to avoid minor variants like versions without an FPU (e.g. the EC versions of the 68020 and 68030). I have included them with some of the x86 lines though, so I need to make a decision and clean up. I think where it's the same part but binned, or the same part but a higher speed version due to process/yield improvements, I won't include it.

Do you have a date for when the 32008 was released? Might have been 1984 but I can't find anything definitive. Do you know of any machines that used it? The only one I have is the Emax sythn which was released in 1986.

Transputers are on the right hand side. I need a date for the T212.

How about the Signetics N3000, it was a bit slice design that comprised the N3001 central microprogram chip that managed 2-bit N3002 processing elements. There was also the later(?) N8X3001 processor.
And the special variant of the 68000 used in the PC XT/370 that was mask-programmed to run the IBM System 370 instruction set. I think that might warrant a node branched off the 68000 for that one.
The chart needs a title, and perhaps a key of company logos.

I can't find much info on the N3000 range. Do you have a list of parts and dates when they were released? Were they used in any computers or games consoles?

Same with the special 68000 version, do you have any details?

I am planning to add a title and a key. I'll try to get all the information in first, before worrying too much about how it looks. I'm no graphic artist though.

 

[Plasmo]

6805 and 6808 are mostly microcontrollers for embedded applications. They usually have lots of embedded flash and some RAM and special IO. They are similar to AVR with many memory/IO options so perhaps they are numerous, but minor branches of 6800 Family tree.

683xx are mostly embedded computers. 68328 is well-known because it is in Palm Pilot. 68332/68336/68340/68376 are mostly embedded engine controllers. 68360 is often an IO companion to 68040. I’ve reverse engineered a SBC based on 68302 and ported CPM68K to it. Originally it was a controller for network server.
https://forum.vcfed.org/index.php?threads/porting-cp-m-68k-to-a-recycled-board.61681/#post-746253

P90CE201 is new to me but I ported CPM68K to it 2-3 years ago. I don’t know whether or what commercial computers were based on it. https://www.retrobrewcomputers.org/doku.php?id=builderpages:plasmo:p90mb:p90mb_r1

I think you should remove Z800 so the family line goes directly from Z80 to Z280. Tilmann Reh famously designed CPU280 for hobbyists based on Z280.

I collect CPU and have several tubes of NS32008 but I don’t know any computers based on 32008.
Bill

 

[marcopolo]

Motorola Coldfire ?