Still a nice chip to learn about microprocessors on(say compared to the now ruling ARM family). Nice flat 24 MB memory, easy to learn instruction set. Had three neat microcomputers based on and it's bigger brothers were the muscle for Apollo and Sun Workstations. Nice lineage.
It's too bad they didn't fix the internal architecture bottlenecks in the 68000 series until MC6833x line came out.
Oddly enough the Zilog ZNEO has a instruction set almost identical to the MC68k, except a unified register set and some DSP hardware thrown in.
This is my favorite collectable...
An Engineering Sample of the 88100 Motorola Risc processor, with a RamBlade (Prop and 512KB SRAM) beside it.
It was given to me hot of the press and under NDA on my visit to Motorola in Austin TX.
The DIP64 were originally 0.9" wide and 0.1" pitch.
But I recall Hitachi brought out a micro (think it was a 6805 derivative) that was a DIP 64 0.6" wide and 0.7" pitch which made it a similar footprint to the DIP40. This was much better than the Rockwell modem chips that were staggered 0.05" pitch and about 0.9"? wide. Very flimsy pins so that sockets often failed.
I mentioned elsewhere, I really like the concept used in the ARM chip with the stacked DDR Ram on top as used in the Raspberry Pi.
Oh the cool things one could do with a 64 pin propeller chip Take off 8 pins for power/xtal etc, and that is still a lot of pins for ram access and general I/O. In a nostalgic sort of way, I think it is a bit of a pity these big DIP chips are considered old fashioned.
This was much better than the Rockwell modem chips that were staggered 0.05" pitch and about 0.9"? wide. Very flimsy pins so that sockets often failed.
We used the Rockwell chipset back in the 80's in a couple of production units. We had about a 10% failure rate because of those pins/sockets.
When my first Mac, an original 128 model, finally burned out its power supply, I took the thing apart and marveled at the 68000 chip in the center. How amazing and modern that board looked -- all through-hole parts, no surface mount, DIP memory chips all layed out so pretty.
If only IBM had gone with that big bad 68000 with it's huge linear address space. Oh no, we had to suffer 64KB segments of memory and funky memory expanders for decade before the 32 bit 386 came out. Then we had to wait forever for MS to produce Windows 95 that actually operated in 32 bit mode.
Yes I know there was Windows NT but that was not for normal people.
If only IBM had gone with that big bad 68000 with it's huge linear address space. Oh no, we had to suffer 64KB segments of memory and funky memory expanders for decade before the 32 bit 386 came out. Then we had to wait forever for MS to produce Windows 95 that actually operated in 32 bit mode.
... Then we had to wait forever for MS to produce Windows 95 that actually operated in 32 bit mode.
Yes I know there was Windows NT but that was not for normal people.
What makes that statement even more funny is that Windows 9x still ran the OS in mostly 16-bit mode, with 32-bit APIs. It wasn't until Windows XP that there was a true 32-bit home-user OS. That's actually the reason that the Pentium Pro was limited to workstation use; it had very poor 16-bit performance, so was slower than the Pentium when running Windows 95. The Pentium 2 was essentially a Pentium Pro with fast 16-bit instruction execution.
Gordon.
Ha, I remember the engineers, more senior than me, opening up the first IBM PC in the company and sneering at how crappy it was.
Back in the 90s I took a class called 'Microcontroller tech', which consisted of ONE period in the classroom, where the teacher gave us the basics on the i8051 and related chips. The rest of the term we were supposed to form groups and build 'something with sensible input, processing and sensible output' based on the i8051.
(A buddy and I made a timer with 4 relays, support for 4 more, 10 independent programs, countdonw timer and correct calendar from 1980 to 2080)
When chatting with the teacher he told us that if any of his students had ever made the PC as a project, he would have failed him.
Is there any other computer where a fault in the Keyboard controller can result in memory errors?
(In fact, the controller can be OK, but a dud keyboard can still cause the same problems...)
If IBM Engineers didn't like the 68000, why couldn't they have picked the Zilog Z8000 instead of the bastardised 8088?
If IBM Engineers didn't like the 68000, why couldn't they have picked the Zilog Z8000 instead of the bastardised 8088?
Intel introducing the 8088:
"The 8086 is okay, but here's the exact same thing with 8 of it's 16 bits chopped off...
What?
No, the package has the same number of pins...
Yes, really...
Relax, 8-bits will save you SO MUCH on PCB routing!
No, we still use the same pins for data and address...
Stop laughing!"
If IBM Engineers didn't like the 68000, why couldn't they have picked the Zilog Z8000 instead of the bastardised 8088?
A the time, the reason doing the rounds were that IBM chose Intel is they were able to buy a chunk of Intel whereas they could not buy a chunk of Motorola. Probably would have been the same with Zilog.
IBM built a board that allowed the IBM to connect to its mainframes. (We built one as well) IBM's was the one that could bring down the whole mainframe. Made us laugh at the time
Thing is to my mind the 8086 was still almost an 8 bit micro.
Yes really. It had 16 bit registers and instructions and so on but like 8 bit micros still only had a 16 bit Program Counter, Stack Pointer and such. The directly addressable memory was 64KB, just like the 8 biters. If you wanted to go over 64K you had to use the segment registers to get you to different areas of a, get this, only 1MB address space. Not much better than adding paging hardware to exapand an 8 bitters memory.
Had it been a proper scale up from 8 to 16 bits the addressing registers would have gone to 32 bits. For a 4Gb memory address space. The would have been no segment registers.
Given that the 8086 is so crippled internally one might as well go the whole hog and make the 8088.
Having designed the worst micro processor in the world you are now poised to conquer the market for decades. Amazing.
I'm no expert on 808x architecture, apart from the limited hacking on the 8088 in my Hero 2000, but wasn't the 8086 at least able to fetch an entire instruction in only one of it's convoluted "address, ALE, then data" cycles?
The 8086 put out a 20 bit address, then tweaked ALE, to latch it externally, then put out 16 bit data or read it in.
8088 did the same but only 8 bit data at a time.
As for fetching instructions, remember that it's instructions were not all the same size. As little as one byte, DAA, for example, upto 4 bytes or more if an address is supplied in the instruction or a segment override prefix is used. I forget, thank God. So, many address-ALE-data cycles could be required to fetch an instruction.
Contrast to the Propeller where all the instructions are 32 bit and take the same time to execute.
If you could work with plastic you could put a small sm
uC and a lot of other stuff into a package that was identical
to that huge DIP. I just bet there are a lot of hobbyists that
would like to buy such things. You could put several ARM
SOCs inside that huge DIP. :-)
Thing is to my mind the 8086 was still almost an 8 bit micro.
Yes really. It had 16 bit registers and instructions and so on but like 8 bit micros still only had a 16 bit Program Counter, Stack Pointer and such. The directly addressable memory was 64KB, just like the 8 biters. If you wanted to go over 64K you had to use the segment registers to get you to different areas of a, get this, only 1MB address space. Not much better than adding paging hardware to exapand an 8 bitters memory.
Had it been a proper scale up from 8 to 16 bits the addressing registers would have gone to 32 bits. For a 4Gb memory address space. The would have been no segment registers.
Given that the 8086 is so crippled internally one might as well go the whole hog and make the 8088.
Having designed the worst micro processor in the world you are now poised to conquer the market for decades. Amazing.
This 16 bit program counter and segment register idea was not much different from what minicomputers of the time did. A 32 bit address space would have been overkill for such a slow cpu, and memory prices at the time would have made use of a 32 bit address space unlikely. A 24 bit address register along with indexed and relative addressing for all applicable instructions would have been great though.
All things said and done, a lot of the microcomputer chip designs showed a lack of foresight and innovation. They pretty much followed the trail blazed by minicomputers and mainframes. There was quite a bit more design innovation in microcontroller market, and I think the Propeller leads the way there.
Thing is to my mind the 8086 was still almost an 8 bit micro.
I was quite upset at the time that Intel and IBM were advertising it as 16-bit. In those days path to memory was IT and 8088 had an 8-bit path to memory no matter what magic was happening inside the plastic. At its core the 808x were really 8080's on steroids; they were meant to be "source code compatibile" with 8080 assembly and that's why they had 16-bit PC and segment registers. You could basically take a 8080 assembly program and recompile it and run it on an 8088 with very little modification, even though the object code was very different. A lot of early PC software was ported from CP/M in this manner, and if it hadn't been I doubt the PC would have ever taken off.
The 8088/8086 were really ahead of their time and as such not very well suited to the tech of their day; 808x software tended to be bloated compared to 8-bit software, it wasn't really much faster (especially the 8088), and the cache destroyed the ability to use cycle-counting to implement reliable time delays, which was a very common technique at the time. But IBM didn't want to use the Z80 because Tandy/RS was using it, they didn't want to use the 6502 because Apple was using it, and the only other similar tech possibility was 808x.
While I will always say the original 8088/8086 were Smile compared to their contemporary competition, the design was visionary in that it was able to maintain object-code compatibility over an unbelievable series of upgrades. I have little utilities I wrote for the non PC compatible HP 150 II which still run on Windows 7. The design is now as old and feeble as it was new and premature in the 1980's, and something needs to kick it to the curb. Unfortunately we seem to have entered another 1970's-like cycle of chaos and mutual incompatibility.
The old 68k was a joy to work with. I can't say that about ARM SOC's or the current Intel MPU's. When embedded ARMs come with manuals the size of the Art of Electronics you know you're dealing with SOCs that aren't meant for the average person. The same thing for the OS's that run on them. Big and complicated.
I suspect it's also reason why retro computing is still popular. The machines and their OS's can still be comprehended by a single person.
Martin Hodge
Posts: 1,246
Comments
-Phil
Is that some kinda crack?
The ADC-LU and SPCD on behalf of Sleepy, Bashful, Doc, Dopey, Happy, Sneezy and Grumpy
Still a nice chip to learn about microprocessors on(say compared to the now ruling ARM family). Nice flat 24 MB memory, easy to learn instruction set. Had three neat microcomputers based on and it's bigger brothers were the muscle for Apollo and Sun Workstations. Nice lineage.
It's too bad they didn't fix the internal architecture bottlenecks in the 68000 series until MC6833x line came out.
Oddly enough the Zilog ZNEO has a instruction set almost identical to the MC68k, except a unified register set and some DSP hardware thrown in.
I saved it because it was soooo huge..
Jeff
An Engineering Sample of the 88100 Motorola Risc processor, with a RamBlade (Prop and 512KB SRAM) beside it.
It was given to me hot of the press and under NDA on my visit to Motorola in Austin TX.
The Plastic DIP's look bigger than the ceramic:
My DECtalk is bigger than yours!
Picture taken today. Still working!
But I recall Hitachi brought out a micro (think it was a 6805 derivative) that was a DIP 64 0.6" wide and 0.7" pitch which made it a similar footprint to the DIP40. This was much better than the Rockwell modem chips that were staggered 0.05" pitch and about 0.9"? wide. Very flimsy pins so that sockets often failed.
I mentioned elsewhere, I really like the concept used in the ARM chip with the stacked DDR Ram on top as used in the Raspberry Pi.
We used the Rockwell chipset back in the 80's in a couple of production units. We had about a 10% failure rate because of those pins/sockets.
http://www.computermuseum.li/Testpage/Chip-Rockwell-R6511.htm
-- Gordon
Yes I know there was Windows NT but that was not for normal people.
Ha, I remember the engineers, more senior than me, opening up the first IBM PC in the company and sneering at how crappy it was.
This sounds very familiar some how.
Back in the 90s I took a class called 'Microcontroller tech', which consisted of ONE period in the classroom, where the teacher gave us the basics on the i8051 and related chips. The rest of the term we were supposed to form groups and build 'something with sensible input, processing and sensible output' based on the i8051.
(A buddy and I made a timer with 4 relays, support for 4 more, 10 independent programs, countdonw timer and correct calendar from 1980 to 2080)
When chatting with the teacher he told us that if any of his students had ever made the PC as a project, he would have failed him.
Is there any other computer where a fault in the Keyboard controller can result in memory errors?
(In fact, the controller can be OK, but a dud keyboard can still cause the same problems...)
If IBM Engineers didn't like the 68000, why couldn't they have picked the Zilog Z8000 instead of the bastardised 8088?
Intel introducing the 8088:
"The 8086 is okay, but here's the exact same thing with 8 of it's 16 bits chopped off...
What?
No, the package has the same number of pins...
Yes, really...
Relax, 8-bits will save you SO MUCH on PCB routing!
No, we still use the same pins for data and address...
Stop laughing!"
IBM built a board that allowed the IBM to connect to its mainframes. (We built one as well) IBM's was the one that could bring down the whole mainframe. Made us laugh at the time
Yes really. It had 16 bit registers and instructions and so on but like 8 bit micros still only had a 16 bit Program Counter, Stack Pointer and such. The directly addressable memory was 64KB, just like the 8 biters. If you wanted to go over 64K you had to use the segment registers to get you to different areas of a, get this, only 1MB address space. Not much better than adding paging hardware to exapand an 8 bitters memory.
Had it been a proper scale up from 8 to 16 bits the addressing registers would have gone to 32 bits. For a 4Gb memory address space. The would have been no segment registers.
Given that the 8086 is so crippled internally one might as well go the whole hog and make the 8088.
Having designed the worst micro processor in the world you are now poised to conquer the market for decades. Amazing.
8088 did the same but only 8 bit data at a time.
As for fetching instructions, remember that it's instructions were not all the same size. As little as one byte, DAA, for example, upto 4 bytes or more if an address is supplied in the instruction or a segment override prefix is used. I forget, thank God. So, many address-ALE-data cycles could be required to fetch an instruction.
Contrast to the Propeller where all the instructions are 32 bit and take the same time to execute.
uC and a lot of other stuff into a package that was identical
to that huge DIP. I just bet there are a lot of hobbyists that
would like to buy such things. You could put several ARM
SOCs inside that huge DIP. :-)
This 16 bit program counter and segment register idea was not much different from what minicomputers of the time did. A 32 bit address space would have been overkill for such a slow cpu, and memory prices at the time would have made use of a 32 bit address space unlikely. A 24 bit address register along with indexed and relative addressing for all applicable instructions would have been great though.
All things said and done, a lot of the microcomputer chip designs showed a lack of foresight and innovation. They pretty much followed the trail blazed by minicomputers and mainframes. There was quite a bit more design innovation in microcontroller market, and I think the Propeller leads the way there.
I was quite upset at the time that Intel and IBM were advertising it as 16-bit. In those days path to memory was IT and 8088 had an 8-bit path to memory no matter what magic was happening inside the plastic. At its core the 808x were really 8080's on steroids; they were meant to be "source code compatibile" with 8080 assembly and that's why they had 16-bit PC and segment registers. You could basically take a 8080 assembly program and recompile it and run it on an 8088 with very little modification, even though the object code was very different. A lot of early PC software was ported from CP/M in this manner, and if it hadn't been I doubt the PC would have ever taken off.
The 8088/8086 were really ahead of their time and as such not very well suited to the tech of their day; 808x software tended to be bloated compared to 8-bit software, it wasn't really much faster (especially the 8088), and the cache destroyed the ability to use cycle-counting to implement reliable time delays, which was a very common technique at the time. But IBM didn't want to use the Z80 because Tandy/RS was using it, they didn't want to use the 6502 because Apple was using it, and the only other similar tech possibility was 808x.
While I will always say the original 8088/8086 were Smile compared to their contemporary competition, the design was visionary in that it was able to maintain object-code compatibility over an unbelievable series of upgrades. I have little utilities I wrote for the non PC compatible HP 150 II which still run on Windows 7. The design is now as old and feeble as it was new and premature in the 1980's, and something needs to kick it to the curb. Unfortunately we seem to have entered another 1970's-like cycle of chaos and mutual incompatibility.
I suspect it's also reason why retro computing is still popular. The machines and their OS's can still be comprehended by a single person.