+1 for removing that battery asap.

I've removed the battery. It was beginning to leak on the hidden side, but nothing serious. I've wiped the board with a cloth imbued with diluted citric acid and then rinsed it with a cloth imbued in distilled water. The board has also a connector for an external battery, so I don't need to solder a replacement.

Just an update... I gave away the 286 board (but I kept the IIT 287, since it was missing from my collection) . I got a smaller board based on the Headland HT12A chipset and... what a difference!

This board allows me to choose between 0 and 1 wait states. With 1 wait states performance is more or less as the other board; but with 0 wait states, I get a nice boost in performance.

System Info 8 went from 9.6x to 15x
Landmark 60 increase was less marked... from 27 MHz to 30 MHz
CheckIt 3 Dhrystone score went from 3800 to 4900

All in all considered, the 286 is a very fast 16bit chip... Imho Intel went overboard by adding a MMU and new addressing modes. A waste of transistors that didn't get really around the many other 8086 limitations and made life more miserable for software developers. It might also have delayed the transition to 'clean' 32 bit OSes on the PC . If Intel went straight from 8086 to 386 without adding an in-between architecture, developers and users alike would have jumped ship much sooner.

However, the speed boost compared to the 8086 is alone worth having it; besides... 640K ought to be enough for anything, right? 😀

MSxyz wrote on 2024-03-08, 18:12:

Just an update... I gave away the 286 board (but I kept the IIT 287, since it was missing from my collection) . I got a smaller board based on the Headland HT12A chipset and... what a difference!

This board allows me to choose between 0 and 1 wait states. With 1 wait states performance is more or less as the other board; but with 0 wait states, I get a nice boost in performance.

Sweet! The Headland 12 chipset is supposed to be nice. Someday people might find the MrBios that was written for it. I'm jealous. The only 286 I have is a giant Packard Bell board with a funky BIOS that's either on the the low end of compatibility or the high end of incompatibility, depending on the task.

MSxyz wrote on 2024-03-08, 18:12:

Just an update... I gave away If Intel went straight from 8086 to 386 without adding an in-between architecture, developers and users alike would have jumped ship much sooner.

That's one way of looking at it. Would have been nice if they did what the market needed on the first try. I think intel was working under the assumption that once their customers tried their new protected mode environment, they would never want to go back to real mode. Turns out people really liked their legacy software. Maybe the 286 protected mode was a necessary step in the learning process to understand what needed to go into the 386. And maybe the first few 286 chips were just to slow to really benefit from prememptive multitasking anyway.

Seems like there was a similar mistake with the 386 and a 32bit bus. Microchannel was less attractive because it had a high cost and you had to junk your old cards to boot. Intel didn't get a 32bit bus that people liked until PCI worked along side ISA. But part of that was timing. A fast 32 bit bus didn't really become imperative until after processors started to > 40 Mhz. Back in the day, microchannel on a 20Mhz 386 couldn't make a compelling case for its existence anymore than OS/2 could on a an 6 Mhz 286. And PCI wouldn't have been as nice as it was if the industry hadn't learned lessons about the usability of reference disks.

douglar wrote on 2024-03-08, 19:47:

Turns out people really liked their legacy software. Maybe the 286 protected mode was a necessary step in the learning process to understand what needed to go into the 386. And maybe the first few 286 chips were just to slow to really benefit from prememptive multitasking anyway.

If your performance requirements don't exceed the performance of an IBM PC/XT, you could run two sessions of protected-mode-compatible productivity software (spread sheets, word processors) on an 8MHz 80286 just fine - you "just" need to find such software. Microsoft XENIX would be a suitable operating system to run those programs, and multiple users can use the same computer using serial terminals.

douglar wrote on 2024-03-08, 19:47:

Seems like there was a similar mistake with the 386 and a 32bit bus. Microchannel was less attractive because it had a high cost and you had to junk your old cards to boot. Intel didn't get a 32bit bus that people liked until PCI worked along side ISA. But part of that was timing. A fast 32 bit bus didn't really become imperative until after processors started to > 40 Mhz.

The 32-bit bus of the 80386DX is not just used for I/O cards, but also for RAM access. As soon as you run any kind of 32-bit software that uses aligned 32-bit accesses, this software profits from 32-bit RAM access. Additionally, 32-bit RAM access provides a slight performance boost on fetching instructions, but that boost wasn't worth the price difference between a 16-bit 386SX and a 32-bit 386DX systems. You shouldn't assume that Intel designed the 80386 as "processor for IBM PC compatible systems" only. In systems designed from scratch with a 32-bit architecture in mind (think 32-bit UNIX workstations), the 80386DX architecture made a lot of sence.

mkarcher wrote on 2024-03-09, 12:03:

If your performance requirements don't exceed the performance of an IBM PC/XT, you could run two sessions of protected-mode-comp […]
Show full quote

douglar wrote on 2024-03-08, 19:47:

Turns out people really liked their legacy software. Maybe the 286 protected mode was a necessary step in the learning process to understand what needed to go into the 386. And maybe the first few 286 chips were just to slow to really benefit from prememptive multitasking anyway.

If your performance requirements don't exceed the performance of an IBM PC/XT, you could run two sessions of protected-mode-compatible productivity software (spread sheets, word processors) on an 8MHz 80286 just fine - you "just" need to find such software. Microsoft XENIX would be a suitable operating system to run those programs, and multiple users can use the same computer using serial terminals.

douglar wrote on 2024-03-08, 19:47:

Seems like there was a similar mistake with the 386 and a 32bit bus. Microchannel was less attractive because it had a high cost and you had to junk your old cards to boot. Intel didn't get a 32bit bus that people liked until PCI worked along side ISA. But part of that was timing. A fast 32 bit bus didn't really become imperative until after processors started to > 40 Mhz.

The 32-bit bus of the 80386DX is not just used for I/O cards, but also for RAM access. As soon as you run any kind of 32-bit software that uses aligned 32-bit accesses, this software profits from 32-bit RAM access. Additionally, 32-bit RAM access provides a slight performance boost on fetching instructions, but that boost wasn't worth the price difference between a 16-bit 386SX and a 32-bit 386DX systems. You shouldn't assume that Intel designed the 80386 as "processor for IBM PC compatible systems" only. In systems designed from scratch with a 32-bit architecture in mind (think 32-bit UNIX workstations), the 80386DX architecture made a lot of sence.

Message received . Its often too easy to look at everything from the perspective of a DOS user. Even the hoary 286 had loftier targets than desktop systems when it was designed. Multitasking, or even task switching, multiple DOS environments wasn’t on the feature list when the 286 first came out.

https://en.m.wikipedia.org/wiki/Intel_80286

Intel did not expect personal computers to use the 286. The CPU was designed for multi-user systems with multitasking applications, including communications (such as automated PBXs) and real-time process control.

douglar wrote on 2024-03-09, 14:52:

Message received . Its often too easy to look at everything from the perspective of a DOS user. Even the hoary 286 had loftier […]
Show full quote

mkarcher wrote on 2024-03-09, 12:03:

If your performance requirements don't exceed the performance of an IBM PC/XT, you could run two sessions of protected-mode-comp […]
Show full quote

douglar wrote on 2024-03-08, 19:47:

Turns out people really liked their legacy software. Maybe the 286 protected mode was a necessary step in the learning process to understand what needed to go into the 386. And maybe the first few 286 chips were just to slow to really benefit from prememptive multitasking anyway.

If your performance requirements don't exceed the performance of an IBM PC/XT, you could run two sessions of protected-mode-compatible productivity software (spread sheets, word processors) on an 8MHz 80286 just fine - you "just" need to find such software. Microsoft XENIX would be a suitable operating system to run those programs, and multiple users can use the same computer using serial terminals.

douglar wrote on 2024-03-08, 19:47:

Seems like there was a similar mistake with the 386 and a 32bit bus. Microchannel was less attractive because it had a high cost and you had to junk your old cards to boot. Intel didn't get a 32bit bus that people liked until PCI worked along side ISA. But part of that was timing. A fast 32 bit bus didn't really become imperative until after processors started to > 40 Mhz.

The 32-bit bus of the 80386DX is not just used for I/O cards, but also for RAM access. As soon as you run any kind of 32-bit software that uses aligned 32-bit accesses, this software profits from 32-bit RAM access. Additionally, 32-bit RAM access provides a slight performance boost on fetching instructions, but that boost wasn't worth the price difference between a 16-bit 386SX and a 32-bit 386DX systems. You shouldn't assume that Intel designed the 80386 as "processor for IBM PC compatible systems" only. In systems designed from scratch with a 32-bit architecture in mind (think 32-bit UNIX workstations), the 80386DX architecture made a lot of sence.

Message received . Its often too easy to look at everything from the perspective of a DOS user. Even the hoary 286 had loftier targets than desktop systems when it was designed. Multitasking, or even task switching, multiple DOS environments wasn’t on the feature list when the 286 first came out.

https://en.m.wikipedia.org/wiki/Intel_80286

Intel did not expect personal computers to use the 286. The CPU was designed for multi-user systems with multitasking applications, including communications (such as automated PBXs) and real-time process control.

Intels period correct 286 documentation goes overboard describing the building and requirements / operation of multiprocessor 286 systems .

To Intels horror the industry demanded 100% PC/XT/AT Compatibility with all future CPUs/ systems and no unique and mostly incompatible advancements.

I mounted the IIT 287 on this board and I was also able to compute the Whetstone benchmark using a native FPU:

938.5 Whetstones using CheckIt 3 and 1061 Whetstones using Navratil System Info. This is a results inline with expectations and nearly twice what I obtained on the other board using the same chip! (Obviously the other board was applying an external divider). I wonder if using a 287XL it will be possible to score higher, since the FPU is supposed to have an internal 1.5x multiplier. Of course IF the chip can be made to run at 30MHz...

rmay635703 wrote on 2024-03-09, 19:29:

Intels period correct 286 documentation goes overboard describing the building and requirements / operation of multiprocessor 286 systems .
To Intels horror the industry demanded 100% PC/XT/AT Compatibility with all future CPUs/ systems and no unique and mostly incompatible advancements.

Motorola was able to capture most of the "low-end" workstation market of the '80s with its 68000 and 68020. Performance wise, the 286 is faster than the 68000 and sometimes even the 68020, unless running optimized 32 bit code which can take advantage of the superior architecture of the 68K. In any case, expecting that a 16 bit chip like the 286 would have been used in anything but a faster version of the existing PCs shows that Intel at the time didn't have a clear focus, not to mention that it also killed any potential market the i432 could have, since the 286, despite its limitations, was faster that this one.

Hence my reasoning that Intel should have made the 286 just a faster version of the 8086; that's what the market wanted.

MSxyz wrote on 2024-03-11, 09:32:

Hence my reasoning that Intel should have made the 286 just a faster version of the 8086; that's what the market wanted.

The 286 was just used as a fast 8088 most of it’s existence but that wasn’t what the market really wanted, in fact they were complaining about it already in late 82 which delayed the chips release as Intel fixed extended memory (protected mode) and multitasking support.

The market wanted the 386 address range, virtualization and multitasking capabilities but also didn’t want to pay for it and didn’t have the software to use it effectively most of the time.

Also motto at the time was give me full compatibility or give me death.

rmay635703 wrote on 2024-03-12, 00:06:

MSxyz wrote on 2024-03-11, 09:32:

Hence my reasoning that Intel should have made the 286 just a faster version of the 8086; that's what the market wanted.

The 286 was just used as a fast 8088 most of it’s existence but that wasn’t what the market really wanted, in fact they were complaining about it already in late 82 which delayed the chips release as Intel fixed extended memory (protected mode) and multitasking support.

The market wanted the 386 address range, virtualization and multitasking capabilities but also didn’t want to pay for it and didn’t have the software to use it effectively most of the time.

So Intel wanted to pitch the 286 as a processor for workstations and advanced systems, common folks wanted just a faster personal computer and power users complained that the features were not on the same level as a 68000/020 with some custom MMU and a fancy multi tasking operating system...

Seems nobody really got what it wanted (except maybe common folks who could finally hope to play the games available for PC with less stuttering compared to a 8088/6 with a CGA 😀 )

rmay635703 wrote on 2024-03-12, 00:06:

The 286 was just used as a fast 8088 most of it’s existence but that wasn’t what the market really wanted, in fact they were complaining about it already in late 82 which delayed the chips release as Intel fixed extended memory (protected mode) and multitasking support.

I can't fault the Intel engineers for dreaming big. If I was making a chip, I'd want to make it cool too, if possible. The problem for Intel was that the groups who wanted to build high performance multitasking systems also wanted rational architectures for faster software development. And Intel's plan B worked out pretty good though because it turned out that the secret to having the highest performing chip in the 1990's was simply to have the largest economy of scale.

douglar wrote on 2024-03-12, 17:18:

rmay635703 wrote on 2024-03-12, 00:06:

The 286 was just used as a fast 8088 most of it’s existence but that wasn’t what the market really wanted, in fact they were complaining about it already in late 82 which delayed the chips release as Intel fixed extended memory (protected mode) and multitasking support.

I can't fault the Intel engineers for dreaming big. If I was making a chip, I'd want to make it cool too, if possible. The problem for Intel was that the groups who wanted to build high performance multitasking systems also wanted rational architectures for faster software development. And Intel's plan B worked out pretty good though because it turned out that the secret to having the highest performing chip in the 1990's was simply to have the largest economy of scale.

That and the fact that the IBM PC started as a semi open architecture which spawned a lot of clones right from the start. Compare it to the original Macintosh which, unlike the Apple II, wasn't designed with interoperability and expandability in mind and that was a commercial failure until Apple introduced the Mac II (btw, Jobs wasn't fired from Apple for this failure... he left because he was a person who couldn't take criticism).

I think Intel just got lucky in the beginning being at the right place, at the right time and with a product that, while not spectacular, at least inspired a certain modicum of confidence (cheap to manufacture or outsource, decent all round performance, built on proven technology which the developers were already familiar with).

+1

Macintosh/Jobs vs Apple II/Woz were a bit like Yin and Yang.

Alas, many contemporary Apple 'haters' don't seem to distinguish here anymore.

They don't seem to be aware that historically the Apple II was made by a "good guy"
and was an open platform with a dozen clones around the globe (about 200 ?).

My own sister is no exception here, btw. She really thinks I'm a hopeless Apple fan boy or diehard,
just because I have an interest in vintage Macs and their technology (PhoneNet/AppleTalk was interesting).

Speaking of Apple II, the Apple IIGS was an interesting convergence of both worlds, maybe.
It had the GUI/Networking of a Macintosh, but the Apple II/6502 heritage of the Apple II.
Really fascintating. Didn't know much about this model until "recently".

Btw, about the open slots. I once read that the slots of the Apple II were being inspired by a HP computer system.
So the origin of the open architecture goes further way down in history, maybe.

Last but not least, the Apple Lisa systems were really fascinating from a workstation point of view (and as pricey, as those).
Very slow and sluggish, but also very expandable. There were HDD upgrades, most notably.
It also had MacWorks or MacWorks XL (optional), which allowed the use of Macintosh programs.
The Macintosh XL was a modified Lisa, essentially.

Edit: No offense, though. I didn't meant to go OT. I was more or less thinking out loud here.
In terms of processors, Motorola vs intel, am I right that the 80286 was doing better (performance wise) than an 68000?
Or rather, a typical m68k machine vs an IBM AT ? I vaguely remember reading some numbers in a book. Have to check..

Edit: About the 80286.. I think that the 8080 heritage of both 8086/80286 was also felt welcome by the countless 8080/Z80 developers of the time.
This eased porting of existing CP/M-80 software. The 8-bit i8080 version of MP/M operating system could use a few MBs of memory even.
It had used bank-switching almost a decade before EMS on DOS platform was being around.

Here in Europe in late 1981 or 1982, the first IBM PC with DOS was barely even available yet.
So DOS as a platform (as we know it today) was literally being irrelevant at this point in time, maybe. So no big compatibility concerns.
I mean, DOS existed, in the states, but international software still was all about CP/M platform (numerous clones).

https://en.wikipedia.org/wiki/MP/M
https://en.wikipedia.org/wiki/Sirius_Systems_Technology

Edit: "The IBM 5150 didn't officially launch in the UK until January 1983"
Source: https://nosher.net/archives/computers/adve_045

Jo22 wrote on 2024-03-18, 01:10:

Btw, about the open slots. I once read that the slots of the Apple II were being inspired by a HP computer system.
So the origin of the open architecture goes further way down in history, maybe.

Well, it could be. Wozniak worked briefly at HP in the period where they were introducing their first "desktop calculators" (Apparently at HP they coined this term because labelling them 'computers' evoked the idea of a very large and expensive machine). I have an HP 9825 from 1976 (still working!) and it's a fully modular system: 5 ROM drawers (one for the operating system / language and 4 to expand the capabilities) and 3 slots on the back accessible without having to open the chassis to install various I/O cards. Memory is upgradeable too, with 8KB modules that are as big as a Micro ATX mainboard, but you'll have to open the machine to install them. All cards and ROM cartridges use edge connectors, similar to those still used in todays computers. However HP machines were never 'open' in the sense that HP was the sole supplier of expansion cards.

In terms of processors, Motorola vs intel, am I right that the 80286 was doing better (performance wise) than an 68000?
Or rather, a typical m68k machine vs an IBM AT ? I vaguely remember reading some numbers in a book. Have to check..

Architecturally wise, the 68K CPUs have more internal registers and a more rational instruction set. This translates in being able to do "more" work with a single machine language instruction thus requiring less instructions to perform a certain complex task. This, and the ability to address linearly 4Gb of memory without complex arrangements of segments and pointers, made the 68K the preferred architecture in the 80s. The 68000 itself however, has been designed in the '70s, already pushing the limits of what was possible to manufacture on a single chip of silicon and had some compromises that made it less performing than would have been possible. By contrast, the later 80286, while being gimped by the constraints of the 8086 architecture, was very fast both at executing instructions (16bit hardware division and multiplication took something like 20-24 cycles, compared to 70-90 of the 68000) and also at accessing memory.

When people think of the 68000, they mostly think of the Amiga and its incredible multimedia capabilities. However, these are more the results of its advanced (for their time) coprocessors than the 68000 itself. Actually, the 68000, both in the Amiga and the original Macintosh, is hampered by their unified memory architecture with the image frame buffer being read every 4th clock cycle. This strict arrangement also caused the 68000 to suffer from extra wait states whenever it couldn't access memory in sync with the video, a rather common occurrence since all operations involving 32bit numbers took an extra of 2 cycles on the original 68000, since its ALU was 16 bit only to save transistors!